Feasibility Report on DOLPHIN SEA FOOD RESTAURANT

 Feasibility  Report on DOLPHIN SEA FOOD RESTAURANT

Read More

Vegetable farming in Bangladesh.

Read More

Thesis Related to Finance

   RISK TAKING IN FINANCIAL
          MARKETS:

Acknowledgements
This thesis studies behavioral characteristics of human beings and how they
influence the risk taking decisions of individuals in financial markets. In a broad sense,
any human action involves consequences, which are typically uncertain, and so we are
constantly assuming risks. More than an alternative way to approach financial problems,
we propose a novel view of how people deal with risks. This work wouldn’t be possible
without the help of several people.
First of all, I have to thank my supervisors Juan Ignacio Peña and Benjamin
Tabak. Ignacio believed in my ideas from the very first day and was able to guide me
positively throughout the creation process. Benjamin was also essential, notably in the
scope definition and in the experiment performed in Brazil. I am eternally in debt with
them.
I am also grateful to my professors from the doctoral programme, and
particularly to Luis Gomez-Mejia and Luis Ubeda Rives, who provided me the first
insights, related to the behavioral topic and first motivated me to accept this challenge
to join psychology and finance in the same framework. Professors Jaime Ortega, Mikel
Tapia, Margarita Samartín and Josep Tribó were also of fundamental importance
teaching me the financial and economic tools to address the finance subject.
My friends from the doctoral programme, Sabi, Jordi, Eduardo, Augusto, Kike,
Peter, Kremena, Silviu, Santiago, Rômulo, among others, were also crucial in my
motivation process.
I want to specially thank my wife Luciana for her support and understanding that
a PhD is a joint project for a couple. To love is to share emotions and thanks for sharing
these years with me.
The help of my family was also infinite. I always have to thank them for giving
me the confidence to go for my objectives. My mother Maria Luiza gave me the wings
to fly away and discover the world. Mom, this thesis is for you !
This thesis also benefited from financial support from the Business
Administration Department of Universidad Carlos III de Madrid. With it I was able to
take part in several international conferences in which I could discuss and better define
my ideas. I also could carry through the experiment which is an important part of the
thesis.
Finally I want to thank the Central Bank of Brazil for the opportunity to take part
in its specialization programme. I am sure that my knowledge is going to be helpful in
my daily tasks at the Central Bank. I want to give a special thank to Isabela Maia,
Antônio Francisco, Paulo Cacella, José Renato Ornelas and my other workmates at the
Executive Office for Monetary Policy Integrated Risk Management for several fruitful
discussions which surely enhanced this thesis.
“Behavior Finance is a fascinating area, a course
of self analysis. The more we learn about it, more
we realize that each of us fail in traditional tests of
rationality in an unsuspected way. Von Newmann-
Morgenstern, despite of their brilliant analysis,
omitted relevant pieces of the history.”
Peter Bernstein (1997)
Against the Gods: The Remarkable Story of Risk.
ABSTRACT
Since the innovative work of Kahneman and Tversky (1979), behavioral finance
has become one of the most active areas in financial economics. As compared to
traditional models in this area, behavioral models often have the degree of flexibility
that permits reinterpretation to fit new facts. Unfortunately, this flexibility makes it hard
either to disprove or to validate behavioral models. In the present thesis we try to
overcome this problem, by proposing a general framework based on stylized facts of
human behavior (invariants); and applying it to three financial contexts. Related to the
individual risk taking decision, we focus: on the role of incentives; on how prior
outcomes influence future decision; and on the portfolio choice problem. Different from
traditional models where risk aversion is usually assumed, in our behavioral framework,
the risk preference of the investor varies depending on how he frames his choices. Our
main conclusion is that absolute evaluations based on final wealth are limited and the
relativity of risk taking decisions, where the perception of gains and losses drives the
process, is a requirement to understand individual’s decisions. As a reply to behavioral
critics, we reach propositions that can be falsified and make several predictions.
 Summing up, under the same theoretical framework we make the following
contributions: (1) Shed more light on how incentives affect risk taking behavior,
proposing a model where the reference plays a central role in the previous relationship;
(2) Empirically test our incentives model considering a comprehensive world sample of
equity funds (cross-section data); (3) Evaluate which effect is stronger in multi-period
risk taking decisions for individual investors: house-money or disposition effect,
proposing a novel experiment treatment; (4) Compare survey to experimental results
finding that they are not necessarily aligned. (5) Generalize the myopic loss aversion
behavior performing an experiment in Brazil and Spain; (6) Propose a theoretical model
of how behavioral individuals choose their portfolios in terms of risk taking behavior;
(7) Include estimation error in the behavioral portfolio analysis; (8) Evaluate whether
theoretical models of portfolio choice should adapt or moderate the individual behavior
characteristics. In this case we used a sample of several country equity indices.
Our propositions suggest that managers in passive managed funds tend to be
rewarded without incentive fee and be risk averse. On the other hand, in active managed
funds, whether incentives will reduce or increase the riskiness of the fund will depend
on how hard it is to outperform the benchmark. If the fund is (un)likely to outperform
the benchmark, incentives (increase) reduce the manager’s risk appetite. Furthermore,
the evaluative horizon influences the trader’s risk preferences, in the sense that if traders
performed poorly (well) in a period, they tend to choose riskier (conservative)
investments in the following period given the same evaluative horizon. Empirical
evidence, based on a comprehensive world sample of mutual funds is also presented in
this chapter and gives support to the previous propositions. Related to the house-money
and disposition effect, we employ a survey approach and find evidence of house money
effect. However, in an experiment performed in a dynamic financial setting, we show
that the house money effect disappears, and that disposition is the dominant effect. We
report supportive results related to existence of myopic loss aversion across countries
and some evidence of country effect. Finally, in terms of asset allocation, our results
support the use of our behavioral model (BRATE) as an alternative for defining optimal
asset allocation and posit that a portfolio optimization model may be adapted to the
individual biases implied by prospect theory without efficiency loss. We also explain
why investors keep on holding, or even buy, loosing investments.
RESUMEN
Desde el trabajo innovador de Kahneman y de Tversky (1979), las finanzas del
comportamiento se han convertido en una de las áreas más activas en la economía
financiera. Comparados con los modelos tradicionales en esta área, los modelos del
comportamiento tienen a menudo el grado de flexibilidad que permite su
reinterpretación, ajustándoles a nuevos hechos empíricos. Desafortunadamente, esta
flexibilidad hace difícil refutar o validar modelos del comportamiento. En la actual tesis
intentamos superar este problema, proponiendo un marco general basado en hechos
estilizados del comportamiento humano (invariables); y aplicándolo a tres contextos
financieros. Relacionado con la decisión individual de toma de riesgo, nos enfocamos:
en el papel de los incentivos financieros; en cómo los resultados anteriores influencian
la decisión futura; y en el problema de la gerencia de cartera. Diferente de los modelos
tradicionales donde la aversión al riesgo se asume generalmente, en nuestro marco del
comportamiento, la preferencia del riesgo del inversionista varía dependiendo de cómo
él enmarca sus opciones. Nuestra conclusión principal es que las evaluaciones absolutas
basadas en la riqueza final son limitadas y la relatividad de las decisiones de riesgo,
donde la percepción de las ganancias y de las pérdidas conduce el proceso, son un
requisito para entender las decisiones del individuo. Como contestación a los críticos
del comportamiento, alcanzamos proposiciones que pueden ser contrastadas y hacemos
varias predicciones.
Resumiendo, bajo el mismo marco teórico hacemos las contribuciones
siguientes: (1) Verificamos cómo los incentivos afectan el comportamiento de toma de
riesgo, proponiendo un modelo donde la referencia desempeña un papel central en la
relación anterior; (2) Empíricamente contrastamos nuestro modelo de incentivos en
vista de una muestra mundial representativa de los fondos de acciones; (3) Evaluamos
qué efecto es más fuerte en un estudio dinámico del comportamiento individual de toma
de riesgo: house-money o disposition effect, proponiendo un nuevo tratamiento
experimental; (4) Comparamos los resultados de una encuesta a los experimentales y
verificamos que no están necesariamente alineados. (5) Generalizamos el
comportamiento miope de la aversión de la pérdida (myopic loss aversion) con un
experimento realizado en Brasil y España; (6) Proponemos un modelo teórico de cómo
los individuos sesgados eligen sus carteras en términos de toma de riesgo; (7) Incluimos
el error de la estimación en el análisis de la cartera; (8) Evaluamos si los modelos
teóricos de la selección de carteras deben adaptarse a las características sesgadas de los
individuos o moderar las características individuales del comportamiento. En este caso
utilizamos una muestra de varios índices de acciones de diversos países.
Nuestras proposiciones sugieren que los gerentes en fondos manejados
pasivamente tiendan a ser recompensados sin el honorario variable y sean adversos al
riesgo. Por otra parte, en fondos manejados activamente, si los incentivos reducirán o
aumentarán la toma de riesgo del fondo dependerá de su posibilidad de superar la
rentabilidad de su cartera de referencia. Si es probable que el fondo supere la cartera de
referencia, los incentivos (aumentan) reducen el apetito del riesgo del gerente. Además,
el horizonte evaluativo influencia las preferencias del riesgo del gerente, en el sentido
que si han obtenido mal (buenos) resultados en un período, ellos tienden para elegir
inversiones (conservadoras) más arriesgadas en el período siguiente dado el mismo
horizonte evaluativo. La evidencia empírica presentada en la tesis considerando una
base representativa de fondos de acciones soporta esas predicciones. Relacionado con el
house money y disposition effect, empleamos una encuesta y verificamos la existencia
de house money. Sin embargo, en un experimento demostramos que desaparece el
efecto house money, y que disposition es el efecto dominante. Nuestros resultados
también soportan la existencia de la aversión miope de la pérdida a través de países y
una cierta evidencia del efecto del país. Finalmente, en términos de gerencia de carteras,
nuestros resultados apoyan el uso de nuestro modelo del comportamiento (BRATE)
como alternativa para definir la asignación óptima de los activos y postulamos que un
modelo optimización de cartera se puede adaptar a los sesgos individuales implicados
por la teoría (prospect theory) sin pérdida de la eficacia.
Contents
Chapter One:
General Introduction 01
Chapter Two:
Delegated Portfolio Management and Risk Taking Behavior 08
2.1 Introduction 08
2.2 Literature Review 11
2.3 The Decision Making Model 17
2.3.1 The Case of Passive Funds 23
2.3.2 The Case of Active Funds 25
2.4 Empirical Analysis 35
2.4.1 Data 35
2.4.2 Empirical Results 37
2.5 Conclusions 48
Chapter Three:
House-Money and Disposition Effect Overseas: An
Experimental Approach 50
3.1 Introduction 50
3.2 Theoretical Background 54
3.3 Experiment Design 61
3.4 Experiment Results 65
3.5 Concluding Remarks 74
Chapter Four:
Behavior Finance and Estimation Risk in Stochastic
Portfolio Optimization 75
4.1 Introduction 75
4.2 The Behavioral Model 82
4.2.1 First Period 86
4.2.2 Second Period 99
4.2.3 Multi-Period Analysis 106
4.2.4 Resampling 107
4.3 Empirical Study 109
4.3.1 Data and Implementation 110
4.3.2 Results 112
4.4 Conclusions 117
Chapter Five:
General Conclusions, Contributions and Lines for Further
Research 119
Appendix A: Experiment Instructions (Chapter 3) 123
Appendix B: Proofs of the Value Function Properties (Chapter 4) 135
References 138
1
Chapter One
General Introduction
Since the seminal work of Kahneman and Tversky (1979) two different
approaches are being used to understand and forecast human behavior in terms of the
decision making process, applied to economy and other social sciences: the traditional
rational approach and behavioral theories.
The traditional finance paradigm seeks to understand financial markets using
models in which agents are “rational”. Barberis and Thaler (2003) suggest that
rationality is a very useful and simple assumption, which means that when agents
receive new information, they update their beliefs and preferences instantaneously in a
coherent and normative way such that they are consistent, always choosing alternatives
which maximize their expected utility. Economists have traditionally used the axioms of
expected utility and Bayes’ rule to serve both the normative and descriptive purposes.
The paradigm of individual behavior in traditional finance theory is that of
expected utility maximization, combined with risk aversion. Ever since it was founded
by von Neumann and Morgenstern (1944) the expected utility assumption has been
Chapter One: General Introduction
2
under severe fire as a descriptive1 theory of investor’s behavior. Allais (1953), Ellsberg
(1961) and Kahneman and Tversky (1979) are three prominent examples of this
critique. Traditional finance has also been empirically rejected in explaining several
financial phenomena as the growing behavioral finance literature shows.
Agency theory has its foundations in traditional economic theory considering
stable risk preferences. In this perspective the existence of a separation between
ownership and management in organization creates conflict as some decisions taken by
the agent may be in his own interest and considered not to be maximizing the
principal’s welfare, which is called “moral-hazard”, and it is a consequence of the
information asymmetry between the agent and the principal. An agency relationship has
arisen between two (or more) parties when one, designated as the agent, acts for the
other, designated the principal, in a particular domain of decision problems.
Behavioral finance suggests an alternative approach to expected utility and
agency theory using models in which agents are not fully rational – a theory that is
consistent with the psychology of the investor. The field has two building blocks: limits
to arbitrage, which argues that it can be difficult for rational traders to undo the
dislocations caused by less rational traders; and psychology, which catalogues the kinds
of deviations from full rationality we might expect to see impacting investors’ beliefs
and preferences.
As pointed out by Shefrin (2005), financial economists are in the midst of a
debate about a paradigm shift, from neoclassical-based paradigm to one that is
behaviorally based. In the essence of the debate, there is the unsolved problem of how
individuals make decisions and specifically, how comfortable they are in assuming
risks. This is the main point we want to address in the present thesis: How and in what
1 Normative theories characterize rational choice, while descriptive theories characterize actual choices.
Chapter One: General Introduction
3
situations individuals are more willing to take risks? We also evaluate the practical
consequences of a behavioral risk decision in terms of financial efficiency. Our work
belongs to the field of behavior finance, which has the objective to enhance the
explanatory power of traditional economic models leading to more realistic
psychological foundations.
Behavioral biases can roughly be grouped in two categories: cognitive and
emotional; though both types yield irrational decisions. Because cognitive biases
(heuristics like anchoring, availability and representative biases) stem from faulty
reasoning, better information and advice can often align them with the traditional
rational theory. Conversely, emotional biases such as regret and loss aversion originate
from impulsive feelings or intuition, rather than conscious reasoning and are hardly
possible to be aligned to traditional rationality. Kahneman and Tversky2 (1979) were the
first to propose the use of behavioral lens to evaluate individuals’ decision-making
process. Their prospect theory is a psychologically based theory of choice under risk
and uncertainty.
In general terms, prospect theory or its latter version cumulative prospect theory3
posits four novel concepts in the framework of individuals risk preference, which can be
classified as emotional biases that cannot be eliminated: investors evaluate assets
according to gains and losses and not according to final wealth (mental accounting),
individuals are more averse to losses than they are attracted for gains (loss aversion);
individuals are risk-seeking in the domain of losses and risk averse in the domain of
2 In 2002 their work has been rewarded with the Nobel prize in economics.
3 CPT (Cumulative Prospect Theory) was proposed by Tversky and Kahneman (1992) as an improvement
on, and development from, their earlier Prospect Theory (Kahneman and Tversky, 1979) and is a
combination of Rank Dependent Utility (first proposed by Quiggin, 1982) with reference point to
differentiate between gains and losses.
Chapter One: General Introduction
4
gains (asymmetric risk preference); individuals evaluate extreme probabilities in a sense
of overestimating low probabilities and underestimating high probabilities (probability
weighting function). The theory holds that individuals do not utilise objective
probabilities in their decisions, but rather transform the objective probabilities using
non-linear decision weighting function.
The choice process under prospect theory starts with the editing phase, followed
by the evaluation of the edited prospects and at the end the alternative with the highest
value is chosen. During the editing phase agents code outcomes into gains and losses
and implement mental calculations over the probabilities. In the valuing phase the
agents attach a subjective value to the lottery and then chose the prospect which
generates the highest value.
All previous concepts were already identified in several experiments and we
consider them stylized facts or invariants of human behavior. Under this perspective, we
develop a general theoretical framework to infer the risk taking behavior of human
beings. The main goal of this thesis is to analyze risk taking by individual investors in
financial markets under a behavioral perspective. We approach this topic through three
special cases: the role of incentives in risk taking behavior, risk attitudes in a multi
period analysis; and behavioral risk taking in portfolio choice. The main message we
want to convey is that we live in a behavioral world and we should study financial
markets as such. It doesn’t mean that people are completely irrational or that financial
markets are not efficient. But instead, standard models based on the traditional
paradigm are not capable to fully describe the human nature in terms of decisionmaking.
Standard models of moral hazard predict a negative relationship between risk
and incentives, however empirical studies on mutual funds present mixed results. In
Chapter One: General Introduction
5
Chapter 2 we propose a behavioral principal-agent model to the professional manager’s
context, focusing on the situation of active and passive investment strategy. In this
general framework we evaluate how incentives affect the manager’s risk taking
behavior, where the standard moral hazard model is just a special case.
Our propositions suggest that managers in passive managed funds tend to be
rewarded without incentive fee and be risk averse. On the other hand, in active managed
funds, whether incentives will reduce or increase the riskiness of the fund will depend
on how hard it is to outperform the benchmark. If the fund is (un)likely to outperform
the benchmark, incentives (increase) reduce the manager’s risk appetite. Furthermore,
the evaluative horizon influences the trader’s risk preferences, in the sense that if traders
performed poorly (well) in a period, they tend to choose riskier (conservative)
investments in the following period given the same evaluative horizon. Empirical
evidence, based on a comprehensive world sample of mutual funds is also presented in
this chapter and gives support to the previous propositions.
Recent literature has advocated that risk-taking behavior is influenced by prior
monetary gains and losses. On one hand, after perceiving monetary gains, people are
willing to take more risk. This effect is known as the house-money effect. Another
stream of the literature, based on prospect theory and loss aversion, suggests that people
are risk averse/seeking in the gain/loss domain – disposition effect. Also, behavior
economic research has tended to ignore the role of country differences in financial and
economic decision-making.
The objective of Chapter 3 is twofold: first to clarify which effect is dominant in
a dynamic setting: disposition or house-money; and second to verify the existence of
myopic loss aversion across countries. We employ a survey approach and find evidence
of house money effect. However, in an experiment performed in a dynamic financial
Chapter One: General Introduction
6
setting, we show that the house money effect disappears, and that disposition is the
dominant effect. This finding indicates that care should be taken in generalizing survey
results. We report supportive results related to existence of myopic loss aversion across
countries and some evidence of country effect.
Chapter 4 has two main objectives. The first is to incorporate mental accounting,
loss aversion, asymmetric risk-taking behavior and probability weighting in a multiperiod
portfolio stochastic optimization for individual investors. The previous
behavioral biases have already been identified in the literature; however their overall
impact during the process of determining optimal asset allocation in a multi-period
analysis is still missing. And second, we also take into account the estimation risk in the
analysis. Considering 26 daily index stock data, from the period from 1995 to 2007, we
empirically evaluate our model (BRATE – Behavior Resample Adjusted Technique) to
the traditional Markowitz. Our results support the use of BRATE as an alternative for
defining optimal asset allocation and posit that a portfolio optimization model may be
adapted to the individual biases implied by prospect theory without efficiency loss. We
also explain why investors keep on holding, or even buy, loosing investments.
Finally, in Chapter 5 we provide a brief review of the main findings and
contributions of the thesis and propose several lines for further research. Behavioral
finance is surely an open avenue and we speculate that future paradigms should be
centred on this stream of research.
Our main conclusion is that absolute evaluations based on final wealth are
limited and the relativity of risk taking decisions, where the perception of gains and
losses drives the process, is a requirement to understand individual’s risk decisions. As
a reply to behavioral critics, we reach propositions that can be falsified and make
Chapter One: General Introduction
7
several predictions. Summing up, under the same theoretical framework we make the
following contributions:
- Shed more light on how incentives affect risk taking behavior, proposing a
model where the reference plays a central role in the previous relationship;
- Empirically test our incentives model considering a comprehensive world
sample of equity funds (cross-section data);
- Evaluate which effect is stronger in multi-period risk taking decisions for
individual investors: house-money or disposition effect, proposing a novel experimental
treatment;
- Compare survey to experimental results finding that they are not necessarily
aligned.
- Generalize the myopic loss aversion behavior performing an experiment in
Brazil and Spain, also verifying some country effect;
- Propose a theoretical model of how behavioral individuals choose their
portfolios in terms of risk taking behavior;
- Include estimation error in the behavioral portfolio analysis;
- Evaluate whether theoretical models of portfolio choice should adapt or
moderate the individual behavior characteristics. In this case we used a sample of
several country equity indices.
As a last remark, this thesis was elaborated in a way that any of the following 3
chapters can be read independently. In this sense, Chapters 2, 3 and 4 present complete
researches that, although sharing the behavioral assumptions explained previously,
consider different contexts and lead to independent conclusions.
8
Chapter Two
Delegated Portfolio Management and Risk Taking Behavior
2.1. Introduction
This Chapter deals with a relevant financial phenomenon that occurs in several
markets. There has been tremendous and persistent growth in the prominence of mutual
funds and professional investors over the recent years, which is relevant for both
academics and policy makers (Bank for International Settlements, 2003). Nowadays,
most real world financial market participants are professional portfolio managers
(traders), which means that they are not managing their own money, but rather are
managing money for other people (e.g. pension funds, hedge funds, central banks,
mutual funds, insurance companies). The value of the assets managed by mutual funds
rose from $50 billion in 1977 to $4.5 trillion in 1997. Similarly, the assets managed by
pension plans have grown from around $250 billion in 1977 to 4.2 trillion in 1997
(Cuoco and Kaniel, 2003). Considering only the United States market during the
nineties, assets managed by the hedge fund industry experienced exponential growth;
assets grew from about US$40 billion in the late eighties to over US$650 billion in
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
9
2003. Assets managed by mutual funds exceed those of hedge funds, as total assets
managed by mutual funds are in excess of US$6.5 trillion 4(2003). US equity mutual
funds had total net assets of US$ 4.4 trillion at the end of 2004 (Sensoy, 2006).
The main reasons for the investor to delegate the right of investing their money
to traders include: customer service (including record keeping and the ability to move
money around among funds); low transaction costs; diversification; and professional
management (traders task). Individual investors expect to receive better results, as they
are provided a professional investment service. However, an important stylized fact of
the delegated portfolio management industry is the poor performance of active funds
compared to passive ones (Stracca, 2005). Fernández et al. (2007a,b) found that just 23
of 649 Spanish funds outperformed their benchmarks. Gil-Bazo and Ruiz-Verdú (2007)
found that for active US funds, the ones that charge higher fees often obtained lower
performance. Thus, active management appears to subtract, rather than add value5. A
way to justify the previous empirical evidence is to assume that the delegated portfolio
management context generates an agency feature that has relevant negative
consequences. As investors usually lack specialized knowledge (information
asymmetry), they may evaluate the trader just based on his performance, generating
early liquidation of the trader’s strategy, and can lead to mispricing. This is called the
“separation of capital and brains” (Shleifer and Vishny, 1997). Also, Rabin and
Vayanos (2007), show that investors move assets too often in and out mutual funds, and
exaggerate the value of financial information and expertise.
4 Data provided by HedgeCo.net
5 Fernandez et al. (2007) show that during the last 10 years (1997-2006), the average return of mutual
funds in Spain (2.7%) was smaller than average inflation (2.9%).
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
10
Despite relevant research on incentives produced in both scientific areas,
management and economics, the search for integrative models has been neglected. In
general, management papers usually provide good intuition and interpretation but lack a
more precise methodology and often reach ambiguous results. On the other hand,
economic papers are usually tied to classical rationality assumptions and just capture
one side of an issue. Moreover, standard models of moral hazard predict a negative
relationship between risk and incentives, but empirical work has not confirmed this
prediction (Araújo, Moreira and Tsuchida, 2004).
Building on agency and prospect theory, Wiseman and Gomez-Mejia (1998)
first proposed a behavioral agency model (BAM) of executive risk taking suggesting
that the executive risk propensity varies across and within different forms of
monitoring, and that agents may exhibit risk seeking as well as risk averse behaviors.
However, this study considered only a single period model applied to the case of
company CEOs.
In this Chapter, considering BAM to the professional portfolio manager’s
context, and using the theory of contracts and behavior-inspired utility functions, we
propose an integrative model that aims to explain the risk taking behavior of the traders
with respect to active or passive investment strategies. Our focus is on relative risk
taking measured against a certain benchmark. We argue that BAM can better explain
the situation of professional portfolio managers, elucidating the way incentives in active
or passive investment strategies affect the attitudes of traders towards risk6. Our
6 A portfolio manager decides the scale of the response to an information signal (he also decides the
required effort) and so influences both the level of the risk and the portfolio returns. As pointed out in
Stracca (2005), in a standard agency problem, the agent controls either the return or the variance, but not
both. The previous specific characteristic offers its own challenges as the fact that the agent controls the
effort and can influence risk makes it more difficult for the principal to write optimal contracts.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
11
propositions suggest that managers in passively managed funds tend to be rewarded
without an incentive fee and are risk averse. On the other hand, in actively managed
funds, whether incentives reduce or increase the riskiness of the fund will depend on
how hard is to outperform the benchmark. If the fund is likely to outperform the
benchmark, incentives reduce the manager’s risk appetite, while the opposite is true if
the fund is unlikely to outperform the benchmark. Furthermore, the evaluative horizon
influences the trader’s risk preferences, in the sense that if traders performed poorly in a
period, they tend to choose riskier investments in the following period given the same
evaluative horizon. Conversely, if traders performed well in a given time period, they
tend to choose more conservative investments following that period. We test our
propositions in a world sample of equity mutual funds, finding supportive results.
The remainder of this Chapter is structured as follows. In Section 2, we first
offer a brief literature review. Section 3 describes the professional portfolio manager’s
context and formally presents the model, positing the propositions. Section 4 provides
some empirical evidence supporting the model and Section 5 concludes with a summary
of the main findings.
2.2. Literature Review
The traditional finance paradigm seeks to understand financial markets using
models in which agents are “rational”. Barberis and Thaler (2003) suggest that
rationality is a very useful and simple assumption. This means that when agents receive
new information, they instantaneously update their beliefs and preferences in a coherent
and normative way such that they are consistent, always choosing alternatives which
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
12
maximize their expected utility. Unfortunately, this approach has been empirically
challenged in explaining several financial phenomena, as demonstrated in the growing
behavioral finance literature7. The increase in price of a stock which has been included
in an Index (Harris and Gurel, 1986) and the case of the twin shares which were priced
differently (Barberis and Thaler, 2003) are examples of the empirical market anomalies
found in the literature.
Agency theory has its foundations in traditional economics assuming the
previous “rationality” paradigm. The perspective of a separation between ownership and
management creates conflict as some decisions taken by the agent may be in his own
interest and may not maximize the principal’s welfare (Jensen and Meckling, 1976).
This is known as “moral-hazard”, and it is a consequence of the information asymmetry
between the agent and the principal. We say that an agency relationship has arisen
between two (or more) parties when one, designated as the agent, acts for the other,
designated as the principal, in a particular domain of decision problems (Eisenhardt,
1989).
Related to the main assumptions, agency theory considers that humans are
rationally bound, self-interested and prone to opportunism. It explores the consequences
of power delegation and the costs involved in this context characterized by an agent
which has much more information than the principal about the firm (information
asymmetry). The delegation of decision-making power from the principal to the agent is
problematic in that: (i) the interests of the principal and agent will typically diverge; (ii)
the principal cannot perfectly monitor the actions of the agent without incurring any
costs; and (iii) the principal cannot perfectly monitor and acquire information available
to or possessed by the agent without incurring any costs. If agents could be induced to
7 Allias paradox and Ellsberg paradox.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
13
internalize the principal’s objectives with no associated costs, there would be no place
for agency models (Hart and Homstrom, 1987).
Moreover, while focusing on divergent objectives that principals and agents may
present, agency theory considers principals as risk neutrals in the individual actions of
their firms, because they can diversify their shareholding across different companies.
Formally, principals are assumed to be able to diversify the idiosyncratic risk but they
still bear market risk. On the other hand, since agent employment and income are tied to
one firm, they are considered risk averse in order to diminish the risk they face to their
individual wealth. (Gomez-Mejia and Wiseman, 1997).
Hence, current agency literature considers that principals and agents have
predefined and stable risk preferences and that risk seeking attitudes are irrational.
Highlighting this fact, Grabke-Rundell and Gomez-Mejia (2002) posit that agency
theorists give little consideration to the processes in which individual agents obtain their
preferences and make strategic decisions for their firms. Some empirical studies have
shown that people systematically violate previous risk assumptions when choosing risky
investments, and depending on the situation, risk seeking attitudes may be present. This
occurrence of risk seeking behavior was already identified by several studies related to
choices between negative prospects, and the most prominent of these studies is that of
Kahneman and Tversky (1979) which proposes the prospect theory.
In general, prospect theory8 posits four novel concepts in the framework of
individuals risk preferences: investors evaluate financial alternatives according to gains
and losses and not according to final wealth (mental accounting); individuals are more
averse to losses than they are attracted to gains (loss aversion); individuals are risk
seeking in the domain of losses, and risk averse in the gains domain (asymmetric risk
8 And in its latter version (Kahneman and Tversky, 1992) known as cumulative prospect theory.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
14
preference); and individuals evaluate extreme events in a sense of overestimating low
probabilities and underestimating high probabilities (probability weighting function). In
this Chapter, we consider a behavior inspired utility function, in the framework of
delegated portfolio managers, which takes into account the first three stated concepts.
Coval and Shumway (2005) found strong evidence that CBOT traders were
highly loss-averse, assuming high afternoon risk to recover from morning losses. In an
interesting experiment, Haigh and List (2005) used traders recruited from the CBOT
and found evidence of myopic loss aversion, supporting behavioral concepts. They
conclude that expected utility theory may not model professional trader behavior well,
and this finding lends credence to behavioral economics and finance models as they
relax inherent assumptions used in standard financial economics. Aveni (1989) in a
study about organizational bankruptcy posit that creditors wish to avoid recognizing
losses and thus tend to assume more risk then they would otherwise take.
Wiseman and Gomez-Mejia (1998) argue that prospect and agency theories can
be understood as complementing each other for reaching better predictions of risk
taking by managers. Fernandes et al. (2007), in an analysis of risk factors in forty-one
international stock markets, show that tail risk is a relevant risk factor. We argue that
tail risk can be associated with loss aversion and therefore the BAM offers more fruitful
results in the professional managers’ context.
Now, we will comment on the main criticism received by this approach.
Traditional rational theorists believe that: (i) people, through repetition, will learn their
way out of biases; (ii) experts in a field, such as traders in an investment institution, will
make fewer errors; and (iii) with more powerful incentives, the effects will disappear.
While all these factors can attenuate biases to some extent, there is little evidence that
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
15
they can be completely eliminated9. Thaler (2000) suggests that “homo economicus”
will become a slower learner. In this Chapter, we address the argument of incentives
(iii), showing that in some cases, compensation contracts may even induce risk seeking
attitudes.
As noted by Hart and Holmstrom (1987), underlying each agent model is an
incentive problem caused by some form of asymmetric information. The literature on
incentives and compensation contracts is very extensive, both on theoretical and
empirical studies. Among them there is a consensus about the usefulness of piece-rate
contracts in order to increase productivity10. In our study, we approach the professional
portfolio manager's setting considering a widely used piece-rate contract.
Baker (2000) concludes that most real-world incentive contracts pay people on
the basis of risky and distorted performance measures. This is powerful evidence that
developing riskless and undistorted performance measures is a costly activity. We
extend the previous argument showing that the use of risky performance measures
might be in the interest of companies to induce risk seeking behavior of the agent.
Araujo, Moreira and Tsuchida (2004) discuss the negative relationship between
risk and incentives, predicted by conventional theory but not verified by empirical
9 Behavioral literature suggests two types of biases: cognitive and emotional. Cognitive biases
(representativeness, anchorism, etc) are related to misunderstanding and lack of information about the
prospect, and can be mitigated through learning. On the other hand, emotional biases (loss aversion,
asymmetric risk taking behavior, etc) are human intrinsic reactions and may not be moderated.
10 Lazear (2000a), analyzing a data set for the Safelite Glass Corporation found that productivity
increased by 44% as the company adopted a piece-rate compensation scheme. Bandiera, Barankay and
Ransul (2004) found that productivity is at least 50% higher under piece rates, considering the personnel
data from a UK soft fruit farm for the 2002 season. Lazear (2000b) stresses that the main reason to use
piece-rate contracts is to provide better incentives when the workforce is heterogeneous.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
16
studies. They propose a model with adverse selection followed by a moral hazard,
where the effort and degree of risk aversion is the private information of an agent who
can control the mean and the variance of profits, and conclude that more risk adverse
agents provide more effort in risk reduction.
Palomino and Prat (2002) develop a general model of delegated portfolio
management, where the risk neutral agent can control the riskiness of the portfolio.
They show that the optimal contract is simply a bonus contract. In an empirical study,
Kouwenberg and Ziemba (2004) evaluate incentives and risk taking in hedge funds,
finding that returns of hedge funds with incentive fees are not significantly more risky
than the returns of funds without such a compensation contract.
Our approach is distinguished from the previous approaches as we consider
changes in risk preference of the agents depending on how they frame their optimization
problem rather than assuming risk aversion or risk neutrality from the beginning.
Agents are still considered to be value maximizers, but we are using behavior-inspired
utility functions, based on prospect theory. We also focus on relative risk measured
against a certain benchmark (tracking error), instead of total risk, as this is the relevant
variable of interest for individual investors to decide whether to put their money in
passive or active funds.
The key element to apply prospect theory to our context is to identify what the
trader perceives as a loss or a gain, in other words, to determine what their reference
point should be. In the mutual funds industry, benchmarks are widely used and are
published in their prospects. It is safe to assume the return of the benchmark as the
trader’s reference point. If he can anticipate a negative frame problem, his loss aversion
behavior will lead him to go on riskier actions in order to avoid his losses even if there
are other less risky alternatives which could minimize the loss. This is based on a
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
17
behavioral effect called "escalation of commitment". The intuition is that, due to the
convex shape of the value function in the range of losses, risk seeking behavior will
prevail in the case of prior losses.
Daido and Itoh (2005) propose an agency model with reference-dependent
preferences to explain the Pygmalion effect (if a supervisor thinks her subordinates will
succeed, they are more likely to succeed) and the Galatea effect (if a person thinks he
will succeed, he is more likely to succeed). They show that the agent with high
expectations about his performance can be induced to choose a high effort with lowpowered
incentives. Empirical evidence of the escalation situation can be found in
Odean (1998) and Weber and Camerer (1998). They found that investors sell stocks that
trade above the purchase price (winners) relatively more often than stocks that trade
below the purchase price (losers). Both papers interpreted this behavior as evidence of
decreased risk aversion after a loss and increased risk aversion after a gain.
2.3. The Decision Making Model
We consider professional portfolio managers to be traders who are responsible
for managing the financial resources of others who work for financial institutions such
as: pension funds, mutual funds, insurance companies, banks, and central banks. Their
jobs consist of investing financial resources, selecting assets (e.g. stocks, bonds), and
often using an index as a reference. Despite high competition in financial markets, we
argue that traders, as any human beings, are continuously dealing with their own
emotional biases which make their attitudes toward risk different depending on how
they frame the situation they face.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
18
A characteristic that can affect trader behavior is if the funds they manage have a
passive or active investment strategy. Under active management, securities in the
portfolio and other potential securities are regularly evaluated in order to find specific
investment opportunities. Managers make buy/sell decisions based on current and
projected future performance. This strategy, while tending toward more volatile
earnings and transaction costs, may provide above-average returns. In this case, traders
must be much more specialized because results are directly related to how they choose
among different assets and allocate the resources of the fund in order to obtain better
profits.
On the other hand, in the passive strategy, part or the entire portfolio is settled to
follow a predetermined index, such as the S&P500 or the FTSE100, with the idea of
mimicking market performance (tracking the index). Traders are much more worried
about constructing a portfolio similar to the index than in trying to find investment
opportunities. In this situation, a trader’s activity can be specified in advance as it
consists of allocating the resources closely to a predetermined public index, and then it
is much more programmable and predictable, which raises the possibility for better
control. This strategy requires less administrative costs, tends to avoid under-market
returns and lessens transaction costs. However, because of their commitment to
maintaining an exogenously determined portfolio, managers of these funds generally
retain stocks, regardless of their individual performance.
The approach suggested by Eisenhardt (1985) yields task programmability,
information systems, and uncertainty as determinants of control strategy (outcome or
behavior based). Outcome-based contracts transfer risk from the principal to the agent
and it is viewed as a way of mitigating the agency costs involved. But this rewarding
package has a side effect, as appropriate behaviors can lead to good or bad outcomes. It
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
19
is a very complex problem to isolate the effect of the specific agent’s behavior on the
outcome, especially in businesses with high risk. Contingent pay will be more effective
in motivating agents when outcomes can be controlled or influenced by them. Bloom
and Milkovich (1998) posit that higher levels of business risk not only make it more
difficult for principals to determine what actions agents take, but also make it more
difficult for principals to determine what actions agents should take.
In line with the agency literature (Holmstrom and Milgrom, 1987; 1991), we
model the interaction between a risk neutral, profit maximizer principal and a valuemaximizing
agent in a competitive market. The principal delegates the management of
his funds to the agent, whose efforts can affect the probability distribution of the
portfolio excess return - differential return for a given portfolio, relative to a certain
benchmark11, x x x N( (t), 2 (t)) p b = − → μ σ . The agent's task is related to obtaining
information about expected returns and defining portfolio strategies. The agent chooses
an effort level “t” incurring in a personal cost C(t). We consider the general differential
assumptions for C(t): C’(t) > 0 and C’’(t) > 0. Also, let’s call 0 C the agent’s minimum
cost of effort required to follow a passive strategy and just replicate the benchmark12.
Consider:
2
( )
2
0
C t = C + t (Eq. 01)
And, the portfolio excess return is given by:
x = μ (t) +ε (t) (Eq. 02)
11 xp is the portfolio return and xb is the return of the benchmark.
12 This cost is related to the index tracking activity and can be estimated considering the ETF’s (Exchange
Traded Funds) total management fee.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
20
where μ (t) is concave and increasing, referring to the part of the return due to his level
effort (t). Also take ε(t) ~ N(0, σ2(t)). In order to simplify, we assume that the
performance of the trader has a linear relationship with his efforts plus a random
variable, so that:μ (t) = μt , and then:
x = μt +ε (t) (Eq. 03)
Moreover, the timing of the proposed principal-agent game is: (i) the principal
proposes a contract to the agent; (ii) the agent may or may not accept the contract, and if
he accepts, he receives an amount of funds to invest; (iii) the reference point of the
agent is defined; (iv) the agent chooses the level of effort (related to his personal
investment strategy) to spend; (v) the outcome of the investment is realized and the
principal pays the agent using part of the benefits generated by the chosen strategy and
keeps the remaining return.
In this case the certainty equivalent of the agent’s utility, as proposed in
Holmstrom and Milgrom (1991), can be given by:

where E[w(x)] is the expected wage of the trader, considered as a function of the
information signal (excess return), α is the performance pay factor, and v(x) is the
trader’s value function, which depends on x , the agent’s perceived gain or loss related
to his reference point (benchmark). In the previous model,
w(x) =αx + β =αμt +αε (t) + β , and so Var[w(x)] =α 2σ (t)2 .
The value function was proposed in the prospect theory of Kahneman and
Tversky (1979) and is an adaptation of the standard utility function in the case of the
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
21
behavior approach. The ratio
' ( )
' ' ( )
μ
μ
v
v is the coefficient of absolute risk aversion. For a
risk averse agent, this ratio is negative and the certainty equivalent is less than the
expected value of the gamble as he prefers to reduce uncertainty. This is the origin of
the negative relationship between risk and incentives in moral hazard models.
Let t* denote the agent’s optimal choice of effort, given α. Note that t* is
independent of β. The resulting indirect utility is given by: V (α ,β ) = β + v(α ) , where
( *)
' ( )
' ' ( )
2
( ) ( *) ( *) 1 2 2 t
v x
v α =αμ t −C t + α v x σ is the non-linear term. The marginal utility of
incentives can then be derived:
( *)
' ( )
( *) ' ' ( ) 2 t
v x
v v μ t α v x σ
α α = = +
∂
∂ (Eq. 05)
and if we were considering risk averse agents, it would represent the mean of the excess
profits minus the marginal risk premium.
The effort of the agent leads to an expected benefits function B(t) which accrues
directly to the principal. Let’s consider B(t) = xb + x. The principal’s expected profit
(which equals certainty equivalent as he is risk neutral) is given by:
CE B(t) E[w(x)] p = − (Eq. 06)
Hence the total certainty equivalent (our measure of total surplus) is:
( )
'( )
''( )
2
1
2
2 2
2
0 t
v x
TCE CE CE x x C t v x a p b = + = + − − + α σ (Eq. 07)
The optimal contract is the one that maximizes this total surplus subject to the
agent’s participation constraint (CEa≥0). Adapting the previous model to the
professional manager’s case and considering mental accounting, loss aversion and
asymmetric risk taking behavior, we assume the value function as follows:
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
22
⎩ ⎨ ⎧
− <
− ≥
=
−
, if 0
1 , if 0
( )
e x
e x
v x rx
rx
λ λ
(Eq. 08)
where r is the coefficient of absolute risk preference, λ is the loss aversion factor which
makes the value function steeper in the negative side; and x is the perceived gain or loss,
rather than final states of welfare, as proposed by Kahneman and Tversky (1979). It is
useful to consider the previous form for the value function because of the existence of a
CAPM equilibrium (Giorgi et al., 2004) and because we reach constant coefficients of
risk preference. The following graph indicates v(x) when α = 0.88, λ− = 2.25 and λ+ = 1
(using values suggested by Kahneman and Tversky).
-1,5
-1
-0,5
0
0,5
1
-1,5 -1 -0,5 0 0,5 1 1,5
Figure 1 – Prospect theory value function for α = 0.88, λ− = 2.25 and λ+ = 1
We assume a general symmetric compensation contract applied to the situation
presented in this paper. Starks (1987) shows that the “symmetric” contract, while it does
not necessarily eliminate agency costs, dominates the convex (bonus) contract in
aligning the manager’s interests with those of the investor. Also, Grinblatt and Titman
(1989) posit that penalties for poor performance should be at least as severe as the
rewards for good performance.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
23
w(x) =αx +β (Eq. 09)
This indicates that the agent is paid a base salary β plus an incentive fee
calculated as a proportion α of the total return of the fund (the performance indicator).
The previous contract arrangement follows the optimal compensation scheme defined in
Holmstrom and Milgrom (1987), and was also used in Carpenter (2000). Lazear
(2000b) argues that continuous and variable pay is appropriate in case of worker
heterogeneity as in the case of professional portfolio managers. Finally, let’s call ψ the
probability that the fund outperforms the benchmark and (1 – ψ) the likelihood that it
performs poorly. So, we can re-write the TCE, CEa and CEp as follows:
2.3.1. The Case of Passive Funds
Investors in passive funds have expectations of receiving average market returns
(E(xp) = E(xb) and E(x) = 0), and trader actions are limited and tied in relation to the
process of buying and selling assets to adjust stock weights in the portfolio in order to
follow the benchmark. The agent’s task is more programmable and his behavior is easy
to monitor (“t” is observable by the principal). As the principal has no interest that the
agent goes on riskier strategies than that of the benchmark, he should set α = 0. Thus,
the certainty equivalent of the agent would be given by:
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior

which implies that t* = 0. Optimally, the agent will make no effort to beat the
benchmark. An important aspect considered in this paper is the competitive situation in
the market of professional portfolio managers, which is of crucial importance in
determining who extracts the surplus from the agency contract. We considered, as it is
usual in the delegated portfolio managers' literature, a perfect competition among agents
with the entire surplus accrued to the principal. This situation implies that: CEa = 0 and
0 β = C . The certainty equivalent of the principal would be given by: CEp = xb – C0.
The principal pays the agent a base salary which is equal to the agent’s cost of
effort to ensure the investor receives the return of the benchmark (say the agent's choice
of effort represents the minimum level needed to replicate the benchmark portfolio).
Moreover, if the agent chooses a level of effort different from C0, the performance of
the fund will not be tied to the performance of the benchmark and so σ2(x) > 0
(increased the risk). If the agent just receives the base salary alone, he doesn’t have any
incentive to choose a level of effort different from 0 and so performs in a risk averse
way. Also, because of employment risk, managers tend to decrease risk in order to
prevent potential job loss (Kempf et al, 2007).
In this incentive scheme, there’s no risk premium associated with the agent’s
decisions. Recall that in this case t is observable, and then if the trader chooses t ≠ 0, the
investor will notice and just fire him. Finally, in this case, there is no reason for using
incentive fees, as the trader is not responsible for the earnings of the fund, which should
be equal to the performance of the benchmark. Observe that the previous result is robust
for different levels of risk preference as it is independent of the value function of the
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
25
agent, regardless of whether he is risk averse or risk seeking. Summing up, we can
construct the following table:
Table 1 - Agent’s Choice of Effort in a Passive Fund
Agent’s choice of t Compensation Performance Risk Result
t = 0 w = β x = xb σ2= 0 optimum
t ≠ 0 w = β x ≠ xb σ2> 0 agent is fired
Proposition One: Traders in passively managed funds tend to be rewarded with a base
salary (α = 0).
Proposition Two: Traders in passively managed funds are more likely to perform as
risk adverse agents (t = tp*).
2.3.2. The Case of Active Funds
In the case of an active fund, investors are usually expecting to receive aboveaverage
risk adjusted returns as they consider it linked to the expertise of the traders.
The trader has to make investment decisions, and a great number of these decisions are
based on his own point of view of the market, raising a relevant problem of information
asymmetry (moral hazard). In this case, the first best results are no longer feasible and
outcome-based rewards are often used as part of their contracts and the agent is
stimulated to go on risky alternatives in order to reach above-average returns. Hence,
the idea of the contract is to reduce objective incongruence between the principal
(investor) and agent (trader), and to transfer risk to the agent.
We now examine two cases. In the single task case, the agent’s effort affects
only the mean of the excess return. In the multitask case, the agent’s effort influences
both the expected return and the risk of the portfolio.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
26
a) Single Task
We first analyze the case in which the agent’s effort controls only the mean of
the excess profits and so the risk is exogenous: σ 2 (x) =σ 2 . Consider a loss averse
agent with a value function given by (Eq. 08). The main point in applying prospect
utility is to define the reference point by which the manager measures his gains and
losses. It seems reasonable in the funds industry to assume the returns of the public
benchmark published by the fund as a reference point, since it is the one used by
individual investors when deciding which fund to invest in. Thus, the total certainty
equivalent would be given by:
⎟ ⎟⎠ ⎞
⎜ ⎜⎝
⎛
+ − − + − + ⎟
⎟⎠
⎞
⎜ ⎜⎝
⎛
= + − − − 2 2
2
0
2 2
2
0 2
1
2
(1 )
2
1
2
TCE ψ x μt C t α rσ ψ x μt C t α rλσ b b .
(Eq. 14)
Taking into account the agent’s maximization problem, we reach the following
results:
[ ] ⎟
⎟⎠
⎞
⎜ ⎜⎝
⎛
+ − − + − + ⎟
⎟⎠
⎞
⎜ ⎜⎝
⎛
= + − − − 2 2
2
0
2 2
2
0 2
1
2
(1 )
2
1
2
max CE ψ αμt β C t α rσ ψ αμt β C t α rλσ ta aA
(Eq. 15)
so, t* =αμ and
2
( *) * 0
C t = C + t . As expected, efforts in outperforming the benchmark
increases with incentives. The agent’s marginal utility of incentives is given by:
αμ 2 α σ 2 (ψ (1 ψ )λ ) α v = − r − − (Eq. 16)
So the effect of incentives on the agent’s utility will depend on whether the
benchmark is likely to be outperformed. Suppose that the fund can easily outperform
the benchmark. In this case, the probability that the return of the fund is greater than the
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
27
benchmark, ψ, is close to one and μ > 0 . Then, αμ 2 α σ 2 α v = − r , which is the usual
solution found by moral hazard models. This implies that an increase in incentives has
both positive and negative effects on the utility of the agent. The positive effect results
from the share of the positive excess return, and the negative effect comes from the
increased risk of the wage. Finally, when we maximize the total surplus:
⎟ ⎟⎠
⎞
⎜ ⎜⎝
⎛
+ − − + − + ⎟
⎟⎠
⎞
⎜ ⎜⎝
⎛
= + − − − 2 2
2
0
2 2
2
0 2
1
2
(1 )
2
1
2
max TCE ψ x μt C t α rσ ψ x μt C t α rλσ t b b
(1 ) 0 2 2
− =
−
= − +
∂
∂
μ
ψα σ
μ
ψ αλ σ
μ αμ r r
t
TCE
then
μ σ [ψ ψ λ ]
μ
α
2 2 (1 )
2
+ − −
=
r (Eq. 17)
So the relationship between risk and return is ambiguous, depending on how
likely it is to outperform the benchmark. As previous experiments have shown that the
value for λ is around 2 (Kahneman and Tversky, 1979) if ψ is higher than 67%, then a
negative relationship between risk and incentives is predicted by the model. However,
as we decrease ψ, a positive relation between risk and return appears.
If we consider a benchmark that is easy to be outperformed, then ψ approaches 1
and so
2 2r
2
μ σ
μ
α
+
=
(Eq. 18)
and therefore, increases in σ 2 and r imply decreases in α . The previous negative
relationship between risk (σ2) and incentives (α) is the usual standard result obtained by
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
28
moral hazard models. However our model generalizes this, and the previous result is
simply a special case. If we consider a benchmark that is difficult to outperform, then ψ
approaches 0 and so
μ σ λ
μ
α
2 2r
2
−
=
(Eq. 19)
and, therefore, increases in σ 2 and r imply increases in α . Some empirical papers
have found previous positive relationships between risk and return.
Recall that σ 2 in our model represents a variance in the differential portfolio
which uses the benchmark as its reference (tracking error). The performance of the
benchmark XB and the performance of the chosen portfolio XP are respectively given by:
B B B X E X ε + = ) ( , with ) , 0 ( ~ 2B
B ε N σ
( ) , P P P X E X ε + = with ) , 0 ( ~ 2P
P ε N σ
Also we consider that the expected return of the benchmark is normalized to
zero {E(XB) = 0} and the expected return of the portfolio equals the agent’s choice of
effort {E(XP) = μt}. Therefore, the return of the differential portfolio is given by:
( - ) P B P B t X X ε ε μ + = ⎟
⎟⎠ ⎞
⎜ ⎜⎝
⎛
−
then
( ) ~ N( t, 2 ) P
2
P
2B
μ σ σ ρσ σ P B B X − X + −
so
2 2 2 (1 )
P P
2
P
2B
σ +σ − ρσ σ = σ − ρ B (Eq. 20)
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
29
Finally, consider the simplified assumption that 2
P
2B
σ =σ (i.e. the total risk of the
portfolio selected by the manager is the same as the total risk of the benchmark
portfolio, so based on portfolio theory, both portfolios should be equivalent in terms of
risk/return trade-off), where ρ is the correlation coefficient between the chosen portfolio
and the benchmark. Therefore, α can be rewritten as:
μ σ ρ [ψ ψ λ ]
μ
α
2 2 (1 ) (1 )
P
2
2
+ − − −
=
r (Eq. 21)
which suggests that increases in α imply increases in 2 (1 2 )
P σ − ρ , and also implies a
decreasing correlation (ρ), for low values of ψ. Thus, using the benchmark as a filter
reduces uncontrollable risk by (1 – ρ). If the agent just reproduces the benchmark
(passive strategy), the correlation is equal to 1 (perfect correlation), all risk can be
filtered out, and the first best can be achieved. Because of the agency problem, we see
that the agent’s choice will depend on the degree of idiosyncratic risk associated with
his contract, as measured by 2 (1 )
P σ −ρ . Unlike standard portfolio theory (Markowitz,
1952), idiosyncratic risk will play a role in incentive schemes.
Proposition Three: Traders in actively managed funds tend to be rewarded in
incentive-base pay (α > 0).
Proposition Four: The relationship between incentives and risk can either be positive
or negative depending on the likelihood ψ of outperforming the benchmark. High (low)
values of ψ imply a negative (positive) relationship.
Table 2 - Agent’s Choice of Effort in an Active Fund
Agent’s choice of t Compensation Performance Risk Result
t = 0 w = β x = 0 σ2= 0 agent is fired
t = t* w = αx + β x > 0 σ2> 0 optimum
t = t* w = αx + β x < 0 σ2> 0 agent is fired
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
30
b) Multi Task
We now introduce the possibility that the agent can also influence the risk of the
portfolio’s excess return. Let tμ and tσ be the effort in mean increase and in variance
reduction. We assume the cost is quadratic and separable:
2 2
( )
2 2
0
tμ tσ
C t = C + + . Also,
let μ μ (t) = μt and ( )2
0
2 ( ) σ σ t = σ − t , where the exogenous variance 2
0 σ is the
variance of the excess return when no effort is provided to change it. Taking into
account the agent’s maximization problem, we reach the following results:
[ ] ( )
( ) ⎟
⎟
⎠
⎞
⎜ ⎜
⎝
⎛
+ − + − − − + −
⎟ ⎟
⎠
⎞
⎜ ⎜
⎝
⎛
= + − − − − −
2
0
2
2 2
0
2
0
2
2 2
0
2
1
2 2
(1 )
2
1
2 2
max
σ
μ σ
μ
σ
μ σ
μ
ψ αμ β α λ σ
ψ αμ β α σ
r t
t t
t C
r t
t t
CE t C ta aA
so, αμ μ t* = and
( )
2 ( ) 0
2
*
1 (1 )
(1 ) σ
α ψ ψ λ
α ψ ψ λ
σ + − −
− −
=
r
t r . As expected, efforts to outperform the
benchmark increase with incentives. The endogenous variance is then given by:
( )
2
0
2
2
2
1 (1 )
( ) 1 σ
α ψ ψ λ
σ ⎟ ⎟⎠
⎞
⎜ ⎜⎝
⎛
+ − −
=
r
t (Eq. 22)
which implies that endogenous risk can be lower or greater than exogenous risk
depending on whether the agent is framing a gain or loss situation. If the benchmark is
easily outperformed, ψ approaches 1 and 2
0
2
2
2
1
( ) 1 σ
α
σ ⎟⎠
⎞
⎜⎝
⎛
+
=
r
t , and so endogenous risk
and incentives are negatively related. On the other hand, if the agent is framing a loss
situation, the endogenous risk would be given by 2
0
2
2
2
1
( ) 1 σ
α λ
σ ⎟⎠
⎞
⎜⎝
⎛
−
=
r
t , and a positive
relationship between risk and incentives is predicted.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
31
Summing up, our model predicts that when the fund manager is facing a
situation of high likelihood to outperform the benchmark, he will frame the portfolio
construction problem in the gain domain, and will act in a risk averse way, and
incentives will stimulate him to exert efforts to reduce risk and improve the expected
excess return. Incentives are lower in riskier portfolios. On the other hand, when he is
facing a situation of low likelihood to outperform the benchmark, the agent is likely to
frame the investment problem in the loss domain, and incentives will make him look for
riskier alternatives. Incentives are higher in riskier portfolios.
c) Multi-Period Analysis
In this section, we discuss the effect of previous outcomes in the future risk
appetite of the agent. Wright, Kroll and Elenkov (2002) posit that institutional owners
exerted a significant positive influence on risk taking in the presence of growth
opportunities. Gruber (1996) showed that in the American economy, actively managed
funds assumed greater risk, but reached lower average returns compared to passively
managed funds.
Hence, in some sense, we have the investment strategy and the contract
arrangements disciplining the risk taking behavior of the agent. However we are aware
that the trader’s cognitive biases moderate this relationship. In this study, we do not deal
with the way these biases moderate the relationship as a deeper psychological analysis
of the trader in his context is required, and we also assume that cognitive biases can be
moderated.
Going further in the analysis of the relationship with risk-return, we can apply
Miller and Bromiley´s (1990) multiperiod approach to the professional investor
environment, taking into account the evaluative period. We assume that a company has
a target performance level which for instance corresponds to the performance of a
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
32
chosen index and the firm provides a report annually to the investors. Investors and
traders are likely to consider this target as the reference point for gain/loss analysis.
Supposing that in the first semester, this company performed poorly and so the
likelihood of outperforming the benchmark is lower, the loss aversion of the agent will
make him choose risky projects in the second semester hoping to convert losses into
gains until the end of the year. On the other hand if the company performed well in the
first period, the agent will only accept an increase in risk if the investment opportunity
offers high expected returns. In this case, the trader tends to reduce his relative risk
exposure and follow the index in the second semester in order to guarantee the return
obtained in the previous period. This is based on a behavioral effect called "escalation
of commitment". In other words, if the fund performed well in the first period, the
likelihood of outperforming the benchmark is higher (greater ψ) and the trader is more
likely to perform in a risk averse way (gain domain). Weber and Zuchel (2003) found
that subjects in the "portfolio treatment" take significantly greater risks following a loss
than a gain.
Deephouse and Wiseman (2000) found supportive evidence to these risk-return
relationships in a large sample of US manufacturing firms. Odean (1998) and Weber
and Camerer (1998) provide empirical evidence of the escalation situation; these studies
found that investors sell stocks that trade above the purchase price (winners) relatively
more often than stocks that trade below purchase price (losers). Both works interpreted
this behavior as evidence of decreased risk aversion after a loss and increased risk
aversion after a gain. Chevalier and Ellison (1997) also found supportive empirical
evidence that an agent with a low interim result is tempted to look for high-risk
investments.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
33
Proposition Five: If traders performed (well) poorly in a period, they tend to choose
(less risky) riskier investments in the following period, considering both in the same
evaluative horizon.
Basak et al. (2003) state that as the year-end approaches, when the fund's yearto-
date return is sufficiently high, fund managers set strategies to closely mimic the
benchmark; however they argue that this is because of the convexities in the manager's
objectives. We extend this approach, stating that the previous proposition is a direct
consequence of the individual behavior inspired utility function.
d) Asymmetric Contract
Despite the fact that most mutual funds adopt a symmetric compensation
contract, there are a few which use asymmetric option-based contract as follows:

where γ is usually called the performance fee. If we considered an incentive scheme, as
defined in Eq. 23, the main conclusions of our model would remain with the expression
(21) now given by:

(Eq. 24)
As can be seen, the only effect of γ would be to increase the negative relation
between incentives and risk, in the case of an easy to be outperformed benchmark. If the
benchmark is difficult to outperform, the performance fee has no effect. Probably due to
its diminished effect on risk, the performance fee is not common. Empirical papers
(Kouwenberg and Ziemba, 2004; Golec and Starks, 2002) found mixed results related to
the impact of performance fees on risk taking behavior.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
34
Table 3 provides a summary of the main formulas for α in all the cases
considered. From the model, we can state the following predictions to be tested in the
empirical section of this chapter:
1. Passive funds have lower management fees than active funds13;
2. Asymmetric contracts are less common than symmetric contracts;
3. Active funds which are likely to outperform the benchmark show a negative
relationship between relative risk (tracking error) and incentives.
4. Active funds which are likely to under perform the benchmark will show a
positive relationship between relative risk (tracking error) and incentives;
5. Active funds under performing the benchmark in one given period tend to
increase their relative risk (tracking error) in the subsequent period.
13 We can consider the management fee of a passive fund as a proxy for the β in our compensation
scheme, and in this case, as predicted by the model, the management fee for passive funds should be
lower than for active funds.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
35
Table 3 - Summary of the Equations
Passive Funds α = 0
Active Funds
Symmetric Contract

2.4. Empirical Analysis
2.4.1. Data
For our empirical investigation, we used the Bloomberg cross-sectional equity
mutual funds database for February 2007. There are 4584 funds using 26 different
equity benchmarks (stock indices) from 15 countries14. The database includes emerging
markets (Brazil, Mexico) as well as developed countries (United States, United
Kingdom, Germany)15. As the theoretical model proposed in this paper is always
14 All funds in the database are alive as of January, 2007. Unfortunately data on the dead funds for the
sample used was not available.
15 We also used daily data from January 2002 to February 2007 for 739 funds from France, United States,
Brazil and Japan in a total of 787.216 day-fund-return data from the Bloomberg time series database in
order to validate the results based on the cross-sectional Bloomberg data.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
36
dependent on the reference considered, all the analysis is performed separately for each
benchmark. As in the funds market, the use of a public benchmark is widespread (Elton
et al., 2003), we assume that fund managers tend to be evaluated and compensated using
the benchmark as the reference point, to which gains and losses are defined.
Table 4 provides some descriptive statistics of the funds in the database. The
funds were grouped by the benchmark they use to evaluate their performance, and so we
can consider that they compete for the same class of investors. The number of funds for
each benchmark varies from 32 (Austrian Stock Exchange) to 1332 (S&P500). From the
list of funds, just 261 (5.67%) use performance fees indicating that this sort of
asymmetric compensation contract is not common, except for Brazil (IBOVESPA),
where 18.10% of the funds charge performance fees.
The mean management fee among the entire sample is 1.36% (median 1.25%).
Mexican funds charge the highest management fees (mean 4.84%) and U.S. funds,
which use the Russell 3000 index benchmark, have the lowest management fee (mean
0.68%). The average volatility of management fees is 0.97, highest in Brazil (1.77) and
lowest in Taiwan (0.25). In terms of net asset value (given in the country’s currency),
the mean is usually much higher than the median indicating the concentration of the
market, with few large funds and many small ones. In terms of fund age, we have a
sample of established funds with an average age of 10.55 years and a median ranging
from 5.99 (IBOVESPA) to 16.79 (Germany REX Total Index).
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
37
Table 4 Descriptive Statistics of the Funds’ Data
Descriptive statistics of the funds in the database are displayed. The cross-sectional mean, median and
standard deviation of the management fee (values in %), the net asset value (millions unit in the country’s
currency), and the age (in years) are listed, respectively, for each country. The number of funds for each
country, and those using a performance fee are also displayed.
We have a representative cross country sample of established mutual funds,
diversified in terms of size. The global nature of the data can surely provide good
insight for testing the theoretical model proposed in Section 3 of this chapter.
2.4.2. Empirical Results
In order to test our propositions, we will first distinguish between active and
passive funds. From our predictions, passive funds should have a very small variable
pay factor (Propositions 1 and 3). Typically, funds may charge investors management
fees as a proportion of the total assets value, and performance fees, paid if the return of
the fund outperforms the one obtained by the benchmark. We already observed that
performance fees (asymmetric contracts) are not common. Thus, funds charge the
management fee and use it to compensate the traders and face other operational costs.
As we previously discussed in the theoretical model, management fees and performance
fees act in the same direction in terms of influencing trader behavior, and thus the latter
Perf
Index Mnemonic Index Country OBS Fee Mean Median Std. Mean Median Std. Mean Median Std.
'ASE' Athens Stock Exchange Greece 36 1 1.75 1.50 0.96 106.09 32.41 161.63 10.72 10.08 5.61
'ASX' FTSE All Share Index UK 256 12 1.17 1.25 0.64 272.97 77.12 607.08 13.04 9.28 11.04
'ATX' Austrian Traded Index Austria 32 7 1.28 1.38 0.66 125.11 61.62 155.37 12.35 9.09 8.45
'CAC' CAC 40 Index France 262 9 1.65 1.50 0.70 227.54 59.68 494.75 10.33 8.82 6.89
'CCMP' NASDAQ US (Nasdaq) 40 6 1.59 1.50 0.96 610.58 38.91 1,848.38 9.54 7.23 6.11
'DAX' DAX Germany 106 4 1.22 1.20 0.56 347.06 92.37 742.08 16.52 13.44 11.45
'DJST' Dow Jones US 227 6 1.53 1.50 0.63 100.29 28.95 295.93 7.07 6.45 4.68
'E100' FTSE EuroGroup 100 Europe 38 3 1.34 1.50 0.61 172.12 47.78 711.57 8.77 8.80 2.09
'IBEX' IBEX35 Spain 215 1 1.45 1.45 0.62 78.21 37.47 142.61 9.06 9.51 4.26
'IBOV' IBOVESPA Brazil 337 61 2.09 2.00 1.77 98.87 32.56 187.31 7.00 5.99 6.26
'INDU' Dow Jones Indust US 47 4 1.92 1.50 1.61 237.01 32.25 953.53 7.98 7.67 3.84
'KLCI' Kuala Lumpur Cap Index Malaysia 122 0 1.45 1.50 0.25 145.13 49.51 225.23 11.11 7.42 9.28
'MEXBOL' Mexico Bolsa Index Mexico 80 0 4.84 5.00 0.72 982.81 373.29 1,963.08 10.49 11.52 5.02
'MID' S&P400 Mid Cap US 46 0 0.80 0.75 0.44 1,522.18 140.02 3,499.62 8.24 7.10 5.10
'NKY' NIKKEI 225 Japan 170 8 1.25 1.20 0.69 13,035.77 150.31 50,786.16 10.82 7.98 7.60
'RAY' Russell 3000 Index US 45 1 0.68 0.63 0.51 664.05 256.58 1,121.88 9.09 8.27 6.10
'REX' Germany REX Total Germany 56 2 0.88 0.70 0.55 210.72 67.01 424.07 18.18 16.79 8.93
'RLG' Russell 1000 Growth US 48 0 0.75 0.72 0.40 1,196.50 357.88 2,113.63 17.99 10.47 18.19
'RLV' Russell 1000 Value US 65 0 0.74 0.66 0.53 2,275.07 434.88 6,932.95 13.98 7.59 17.90
'RTY' Russell 2000 US US 182 10 1.06 1.00 0.50 858.81 150.41 3,301.48 10.65 8.92 8.56
'SENSEX' Mumbai Stock Index India 72 0 1.07 1.19 0.26 4,241.40 932.46 7,279.66 9.61 9.04 3.87
'SET' Stock Exchange of Thailand Thailand 126 2 1.29 1.50 0.34 706.07 286.15 1,580.99 8.59 9.55 4.73
'SPX' S&P500 US 1332 98 1.05 0.85 0.75 1,265.58 101.55 7,437.18 11.85 8.61 11.00
'TPX' Topix Index Japan 427 19 1.28 1.48 0.50 14,521.65 1,792.00 60,286.21 8.32 7.02 6.02
'TWSE' Taiwan Stock Exchange China 109 2 1.48 1.60 0.25 1,454.02 897.58 1,565.13 10.58 9.32 3.88
'UKX' UK Index UK 107 5 1.47 1.50 0.63 91.10 13.32 272.76 9.05 8.69 4.70
All 4584 261 1.36 1.25 0.97 10.55 8.36 9.03
Management Fee Total Assets Age
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
38
is not really a requirement. We expect that passive funds charge lower management fees
as they use it to set the trader’s base salary (β), since incentives (α) are not necessary.
Table 5 provides the average mean of the management fee for the funds for each
benchmark, distinguishing between active and passive16 funds. The last column shows
the t statistics and p-values for the differences among means.
Table 5 Passive vs. Active Funds: Management Fee
Management fees for active and passive funds for the various benchmarks considered. Indices with less
than 10 passive funds were excluded from the analysis. The management fee is non-negative by definition
and in this case we run the test on the natural logarithm of the measure in order to improve the normality
of the variable.
Index Passive Active t-stat p-value
'ASX' 0.88 1.18 2.00 0.05
'CAC' 1.39 1.68 3.27 0.00
'DAX' 0.86 1.25 3.29 0.00
'DJST' 1.57 1.52 0.17 0.86
'IBEX' 1.25 1.46 0.39 0.70
'IBOV' 2.07 2.09 0.15 0.88
'INDU' 0.81 2.19 0.41 0.69
'KLCI' 1.33 1.46 1.55 0.12
'MEXBOL' 4.21 4.14 -0.20 0.84
'MID' 0.42 0.88 0.51 0.61
'NKY' 0.78 1.58 4.04 0.00
'RTY' 0.40 1.10 3.91 0.00
'SPX' 0.66 1.08 3.71 0.00
'TPX' 0.56 1.38 7.90 0.00
'UKX' 1.18 1.52 0.04 0.97
All 1.06 1.37 3.57 0.00
Management Fee t-Test
From the results, it can be seen that active funds charged higher management
fees in 14 of the 16 cases and that this difference is significant at the 5% level in 7 of the
16 indices considered. If we consider the entire sample of mutual funds (All), evidence
suggests that active funds charge higher fees. In this sense, propositions 1 and 3 are
given empirical support, and the level of the management fee for passive funds can be
used as a proxy for the base compensation considered in the model. For instance,
considering the benchmark SPX, funds charge 0.66% of the total assets value to pay the
16 Passive funds are the mutual funds classified as Index Funds by Bloomberg.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
39
trader’s base salary and other operational costs, and any increments on the management
fee are used as incentives17.
Consider the implications of proposition 2, which implies that in general, active
managed funds assume a higher risk than passive funds (passive fund managers are risk
averse). Recall, that the risk we are considering in the model is relative to the
benchmark (tracking error). We use the variable (β −1)2 as a proxy of the tracking
error18. Table 6 provides the average mean of the previous risk variable for the funds for
each benchmark, distinguishing between active and passive funds. We considered both
a short term beta calculated over the previous 6 months (using daily data) and a long
term beta calculated over the previous 2 years (using monthly data).
The results indicate that both the short-term and long-term tracking error for
passive funds are lower than for active funds, as predicted by proposition 219. The
difference is statistically significant (5% level) for 22 out of 32 cases. If we consider the
entire sample, the results are similar.
17 In the case of benchmarks DJST (US) and MEXBOL (Mexico), the management fee for passive funds
is higher than for active funds; however the difference is not statistically significant.
18 This proxy is valid if we assume CAPM and the same market portfolio (benchmark) for all funds.
( 1)2 2Bench TE = β − σ . (Carroll et al., 1992)
19 The two indexes where the tracking error was greater for passive funds than for active funds is in the
case of the Dow Jones Industrial Index (INDU) (the difference is not significant) and in the case of
MEXBOL (significant difference).
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
40
Table 6 Passive vs. Active Funds: Tracking Error
The tracking error proxy (β −1)2 for active and passive funds for the various benchmarks considered.
Indices with less than 10 passive funds were excluded from the analysis. The proxy used was nonnegative
by definition, and in this case we run the test on the natural logarithm of the measure in order to
improve the normality of the variable.
Index Passive Active t-stat p-value Passive Active t-stat p-value
'ASX' 0.05 0.12 2.04 0.04 0.03 0.08 1.24 0.22
'CAC' 0.07 0.17 3.23 0.00 0.07 0.16 2.34 0.02
'DAX' 0.08 0.11 1.09 0.28 0.05 0.08 1.48 0.14
'DJST' 0.11 0.15 2.48 0.01 0.07 0.11 1.38 0.17
'IBEX' 0.00 0.12 6.03 0.00 0.00 0.10 5.42 0.00
'IBOV' 0.07 0.13 2.57 0.01 0.00 0.04 3.37 0.00
'INDU' 0.11 0.09 -0.57 0.57 0.22 0.38 -0.69 0.49
'KLCI' 0.02 0.07 2.18 0.03 0.01 0.07 3.17 0.00
'MEXBOL' 0.40 0.26 -2.52 0.01 0.02 0.16 3.93 0.00
'MID' 0.02 0.18 1.92 0.06 0.06 0.26 -0.79 0.44
'NKY' 0.06 0.21 14.28 0.00 0.05 0.12 15.88 0.00
'RTY' 0.05 0.12 10.63 0.00 0.13 0.13 -1.76 0.08
'SPX' 0.08 0.13 7.23 0.00 0.07 0.11 7.73 0.00
'TPX' 0.01 0.09 16.27 0.00 0.00 0.05 15.39 0.00
'UKX' 0.02 0.07 11.15 0.00 0.07 0.14 1.16 0.25
All 0.08 0.13 13.84 0.00 0.05 0.10 19.24 0.00
(Beta6M - 1)^2 t-Test (Beta2Y - 1)^2 t-Test
From our model (proposition 4), the relationship between incentives and risk
depends upon the likelihood of outperforming the benchmark. In order to test this, we
assume that, considering the sample of funds for each benchmark, the group of funds
with better past-performance is more likely to frame a gain situation and so risk and
incentives should have a negative relationship. On the other hand, the group of funds
with the worse past-performance is more likely to frame a loss situation and act in a risk
seeking way (risk and incentives should have a positive relationship). In this sense, we
evaluated both the short term and long term relative risk strategies of the fund
managers.
For the short term strategy we considered the returns in the first half of the year.
The funds classified as winners are those with a previous return in the top 25%
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
41
percentile, and the losers are those with returns in the bottom 25%20. We then
regressed21 the following 6-month tracking error, taking the management fee as the
explanatory variable, as follows22:
( )
i N
for
C C dl mf C dw mf i i i i i i
1..
log 1 0 1 2
2
=
β − = + ⋅ ⋅ + ⋅ ⋅ +ε
where i is the reference the fund, N is the number of funds for a specific benchmark, C0,
C1 and C2 are the regression coefficients, β is the 6-month beta for the second half of
the year, mf is the fund’s management fee, dl is a dummy variable which equals 1 if the
fund is a loser (considering the past 6 month return) and zero otherwise, and dw is a
dummy variable which equals 1 if the fund is a winner (considering the past 6 month
return) and zero otherwise. This is a cross-sectional regression for all funds in each
benchmark index. The results are presented in Table 7.
We can observe that C1 is positive in 24 out of 26 cases and significant in 14
cases. C2 is negative in 12 out of 26 cases and significant in 11 out of 26. Therefore the
data suggest that for loser funds, the relationship between incentives and risk is positive,
and for winners this relationship is negative. The empirical results give some support to
proposition 4, especially in the case of loser funds. However, both C1 and C2 were
significant with the expected signs in only four cases.
20 This definition for the dummy variables will be the same for the remaining regressions. Observe that
they are not complementary as there are funds which are classified neither as losers nor as winners.
21 We used White’s (1980) heteroskedasticity consistent covariance matrix.
22 Our measure of tracking error is non-negative by definition and then we run the regression on the
natural logarithm of the measure in order to improve the normality of the residuals
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
42
Table 7 Active Funds: Short Term Tracking Error
Tracking error proxy (β −1)2 for active funds for the various benchmarks considered. The explanatory
variables are management fee and dummies for past performance which equals 1 if the fund is a loser
(considering the past 6-months return) and zero otherwise, and equals 1 if the fund is a winner
(considering the past 6-months return) and zero otherwise.
Index C0 p-value C1 p-value C2 p-value R2
'ASE' -3.27 0.00 0.00 0.99 -0.26 0.12 0.03
'ASX' -2.76 0.00 0.23 0.08 0.09 0.44 0.01
'ATX' -1.00 0.00 0.06 0.56 0.05 0.62 0.01
'CAC' -4.29 0.00 1.22 0.00 -0.60 0.20 0.14
'CCMP' -2.78 0.00 0.28 0.01 -0.03 0.90 0.05
'DAX' -3.27 0.00 0.75 0.02 0.42 0.15 0.05
'DJST' -3.84 0.00 1.12 0.00 0.65 0.01 0.10
'E100' -2.26 0.00 0.84 0.01 -0.62 0.02 0.25
'IBEX' -4.29 0.00 1.87 0.00 -1.48 0.00 0.34
'IBOV' -2.39 0.00 0.03 0.25 0.16 0.00 0.08
'INDU' -2.25 0.00 0.26 0.00 0.24 0.13 0.06
'KLCI' -3.88 0.00 0.48 0.12 -1.33 0.00 0.19
'MEXBOL' -1.16 0.00 0.05 0.18 0.05 0.08 0.03
'MID' -5.34 0.00 0.83 0.36 -0.97 0.61 0.03
'NKY' -1.69 0.00 0.07 0.39 -0.56 0.05 0.15
'RAY' -4.38 0.00 0.88 0.16 0.28 0.59 0.02
'REX' -0.76 0.00 -0.01 0.65 -0.06 0.45 0.06
'RLG' -5.92 0.00 2.11 0.00 -0.46 0.74 0.19
'RLV' -6.76 0.00 1.73 0.00 1.48 0.03 0.09
'RTY' -4.49 0.00 0.48 0.15 0.30 0.39 0.01
'SENSEX' -4.37 0.00 0.87 0.04 -0.95 0.10 0.15
'SET' -6.08 0.00 -0.12 0.86 0.43 0.59 0.01
'SPX' -4.06 0.00 0.81 0.00 0.63 0.00 0.04
'TPX' -5.96 0.00 1.18 0.00 0.55 0.11 0.05
'TWSE' -6.36 0.00 1.24 0.00 -0.38 0.31 0.11
'UKX' -2.54 0.00 0.47 0.07 0.13 0.68 0.04
For the long term strategy we considered the returns obtained in the first 2 years
of the last 4 years, and classified them as winners and losers, depending on whether the
fund was in the top 25% or in the bottom 25% of funds. We then regressed the tracking
error for the following 2 years taking the management fee as the explanatory variable,
as follows:
( )
i N
for
C C dl mf C dw mf i i i i i i
1..
log 1 0 1 2
2
=
β − = + ⋅ ⋅ + ⋅ ⋅ +ε
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
43
where i refers to the fund, N is the number of funds for a specific benchmark, C0, C1 and
C2 are the regression coefficients, β is the 2 years beta for the last 2 years, mf is the
fund’s management fee, dl is a dummy variable which equals 1 if the fund is a loser
(considering the past 2 years return) and zero otherwise, and dw is a dummy variable
which equals 1 if the fund is a winner (considering the past 2 years return) and zero
otherwise. This is a cross sectional regression for all funds in each benchmark index.
The results are presented in the following Table.
Table 8 Active Funds: Long Term Tracking Error
Tracking error proxy (β −1)2 for active funds for the various benchmarks considered. The explanatory
variables are the management fee and dummy variables for past performance, which equal 1 if the fund is
a loser (considering the past 2 years return) and equal 1 if the fund is a winner (considering the past 2
years return).
Index C0 p-value C1 p-value C2 p-value R2
'ASE' -3.35 0.00 0.13 0.64 -0.33 0.21 0.05
'ASX' -4.44 0.00 0.82 0.00 0.96 0.00 0.13
'ATX' -1.65 0.00 0.47 0.01 -1.53 0.00 0.60
'CAC' -4.55 0.00 1.45 0.00 0.31 0.18 0.30
'CCMP' -4.14 0.00 0.72 0.03 0.28 0.56 0.08
'DAX' -3.70 0.00 1.04 0.00 -0.28 0.33 0.13
'DJST' -3.99 0.00 0.92 0.00 -0.05 0.85 0.10
'E100' -3.22 0.00 1.38 0.00 0.23 0.51 0.17
'IBEX' -4.44 0.00 1.79 0.00 -1.00 0.00 0.30
'IBOV' -5.53 0.00 0.39 0.00 0.00 0.99 0.09
'INDU' -2.53 0.00 0.25 0.09 -0.98 0.21 0.13
'KLCI' -3.99 0.00 0.52 0.09 -0.39 0.28 0.07
'MEXBOL' -2.71 0.00 0.11 0.30 -0.11 0.18 0.06
'MID' -4.53 0.00 -0.38 0.67 0.89 0.53 0.04
'NKY' -2.47 0.00 0.11 0.38 -0.12 0.44 0.02
'RAY' -4.51 0.00 1.49 0.00 -0.75 0.59 0.18
'REX' -0.78 0.00 -0.24 0.16 -0.05 0.34 0.24
'RLG' -6.51 0.00 1.87 0.00 2.23 0.01 0.18
'RLV' -7.16 0.00 2.80 0.00 1.29 0.03 0.14
'RTY' -3.76 0.00 0.09 0.67 -0.15 0.68 0.00
'SENSEX' -5.79 0.00 0.99 0.29 -0.41 0.45 0.05
'SET' -6.56 0.00 1.87 0.00 1.23 0.05 0.21
'SPX' -4.25 0.00 1.20 0.00 0.08 0.69 0.13
'TPX' -5.52 0.00 1.40 0.00 -0.12 0.62 0.10
'TWSE' -4.42 0.00 0.09 0.80 -0.26 0.34 0.01
'UKX' -3.67 0.00 1.04 0.00 -0.25 0.31 0.13
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
44
We may observe that C1 is positive in 24 out of 26 cases and significant in 16
cases, while C2 is negative in 16 out of 26 cases and significant in only 2 out of 26.
Therefore the data suggest that for loser funds, the relationship between incentives and
risk is positive, but for winners the evidence is weaker. The empirical results give some
support to proposition 4, especially in the case of loser funds. However, only in two
cases are both C1 and C2 significant with the expected signs.
If we add to control variables for size and the age23 of the fund the regression as
the following equation:
( )
i N
for
C C dl mf C dw mf C AGE C SIZE i i i i i i i i
1..
log 1 0 1 2 3 4
2
=
β − = + ⋅ ⋅ + ⋅ ⋅ + + +ε
we reach the results presented in Table 9:
C1 is positive in 24 out of 26 cases and significant in 17 cases, while C2 is
negative in 17 out of 26 cases and significant in only 5 out of 26. Results are similar to
Table 8. The empirical results give some support to proposition 4, especially in the case
of loser funds. However, only in three cases are both C1 and C2 significant with the
expected signs. Also, C3 is negative (positive) in 16 (10) out of 26 cases and significant
in 4 (3) cases, so no clear pattern emerges. C4 is negative in 19 out of 26 cases and
significant in 10 out of 26, which suggests that smaller funds have larger tracking
errors.
23 We actually used the natural logarithm of age and size.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
45
Table 9 Active Funds: Long Term Tracking Error
Tracking error proxy (β −1)2 for active funds for the various benchmarks considered. The explanatory
variables are the management fee and the dummy variables for past performance which equal 1 if the
fund is a loser (considering the past 2 years return) and zero otherwise, and equal 1 if the fund is a winner
(considering the past 2 years return) and zero otherwise. Control variables for log(age) and log(size) are
included.
Index C0 p-value C1 p-value C2 p-value C3 p-value C4 p-value R2
'ASE' -3.94 0.00 0.16 0.48 -0.13 0.63 0.94 0.09 -0.45 0.01 0.23
'ASX' -3.73 0.00 0.77 0.00 0.90 0.00 0.21 0.40 -0.27 0.02 0.16
'ATX' -1.63 0.01 0.53 0.03 -1.66 0.00 -0.43 0.22 0.26 0.04 0.66
'CAC' -3.22 0.00 1.33 0.00 0.32 0.16 -0.18 0.60 -0.19 0.06 0.32
'CCMP' -1.67 0.47 0.76 0.04 0.20 0.69 -1.46 0.11 0.18 0.11 0.17
'DAX' -2.26 0.06 0.90 0.00 -0.21 0.52 -0.53 0.19 0.00 0.98 0.15
'DJST' -1.62 0.08 0.82 0.00 -0.19 0.52 -1.01 0.00 -0.03 0.76 0.14
'E100' -7.45 0.01 1.42 0.00 0.34 0.33 2.10 0.05 -0.09 0.73 0.25
'IBEX' -4.47 0.00 1.81 0.00 -0.76 0.01 0.61 0.31 -0.40 0.02 0.33
'IBOV' -3.51 0.00 0.41 0.00 -0.06 0.55 -0.87 0.01 -0.02 0.79 0.13
'INDU' -1.94 0.23 0.26 0.08 -1.02 0.19 0.46 0.55 -0.43 0.29 0.23
'KLCI' -2.94 0.00 0.37 0.28 -0.39 0.29 -0.03 0.91 -0.23 0.01 0.13
'MEXBOL' -1.40 0.10 0.21 0.05 -0.11 0.14 -0.94 0.01 0.16 0.02 0.17
'MID' -5.60 0.00 -0.58 0.52 1.10 0.46 0.91 0.09 -0.16 0.27 0.10
'NKY' -1.22 0.05 0.04 0.73 -0.10 0.56 -0.42 0.09 -0.06 0.46 0.10
'RAY' -3.21 0.14 0.70 0.03 -1.26 0.30 1.23 0.16 -0.70 0.00 0.34
'REX' -1.04 0.00 -0.25 0.10 -0.01 0.61 0.13 0.01 -0.03 0.05 0.33
'RLG' -3.66 0.05 1.46 0.01 1.95 0.01 -0.73 0.26 -0.13 0.51 0.24
'RLV' -5.06 0.01 2.29 0.00 0.91 0.15 -0.01 0.98 -0.31 0.17 0.17
'RTY' -1.27 0.08 -0.08 0.74 -0.27 0.45 -0.75 0.01 -0.12 0.13 0.07
'SENSEX' -5.28 0.04 0.94 0.31 -1.03 0.07 -1.38 0.22 0.40 0.06 0.11
'SET' -3.84 0.39 1.25 0.05 1.91 0.02 -2.28 0.11 0.45 0.36 0.27
'SPX' -2.16 0.00 1.00 0.00 0.06 0.76 -0.27 0.08 -0.28 0.00 0.20
'TPX' -6.61 0.00 1.04 0.00 -0.17 0.51 1.30 0.00 -0.23 0.00 0.17
'TWSE' -5.13 0.01 0.11 0.73 -0.26 0.33 0.09 0.86 0.07 0.71 0.01
'UKX' -0.29 0.84 0.82 0.01 -0.48 0.05 -0.97 0.12 -0.29 0.08 0.26
The previous result sheds light on the problem of relating incentives to risk
taking behavior, indicating that the mixed results found in previous empirical papers are
probably due to a framing problem. Sensoy (2006) found that tracking error is greater,
among funds with stronger incentives, while the agency theory predicts a negative
relationship. The relationship between incentives and risk seems to depend on the
reference, and this result is robust over various financial markets (developed and
emerging markets). The asymmetry in the risk taking behavior is likely to be an
invariant of the decision making process. Another interesting result is that typically the
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
46
coefficient for C1 is higher than that for C2, indicating the loss of aversion in human
behavior24.
Empirical studies of incentives and risk taking in the literature typically test if
funds with poor performance in the first half of the year increase risk in the second half
of the year (Chevalier and Ellison, 1997). In the framework of prospect theory, this will
happen because loss averse managers will always increase risk as their wealth drops
below the threshold, and this effect will be more pronounced for funds with higher fees.
Related to proposition 5, we implemented the following cross-sectional
regression:
( i ) i i i β − = C +C ⋅ dl +C ⋅ dw +ε 0 1 2
log 1 2
, where C0, C1 and C2 are the regression coefficients, β is the 6 month beta for the
second half of the year, dl is a dummy variable which equals 1 if the fund is a loser
(considering the past 6 month return), and dw is a dummy variable which equals 1 if the
fund is a winner (considering the past 6 month return). The results are presented in
Table 10.
Typically, we verify the relationship that a fund increases the risk if it under
performs in the first half of the year and decreases risk if it outperforms. This supports
proposition 5, and is in line with other empirical papers (Elton et al., 2003).
24 The mean value for abs(C1) in Table 7 is 0.69 while that of abs(C2) is 0.51 (the p value for the
difference = 0.05).
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
47
Table 10 Active Funds: Short Term Tracking Error
Tracking error proxy (β −1)2 for active funds for the various benchmarks considered. The explanatory
variables are dummies for past performance which equals 1 if the fund is a loser (considering the past 6
month return) and equals 1 if the fund is a winner (considering the past 6 month return).
Index C0 p-value C1 p-value C2 p-value R2
'ASE' -3.68 0.00 1.21 0.05 0.01 0.96 0.20
'ASX' -2.70 0.00 0.21 0.29 -0.04 0.82 0.01
'ATX' -1.03 0.00 0.16 0.43 0.10 0.59 0.02
'CAC' -4.09 0.00 2.38 0.00 -1.55 0.04 0.17
'CCMP' -2.61 0.00 0.35 0.50 -0.52 0.52 0.04
'DAX' -3.56 0.00 1.66 0.00 0.93 0.04 0.11
'DJST' -3.99 0.00 2.05 0.00 1.14 0.02 0.12
'E100' -2.43 0.00 1.29 0.01 -0.86 0.12 0.26
'IBEX' -4.23 0.00 2.83 0.00 -2.95 0.00 0.44
'IBOV' -2.42 0.00 0.00 0.98 0.52 0.00 0.13
'INDU' -2.34 0.00 0.92 0.27 0.50 0.34 0.05
'KLCI' -3.94 0.00 0.83 0.06 -1.93 0.00 0.19
'MEXBOL' -1.14 0.00 0.15 0.48 0.22 0.10 0.02
'MID' -5.21 0.00 0.48 0.58 -1.09 0.48 0.04
'NKY' -1.61 0.00 0.04 0.77 -1.03 0.02 0.19
'RAY' -4.21 0.00 0.91 0.29 -0.68 0.46 0.05
'REX' -0.76 0.00 -0.03 0.11 -0.07 0.30 0.04
'RLG' -5.76 0.00 2.14 0.00 -1.10 0.24 0.22
'RLV' -7.10 0.00 2.20 0.03 1.46 0.11 0.08
'RTY' -4.26 0.00 0.19 0.67 -0.26 0.59 0.00
'SENSEX' -4.48 0.00 1.27 0.01 -0.94 0.14 0.17
'SET' -6.18 0.00 0.14 0.88 0.72 0.57 0.01
'SPX' -3.66 0.00 0.21 0.28 -0.33 0.11 0.01
'TPX' -6.33 0.00 2.35 0.00 1.44 0.00 0.09
'TWSE' -6.57 0.00 2.33 0.00 -0.38 0.51 0.16
'UKX' -2.43 0.00 0.52 0.27 -0.10 0.85 0.02
Finally, in terms of the use of the performance fee, we already commented that it
is not common in the sample and that less than 6% of funds use this fee. However, for
Brazil (IBOVESPA) and the United States (S&P500), we could observe more funds
using the performance fee. From the model, funds with a performance fee should have a
higher tracking error. We tested this for the previous two indices and found supportive
results25.
25 Funds that use performance fees have a higher tracking error, indicating that the performance fee acts
to magnify the management fee.
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
48
2.5. Conclusions
In this study we applied the Behavior Agent Model to the professional investor
environment, using the theory of contracts, and focused on the situation of active or
passive investment strategies. In a deductive way, we formulated five propositions
linking investment strategy, compensation and risk taking in a professional investor’s
context.
Our propositions suggest that managers in passively managed funds tend to be
rewarded without incentive fee and are risk averse. On the other hand, in actively
managed funds, whether incentives that reduce or increase the riskiness of the fund
depends on how hard it is to outperform the benchmark. If the fund is likely to
outperform the benchmark, incentives reduce the manager’s risk appetite; conversely, if
the benchmark is unlikely to be outperformed, incentives increase the manager’s risk
appetite. Furthermore, the evaluative horizon influences the trader’s risk preferences, in
the sense that if traders performed poorly in a period, they tend to choose riskier
investments in the following period given the same evaluative horizon. On the other
hand, more conservative investments are chosen after a period of good performance by
a trader. To the best of our knowledge, this is the first time these results have been
illustrated in the literature using a behavioral framework.
We tested the model in an empirical analysis over a large world sample of
mutual equity funds, including developed and emerging markets, and we reached
supportive results to the propositions established in the theoretical model.
Further extensions of this work may include the type of financial institution the
trader works for (banks, insurance companies, pension funds) to take into account
Chapter Two: Delegated Portfolio Management and Risk Taking Behavior
49
regulatory and institutional effects. Also, delegated portfolio management often
involves more than one agency layer and future work could examine how this feature
affects incentives? More generally, studies about general equilibrium implications and
price impact should be interesting, especially for policy-makers, given the relevance of
these funds in all developed financial markets. Also, the consideration of other contract
schemes should be of interest (Sundaram and Yermack, 2006, suggest the use of debt
contracts). Indeed, Manthei and Mohnen (2004) show that factors other than the
performance-dependent part of the compensation influence an individual’s effort
decision. Their experimental data show significantly higher effort levels for very low or
very high fixed payments.
Despite the fact that we use prospect theory assumptions, the impact of cognitive
biases in an agent’s risk preference still needs to be better understood in order to
understand the way psychological states may affect risk preferences in this context. We
applied BAM specifically to the trader’s situation. An extension of this study to other
institutional contexts would be interesting in order to find some external validity to the
propositions settled. Also, in the compensation analysis, only financial compensations
were considered, and we think that including non-financial rewards like recognition and
prestige would enrich the theory and enable better predictions. Finally, inclusion of
career concerns in the model could also improve multi-period analysis. Kempf et al.
(2007) suggest that when employment risk is high, managers that lag behind tend to
decrease risk relative to leading managers in order to prevent potential job loss. All of
these aspects are left for future research.
50
Chapter Three
House-Money and Disposition Effect Overseas: An
Experimental Approach
3.1. Introduction
Behavioral concepts have been used to provide explanations for several financial
market inefficiencies found in empirical studies. Behavioral economists argue that the
prospect of losses seems more damaging than looking at the entire wealth, even if the
average outcome is the same. This sort of myopia in the face of losses may explain
much of the irrationality some agents display in financial markets.
In this sense, several empirical studies (Tversky and Kahneman, 1992) have
shown that when dealing with gains, agents are risk averse, but when choices involve
losses, agents are risk seeking. Moreover in a wide variety of domains, people are
significantly more averse to losses than they are attracted to same-sized gains (Rabin,
Chapter Three: House-Money and Disposition Effect Overseas
51
1998). Loss aversion is a relevant psychological concept that has been imported to
financial and economic analysis, which represents the foundation of prospect theory26.
In the finance literature, there are conflicting results about the effect of prior
outcomes in future risky choices. Thaler and Johnson (1990) show that when faced with
sequential gambles, people are more risk-taking if they earned money on prior gambles
than if they lost. The intuition is that losses are less painful to people when they occur
after a gain rather than when losses happen after a previous loss. This effect is known in
the financial literature as the house-money effect. Gneezy et al. (2003) investigate the
house-money effect in a market experiment, and find a significant positive effect of
lagged profits on expenditures on assets.
On the other hand, another stream of the literature based on individuals’ loss
aversion suggests that subjects take significantly greater risk following a loss than a
gain (disposition effect)27. Weber and Zuchel (2003) provided empirical evidence of the
previous risk behavior in his portfolio treatment. In chapter two, in a theoretical study
about traders' compensation, we proposed that if traders performed poorly in a period,
they tend to choose riskier investments the following period. However, considering the
house-money effect, traders who have earned profits in the first period that exceed some
benchmark level would become less risk averse in the following period, because they
would feel that they were gambling with the house-money (Coval and Shumway, 2005).
Benartzi and Thaler (1995) proposed another behavioral bias— myopic loss
aversion (MLA) as an explanation for the equity premium puzzle (Mehra and Prescott,
1985). MLA combines two behavioral concepts, loss aversion and mental accounting,
26 Coval and Shumway (2005) found strong evidence that traders from the Chicago Board of Trade
(CBOT) appear highly loss-averse, assuming high afternoon risk to recover from morning losses.
27 The increase in risk taking following a loss is also referred as “escalation of commitment”.
Chapter Three: House-Money and Disposition Effect Overseas
52
where the latter is related to how individuals employ implicit methods to evaluate the
consequence of their decisions. Myopic Loss Aversion is related to the behavior of
individuals to evaluate prospects considering a shorter horizon than it should be given
his original investment horizon. The evaluation period tends to fit to the feedback
frequency of the financial strategy.
Gneezy and Potters (1997), Thaler et al. (1997), Gneezy et al. (2003) and Haigh
and List (2005) have provided experimental evidence supporting MLA28. In their
experiments, they change the feedback frequency of investment decisions and find that
agents tend to invest more when the performance of their decisions is assessed more
infrequently.
Another relevant aspect addressed in this paper is that behavioral economic
research has tended to ignore the role of country differences in financial and economic
decision-making, usually making universalistic assumptions of human behavior. As
pointed out by Levinson and Peng (2007), despite the progress of behavioral economics
and its increasing importance in explaining international financial markets, behavior
theory has failed to answer how do systematic cultural and social differences affect
economic, financial, and legal decision-making. Under cultural psychology, people
perceive life through different lenses and interact with diverse environment. It’s
expected that this fact might affect individuals’ decision making process.
The objective of this Chapter is twofold: first to verify the existence of myopic
loss aversion across countries; and second to clarify which effect is dominant in a
28 The reader is also referred to Langer and Weber (2005), who extend the concept of myopic loss
aversion to myopic prospect theory, predicting that for specific risk profiles, where the chance of winning
is much high, myopia will not decrease, but increase the attractiveness of a sequence.
Chapter Three: House-Money and Disposition Effect Overseas
53
dynamic setting: loss aversion or house-money. We test the house-money effect
following Thaler and Johnson (1990), and find strong evidence of it. We then modify
the experiment conducted in Gneezy and Potters (1997) and Haigh and List (2005) to
test which effect is dominant: the house-money effect or the disposition effect. This
experiment was conducted on students in Brazil and Spain, and results suggest that loss
aversion dominates the house-money effect. Therefore, in an experiment that replicates
the dynamics of decision making, the house-money effect disappears. We also report
supportive results related to existence of myopic loss aversion (MLA) across countries
(Brazil and Spain).
The chosen countries, despite their economic differences, have similar cultural
characteristics. Gouveia and Clemente (2000) have shown that Brazil and Spain are
similar in terms of the level of individualism-collectivism, with Spanish people been
slightly more individualists.
In a framework close to ours, Weber and Zuchel (2003) investigate the influence
of prior outcomes on risk attitude. They found that the prospect presentation might
influence the risk attitude of subjects after prior outcomes. They observed that in the
“portfolio treatment”, subjects take significantly greater risk following a loss than a
gain. On the other hand, in the “lottery treatment,” there was greater risk taking after a
gain than after a loss. However, in their experiment, subjects participated in only one
sequence of two periods, so they were not given the opportunity to learn by
participating in several rounds. This could influence the perception of the subjects about
the likelihood of the outcomes (law of small numbers). Also, their within-subject design
aiming to get the impact of prior outcomes is unnatural and unrealistic. There’s clearly a
difference in asking for the behavior of a subject in the face of gains or losses; and the
Chapter Three: House-Money and Disposition Effect Overseas
54
direct observation of his attitude facing risky dynamic decisions. Our experimental
design overcomes these previous limitations.
The remainder of this Chapter proceeds as follows. In the next section, we
provide brief theoretical background of loss aversion, MLA and house-money effects.
In section 3 we describe our experimental design. In section 4 we present the results and
section 5 concludes the chapter, highlighting the main findings.
3.2. Theoretical Background
Benartzi and Thaler (1995) were the first to propose the myopic loss aversion
(MLA) concept to elucidate the equity premium puzzle raised by Mehra and Prescott
(1985). This puzzle refers to the fact that despite stocks have provided a more favorable
risk-return profile; investors were still willing to buy bonds. Benartzi and Thaler took
advantage of two behavioral concepts to solve the puzzle: loss aversion and mental
accounting. They included the impact of anticipated emotions on decision-making.
Anticipated emotions are those that decision-makers expect to experience given a
certain outcome. They demonstrated that the size of the equity premium is consistent
with the investors evaluating their investments annually, and weighting losses about
twice as heavily as gains.
The approach of Benartzi and Thaler (1995) was not direct (experimental)
evidence of MLA. However, Gneezy and Potters (1997) conducted an experiment that
produced evidence to support the behavioral hypothesis of myopic loss aversion. Haigh
and List (2005), in another experiment with undergraduate students and traders from the
CBOT, in a design similar to the one proposed by Gneezy and Potters (1997), found
evidence of myopic loss aversion (MLA), and that traders exhibit behavior consistent
Chapter Three: House-Money and Disposition Effect Overseas
55
with MLA to a greater extent than students. In the experiment, they evaluated the
participation of the agents in a certain lottery, changing the frequency of information
provided to them. Bellamare et al. (2005), in another experiment similar to Gneezy and
Potters (1997), distinguish the effects of information feedback and investment flexibility
over the myopic loss aversion. They found that varying the amount of information,
alone, suffices to induce the behavior that is in line with the myopic loss aversion.
However, Langer and Weber (2003) found a strong MLA effect depending on
the length of commitment, a much less pronounced effect of feedback, and a strong
interaction between both variables. The effect of the feedback frequency was reversed
for a long binding period. Huang and Liu (2007) suggest that an investor with a higher
risk aversion or a longer investment horizon chooses less frequent but more accurate
periodic news updates. Charness and Gneezy (2003) found that myopic loss aversion
treatment participants were willing to pay money to have more freedom to choose, even
though (in line with the documented bias) they invested less when having more freedom
to change their investment.
There is ample evidence that feelings do significantly influence decisionmaking,
especially when the decision involves conditions of risk or uncertainty (see
Lucey and Dowling (2005) for a recent review). The consideration of risk as feeling can
be presented as in the figure 1. The idea is to incorporate the fact that the emotions
people experience at the time of making a decision influence their eventual decision.
Edmans et al. (2007) find significant market decline after soccer losses, which is
assumed to affect negatively the investors’ mood. Supported by psychologist’s research,
the risk as feeling model is based on three premises: cognitive evaluations induce
emotional reactions; emotions inform cognitive evaluations; and feelings can affect
behavior.
Chapter Three: House-Money and Disposition Effect Overseas
56
In our study, we extend the previous model, by considering a multi-period
analysis, suggesting a feedback process in which the experienced outcomes
(experienced utility), whether they were positive or negative, will influence the
following decision-making process (decision utility) inducing a different risk taking
behavior. In other words, realized outcomes have an impact on the anticipated
outcomes, emotions and subject probabilities; which will deviate the risk taking
decision of the subject from what is predicted by standard utility functions.
Figure 1
Risk as Feelings framework
Note: Adapted from Lucey and Dowling, 2005.
The reasons why decision utility and experienced utility fail to coincide are
related to the following empirical facts: the financial decisions are cognitively hard to
make; people usually have poor forecasts of preference dynamics (failure to anticipate
adaptation and visceral effects); and individuals have inaccurate memories of past
hedonic experiences.
Anticipated
outcomes or
emotions
Subjective
probabilities
Extra factors
(mood,
vividness, etc.)
Cognitive
evaluation
Feelings
Decisions Outcomes
Chapter Three: House-Money and Disposition Effect Overseas
57
Assume the following example to highlight the influence of MLA in the
individual decision making process. Suppose an agent is loss averse, and weights losses
relative to gains at a rate of λ > 1. Consider also that the agent is evaluating the
possibility of taking part in a lottery, in which he has probability p of losing $L, and
probability (1 – p) of winning $W. So the expected utility of the lottery is
( ) (1 )( ) ( ) 1 E U = − p W +λp −L , and the agent should accept the gamble if E(U1) > 0 or
( )
(1 )( )
p L
− p W
λ < . Now, if the agent is evaluating the possibility of participating in an nround
lottery, each with the same payoff as the previous one, the expected utility is now
given by:
E(U ) nW(1 p)n [(n 1)W L](1 p)n 1 p ... ( p)n ( nL)
n = − + − −λ − − + +λ − . (Eq. 01)
For instance, for values n =1, L= 1, p= 2/3 and W =2.5, the agent should accept the
gamble if λ < 1.25. When n =3 the agent should accept the gamble if λ < 1.56. As the
number of rounds in the lottery increases, the gamble becomes more attractive to
increasingly loss-averse agents.
Translating the previous framework to the financial market, if agents evaluate
their investments more often, there will be more occasions when the risky asset (stocks)
underperforms the safe asset (bonds). As individuals are loss averse, those occasions
would generate to them a greater dissatisfaction, so they will tend to avoid the risky
asset. Conversely, if agents evaluate their investments over longer time periods, the
risky asset will rarely under perform the safe asset, and the investors will face losses
more seldom. In this case, the loss averse agent will be more comfortable taking more
risk.
Hypothesis 1(MLA): Within a dynamic environment, decision makers who receive
feedback more (less) often will take less (more) risk.
Chapter Three: House-Money and Disposition Effect Overseas
58
The house-money effect, proposed in Thaler and Johnson (1990), suggests that
the risk tolerance of an agent increases as his wealth increases as a consequence of
previous gains. To explain this effect, they developed Quasi-Hedonic Editing (QHE)
Theory, which considers that decision makers segregate recent gains from their initial
position, but they do not segregate recent losses. Because they have segregated their
prior winnings from their initial wealth, future losses will appear less damaging, as they
are losing someone else's money (the house-money). Slattery and Ganster (2002) found
that decision maker who had failed to reach their goals, set lower, less risky goals in
subsequent decisions.
On the other hand, disposition effect would predict that traders with profitable
mornings would reduce their exposure to afternoon risk trying to avoid losses and so
guaranteeing the previous gains. One major explanation for this effect is based on the
value function and loss aversion from prospect theory. When subjects use the wealth
they bring to the experiment plus any initial endowment in the experiment as reference
point, any loss in the experiment will be perceived as a loss, not as a reduced gain as in
the hedonic-editing case. The intuition is that, due to the convex shape of the value
function in the range of losses, risk-seeking behavior will prevail in the case of prior
losses. As empirical evidence of the previous effect, Odean (1998) and Weber and
Camerer (1998) have shown that investors are more willing to sell stocks that trade
above the purchase price (winners) than stocks that trade below purchase price (losers)
– disposition effect. Both works interpreted this behavior as evidence of decreased risk
aversion after a loss, and increased risk aversion after a gain.
It’s important to highlight that myopic loss aversion describes the bias when
evaluating a prospect before taking an investment decision, while house-money and loss
Chapter Three: House-Money and Disposition Effect Overseas
59
aversion (as considered here in a dynamic setting) describe behavioral biases after
persons experienced gains or losses. We evaluate both situations in this work.
Ackert et al. (2006) pointed out the previous contradiction between loss aversion
and house-money, however they argue that prospect theory was developed for one-shot
games, and so it cannot be applied to a multi-period setting. They provide the results of
a multi-period experiment that gives evidence of house-money instead of disposition
effect. Nevertheless, the way they test house-money is by changing the initial amount
given to the subjects, and so the change in the individual wealth is not a result of the
individual’s choices. We argue here, based on the risk as feeling ideas, that an
experienced outcome will impact future individual’s decisions, and so we could
perfectly observe the disposition effect. In this study, we want to address the influence
of previous losses (or gains) in future risk-taking decisions. The following two
competing hypotheses can be formulated, based on disposition and on the house-money
effect, respectively:
Hypothesis 2a(DE): Within a dynamic environment, decision makers who receive
negative (positive) feedback in a period will take more (less) risk in the following
period.
Hypothesis 2b(HM): Within a dynamic environment, decision makers who receive
negative (positive) feedback in a period will take less (more) risk in the following
period.
As pointed out by Brennan (2001), the house-money effect is just one of at least
three behavioral stories about how an investor will respond to good or bad news. Biased
self-attribution causes the investor to become more confident as his previous positive
assessment of the investment is confirmed, and therefore, he will be willing to take
more risk; an effect similar to the one proposed by the house-money effect. On the other
Chapter Three: House-Money and Disposition Effect Overseas
60
hand, the disposition effect suggests that investors become more risk averse as their
stocks increase in value, in line with the loss aversion prediction.
Related to MLA, our experimental design is similar to the experiment proposed
in Gneezy and Potters (1997) and Haigh and List (2005). On the other hand, in order to
test which effect is dominant, loss aversion or house-money, we introduce a novel
treatment allowing participants to accumulate previous gains. We explain our
experiment in depth in the next section.
In terms of the cross-country analysis, based on psychological theories, it is
quite plausible that fundamental differences in how people perceive the world might
predict fundamental differences in how people make financial estimations, economic
decisions, and exhibit cognitive biases. One prominent theory has helped social
psychologists explain systematic cultural differences: the individualism/collectivism
model (Triandis, 1995). In general terms, the individualism conveys two aspects:
segregation of the individual from the social group which he participates; and selfconfidence.
The same happens to collectivism: integration with the social group and
interdependence.
Previous psychological studies have related the level of individualism to age
(Hui and Yee, 1994), gender (Cha, 1994), educational level (Han and Choe, 1994),
economic level (Freeman, 1997) and economic independency (Kagitçibasi, 1994). Age
is considered one of the main factors that explain the level of
individualism/collectivism, in the sense that elder people present a tendency towards
collectivism. In terms of gender, women are more social and centred in the family,
when compared to men, and this might explain their higher level of collectivism. The
educational level has also being shown to influence the level of collectivism. Previous
researches have described that one with lower educational level is more likely to present
Chapter Three: House-Money and Disposition Effect Overseas
61
friendship actions and favorable attitudes towards a member of his social group.
Conversely, individuals with a higher educational level usually adopt individualistic
values. Finally, a collectivism (individualism) model is produced when an economic or
emotional dependency (independency) is observed.
The conclusion of previous social psychological studies is that someone who is
young, man, has a high educational and economic level, and is economic independent,
tends to be more individualist. Individualism is related to behavioral biases like
optimism and overconfidence, which leads to less aversion to risk.
The chosen countries, despite their economic differences, have similar cultural
characteristics. Gouveia and Clemente (2000) have shown that Brazil and Spain are
similar in terms of the level of individualism-collectivism, with Spanish people been
slightly more individualists. The previous difference is probably due to the Spanish
higher economic level and economic independency. So we would expect Spanish
subjects to be more comfortable in assuming more risks. Also young and men subjects
are also supposed to be less risk averse.
3.3. Experiment Design
We have two behavioral biases to investigate: the MLA and the house-money
(against disposition) effect. To study the MLA effect we followed closely the
experimental design proposed by Haigh and List (2005). We used a straightforward 2 X
2 experimental design (see Table 1). With this setting, it’s not only possible to verify the
existence of MLA, but also to investigate if there’s any country effect. Both groups of
students were selected from undergraduate courses of Management and Economics.
Chapter Three: House-Money and Disposition Effect Overseas
62
They were evaluated in two distinct treatments: Treatment F (denoting frequent
feedback) and Treatment I (denoting infrequent feedback).
Table 1 - Experimental Design (MLA)
Subject Type Treatment F Treatment I
Students from Brazil 24 28
Students from Spain 32 32
Note: Number of students in each treatment (F: frequent feedback; I: infrequent feedback) grouped by
country (Brazil and Spain).
In Treatment F, our control group, subjects took part in a twelve-round lottery.
In each round, they were endowed with 100 units29 and had to decide how much to bet
in a lottery, which pays two and a half times the amount invested with one-third of
probability, and zero with two-thirds of probability. The subjects were previously
informed about the probabilities, payoffs and mechanism of the lottery, and so they
were aware that they could win from zero to 35030 in each round, depending on the
amount invested. At the end of each round they were informed about the lottery results
and had to decide the next round bet. The experimental instructions for students in each
treatment are given in the Appendix A. After the twelve rounds, all individual results
were summed and the final amount was paid.
In opposite of Treatment F, we used the Treatment I (infrequent feedback),
which is similar to the frequent treatment, except that subjects place their bets in blocks
of three. Agents decide just before round t, how much to invest in rounds t, t +1 and t +
29 For subjects in Brazil, 1 unit represents 1 cent of Brazilian Real, and for students in Spain, 1 unit
represents 1 cent of Euro.
30 If the subject invests 100 and wins, his total earnings would be 100 x 2.5 = 250 + 100 (his initial
endowment) = 350.
Chapter Three: House-Money and Disposition Effect Overseas
63
2. Following the suggestion made in the literature, we restricted the bets to be
homogeneous across the three rounds. After all the subjects placed their blocks of bets,
they were informed about the combined results.
Based on the MLA effect, it’s expected that students in the Treatment F will
invest less in the lottery, compared to those in Treatment I, as they face losses more
often and are loss averse. Subjects in Treatment I face losses less often, and so should
be willing to take more risk.
In order to check the house-money effect against loss aversion, we again used a
2 x 2 experimental design summarized in Table 2. With this setting it’s not only
possible to verify the existence of loss aversion or house-money, but also to investigate
if there’s any country effect. As in the previous setting, both groups of students were
selected from undergraduate courses of Management and Economics. They were
evaluated in two distinct treatments: Treatment IR (Isolated Results) and Treatment C
(Cumulative Results).
Table 2 - Experimental Design (House-Money Effect)
Subject Type Treatment IR Treatment C
Students from Brazil 24 33
Students from Spain 32 33
Note: Number of students in each treatment (IR: isolated results; C: cumulative results) grouped by
country (Brazil and Spain).
In fact, Treatment IR is the same as Treatment F and so the group is the same
(control group). In opposite of Treatment IR, we used Treatment C (cumulative results),
which is similar to the IR treatment, except that the amount available to invest in each
pair round (t = 2, 4, 6, 8, 10 and 12) was the result individually obtained in the previous
round (t = 1, 3, 5, 7, 9 and 11) respectively, plus the 100 units corresponding to that
Chapter Three: House-Money and Disposition Effect Overseas
64
round. In this setting, subjects tend to place their reference point in the beginning of
each two rounds, in opposite to Treatment F where each rounds’ endowment represent
their reference point. The results were given to the subjects and checked after each
round.
The purpose of this novel treatment arrangement is to capture the real effect on
the subject’s risk preference after perceiving a gain or a loss due to his previous risk
choice. We argue that this framework can definitely elucidate the house-money, loss
aversion contradiction. Observe that in Treatment F, the amount available to be invested
by the subject is the same for all rounds, independent of his choices, while in Treatment
C, this value is affected by the previous outcome. In other words, the subject is
effectively experiencing the effects of his previous risky choice. In this framework, very
close to a real investment decision situation, we expect to distinguish between housemoney
and loss aversion.
Evidence that past winners invest more in the following period is consistent with
the house-money effect. Otherwise, we have an indication that the loss aversion effect is
more relevant, since the past winners are willing to take less risk to avoid future losses.
It could be argued that the decisions in our proposed economic experiment could be
distorted, because the money the subjects risk comes from the experimenter, rather than
their own pockets. However, Clark (2002) found no evidence of the previous distortion,
suggesting that use of "free" initial money endowments does not have an impact on the
final experimental results. On the other hand, there are good reasons for supplying
subjects with starting funds in experiments. Specifically, it facilitates recruitment, and
allows subjects to make decisions in the realm of losses without leaving the experiment
in debt. Also, since we are investigating future risk decisions after experiencing a loss
Chapter Three: House-Money and Disposition Effect Overseas
65
(gain), our results tend to be more reinforced than if subjects were using their own
money.
Some final remarks about the experiment: we recruited 85 subjects for the
Brazilian group from the Universidade Católica de Brasília and 97 subjects for the
Spanish group from the Universidad Carlos III de Madrid. The treatment was run from
January to April 2006, in a classroom in Spain and another in Brazil. Subjects were
seated apart from each other avoiding communication. No more than 15 students were
evaluated at a time. Each subject could take part in only one treatment group. All
treatments were run using pencil and paper. After each round, experimenters circulated
to check whether the subjects were calculating the payoffs correctly. To determine if a
subject wins or loses in a round depends on the winning color previously settled for
each participant: green, yellow or red. If the individual winning color is equal to the one
randomly defined by the experimenter for that round, the subject wins; otherwise she
loses. In order to compare our results with the work of Thaler and Johnson (1990), we
included two questions taken from their questionnaire in our experiment. We also
included two control variables: sex and age.
3.4. Experiment Results
The descriptive statistics of the sample are presented in Table 3. In total, we had
182 subjects taken from Brazil (85 students) and Spain (97 students). In terms of sex
distribution, the sample is equitative except for Treatment C (Brazil), which had 75.8%
of men. The average age of the participants was similar across the treatments, with the
Chapter Three: House-Money and Disposition Effect Overseas
66
Brazilian subjects being slightly older. The average earnings in each treatment was 1084
units (Treatment F), 1076 (Treatment I) and 1611 (Treatment C)31.
Table 3 - Descriptive Statistics
Descriptive statistics given for each treatment (F: frequent feedback; I: infrequent feedback; C:
cumulative results). The Table provides the number of subjects in each group (# Students), the percentage
of men (% Men), the average age of the students (Avg. Age), and the average bet as a percentage over the
total amount available for the subject (Avg. Bet).
The main variable of interest is the amount invested by the students in the
lottery, since we want to infer their risk aversion (the results for the Treatment C were
normalized for a range between 0 and 100, to be comparable with the other treatments).
From Table 3, we observe that the average bet varied from 40.10 to 66.02 among the
treatment groups. Figure 2 shows the average bets for all groups. To maintain
consistency with previous literature, we first make a non-parametric analysis of the
results.
In line with our predictions, we observe a country effect, with Spanish students
being less risk averse than Brazilian subjects in all treatment groups. Since MLA
predicts that subjects in the F treatment should invest less than those in the I treatment,
we directly observe their means in each country. Related to Brazil, the group F had an
31 Throughout our analysis we considered relative bets in order to evaluate the treatment C under the same
norm of the remaining treatments. We argue that house-money and disposition effects might remain in
relative terms.
Country Treatment # Students % Men Avg. Age Avg. Bet
F 24 50.0% 24.08 40.10
Brazil I 28 57.1% 25.64 55.38
C 33 75.8% 21.36 44.45
F 32 56.3% 21.59 54.42
Spain I 32 37.5% 22.09 66.02
C 33 54.5% 21.82 47.24
Chapter Three: House-Money and Disposition Effect Overseas
67
average investment of 40.10, while the I treatment had 55.38. The results for the
Spanish subjects follow the same tendency, with the F group having 54.42, and the I
treatment 66.02. In both samples, the predictions of MLA were verified and our results
were close to the ones found by Haigh and List (2005): 50.89 (group F) and 62.5 (group
I).
Figure 2 Average Bets
Average bets given by the percentage invested over the total amount available, grouped by country
(Brazil and Spain) and treatment (F – frequent feedback; I – infrequent feedback; C – cumulative results).
0,00
10,00
20,00
30,00
40,00
50,00
60,00
70,00
F I C
Brazil Spain
To determine the significance of the differences, we used the non-parametric
Mann-Whitney test32. Table 4 presents the results segregated by groups of three rounds.
The Table can be read as follows: row 1, column 1, at the intersection of “Rounds 1-3”
and “Treatment F”, denotes that the average student in Treatment F bet 45.9 (with a
standard deviation of 30.9) units in rounds 1-3. As a comparison, column 2 in the same
32 We cannot use the parametric t-test. Given the fact that subjects are confronted with an upper and lower
bound for investing, the distribution must be non-normal. Moreover, results from the Kolmogorov-
Smirnov test confirm the non-normality of data.
Chapter Three: House-Money and Disposition Effect Overseas
68
row indicates that the average student in Treatment I bets 55.7 (standard deviation of
28.1) in the same rounds.
Considering the whole sample, the average investment of group F was 48.3 and
that of group I was 61.1, with a Mann-Whitney statistic of -7.49 (p-value of 0.000),
indicating that these averages are statistically different. When we consider the time
evolution of the bets, we observe the same pattern, with group F investing consistently
less than group I. Also, subjects in Treatment I are investing more in latter rounds than
in the first ones, indicating some learning effect. When we compare Treatments F and
C, the previous difference disappears. The average student in Treatment C invested 46.8
(standard deviation of 30.9) over all rounds, and the Mann-Whitney statistic when
compared to the F group equals to 0.71 (p-value of 0.480), hence indicating no
statistical difference. Another interesting observation from Table 4 is that subjects tend
to bet more in latter rounds and this pattern remains among all treatments. Our intuition
is that there is some learning effect and as the rounds are played, subjects realize the
positive expected value of the lottery and are motivated to invest more.
Table 4 - Treatments F, I and C
Note: Columns 1-3 summarize student’s betting behavior over rounds. Standard deviations are provided
in parentheses and p-values in brackets. Columns 4-5 summarize Mann-Whitney tests of the differences
in behavior across treatment type.
In order to investigate age and gender effects, we grouped the results
accordingly, reaching the numbers presented in Tables 5 and 6. The results are not clear,
Rounds 1-3 45.9 (30.9) 55.7 (28.1) 42.8 (27.5) -3.59 [0.000] 0.64 [0.520]
Rounds 4-6 45.3 (34.6) 58.1 (28.4) 41.9 (29.9) -3.96 [0.000] 0.48 [0.634]
Rounds 7-9 46.8 (35.5) 62.6 (30.1) 48.2 (30.4) -4.33 [0.000] -0.91 [0.362]
Rounds 10-12 55.2 (36.6) 67.8 (29.8) 50.5 (34.6) -3.15 [0.002] 1.18 [0.237]
Rounds 1-12 48.3 (34.6) 61.1 (29.4) 45.8 (30.9) -7.49 [0.000] 0.71 [0.480]
Treatment F Treatment I Treatment C Mann-Whitney z
F vs. I F vs. C
Chapter Three: House-Money and Disposition Effect Overseas
69
with some gender effect appearing in the groups I and C, and age effect being
significant in group F. Despite the previous results on gender and age were not very
robust, they go in line with the cultural behavioral prediction that a man and young
subject is more individualist and so less risk averse. Johnson et al. (2006), experiment
results show that age increases loss aversion.
Table 5 - Treatments grouped by gender
Columns 1-2 summarize student’s betting behavior over treatments, grouped by gender. Standard
deviations are provided in parentheses and p-values in brackets. Column 3 summarizes Mann-Whitney
tests of the differences in behavior across age groups.
Rounds 1-12
Treatment F 46.6 (32.0) 49.8 (36.7) -0.57 [0.570]
Treatment I 59.4 (25.9) 62.9 (32.9) -2.52 [0.012]
Treatment C 41.7 (26.2) 48.1 (32.9) -2.33 [0.020]
Women Men Mann-Whitney z
Table 6 - Treatments F and I grouped by age
Columns 1-2 summarize student’s betting behavior over treatments, grouped by gender. Standard
deviations are provided in parentheses and p-values in brackets. Column 3 summarizes Mann-Whitney
tests of the differences in behavior across age groups.
Table 7 presents the investment results of Treatments F, C and I, grouped by
country. We can observe a significant country effect in Treatments F and I with subjects
in Spain being consistently less risk averse, while group C didn’t present a statistically
Rounds 1-12
Treatment F 50.6 (35.0) 40.7 (32.1) -2.98 [0.003]
Treatment I 62.8 (28.9) 59.0 (32.9) -1.24 [0.216]
Treatment C 46.0 (31.1) 46.3 (30.9) 0.10 [0.923]
Age <= 23 Age > 23 Mann-Whitney z
Chapter Three: House-Money and Disposition Effect Overseas
70
significant difference33. This result goes in line with our predictions based on the fact
that Spanish students are wealthier and slightly more individualists. Observe that all
subjects were undergraduate students and so sharing approximately the same
educational level, which could be influencing our results. That’s why the differences in
the risk appetite of Spanish and Brazilian students are probably due to economic factors
and not to the educational level. Also, Spanish students were slightly younger than the
Brazilian subjects which can help explaining the results.
Table 7 - Treatments F and I grouped by country
Columns 1-2 summarize student’s betting behavior over treatments, grouped by country. Standard
deviations are provided in parentheses and p-values in brackets. Column 3 summarizes Mann-Whitney
tests of the differences in behavior across country groups.
To infer the house-money effect against loss aversion, Table 8 presents the
average investment decision in each group, considering the decision taken after a loss
and after a gain. Contradicting the predictions of the house-money effect, subjects
invested more after a loss and less after a gain, supporting loss aversion, in Treatment
C. However house-money prevails in Treatment I. So it seems that if we don’t make the
33 We believe that these results could in part be due to a wealth effect. One unit in the experiment
represented 1 Euro cent for students in Spain and 1 Brazilian Real cent for students in Brazil. Since 1
EUR = 2.2 BRL, European students received 2.2 as much as students in Brazil. However in terms of
purchase power parity, this difference is much less. (1 BigMac = 4.50 BRL in Brazil = 3.50 EUR in
Spain)
Rounds 1-12
Treatment F 48.7 (36.1) 54.4 (35.0) 2.56 [0.011]
Treatment I 58.1 (28.5) 66.0 (29.9) 3.97 [0.000]
Treatment C 45.5 (31.0) 46.5 (29.3) 0.49 [0.625]
Brazil Spain Mann-Whitney z
Chapter Three: House-Money and Disposition Effect Overseas
71
subject feel the monetary consequences of his previous decision affecting his available
income for his future choices (as we did in Treatment C), the results are not clear.
An interesting result is that when we asked the subjects about their risk
preferences, in line with Thaler and Johnson (1990), we found that 79% of the students
would increase their bets after a gain and just 37% would invest after a loss; thereby
supporting the house-money effect. We conclude from these findings that the prospect
of a loss (gain) is different from the experience of a loss (gain), and in a dynamic
environment, the outcome effect dominates the individual’s previous cognitive
evaluation. Previous papers (Haigh and List, 2005; Bellamare et al, 2005; Gneezy and
Potters, 1997) didn’t find conclusive results about this dynamic risk aversion behavior.
Table 8 - Treatments F and I grouped previous gain or loss
Columns 1-2 summarize student’s betting behavior over treatments, grouped by previous result. Standard
deviations are provided in parentheses and p-values in brackets. Column 3 summarizes Mann-Whitney
tests of the differences in behavior across previous result groups.
Rounds 1-12
Treatment F 50.1 (32.3) 49.5 (35.4) 0.61 [0.520]
Treatment I 61.2 (29.7) 59.1 (32.7) 1.27 [0.181]
Treatment C 53.1 (31.5) 43.5 (31.5) 5.09 [0.000]
After Loss After Gain Mann-Whitney z
Although the previous analysis already gives some support to hypothesis 2a, and
rejects hypothesis 2b, we can implement a complementary statistical inference. As in
Haigh and List (2005), to provide a test for robustness, we estimate a regression model,
in which we regressed34 the individual bet on a dummy variable for the country, a
dummy variable for the treatment and on the interaction between the two. The results
are presented in Table 9, in which we can observe a significant country effect.
34 We estimated a Tobit regression (censored regression model) as the dependent variable (individual bet)
is non-negative.
Chapter Three: House-Money and Disposition Effect Overseas
72
Considering students from groups F and I, the average student from Spain
invests 10.40 more than a Brazilian student. The treatment effect is also significant
among groups F and I, indicating that students from Treatment F, on average, invest
16.87 less than the students from group I, again supporting the MLA predictions. When
we consider the students from Treatments F and C, the country effect is still significant
but no treatment effect remains.
Table 9 - Regression Results
Tobit regression, p-values are provided in brackets. The dependent variable is the individual bet. “Brazil”
is the omitted subject category, and therefore represents the baseline group. Country S (Treatment F) is
the country (treatment) indicator variable that equals 1 if the subject was a Spanish student (in Treatment
F), 0 otherwise. Country S*Treatment F is the country indicator variable interacted with the frequent
feedback treatment variable. P-values of each regression coefficient are in brackets. The F-statistic, R2
and the number of observations, N, are also provided.
Variable
Constant 55.13 [0.000] 42.80 [0.000]
Country S 10.40 [0.000] 4.94 [0.050]
Treatment F -16.87 [0.000] -3.68 [0.182]
Country S*Treatment F 4.36 [0.227] 9.83 [0.009]
F stat
R2
N 1464
7.34% 2.08%
Specification
F and I F and C
36.66 10.35
1392
Finally, to investigate the house-money effect, we regressed students’ bets on a
dummy variable indicating if the student had a prior gain or loss, a dummy for the
treatment and their interaction. Table 10 provides the results. When we consider
students from groups F and I, no significant win/lose effect is found, which is in
agreement with previous results found in the literature that use an experimental design
close to ours. The treatment effect is significant among groups F and I, indicating that
Chapter Three: House-Money and Disposition Effect Overseas
73
students from Treatment F, on average, invest 9.88 less than the students from group I,
again supporting the MLA predictions.
If we investigate groups F and C, the fact of having a prior gain induces a
significant reduction in the amount invested by 9.68, gives support to our hypothesis 2a,
and rejects the house-money effect. It also indicates that our experimental design
(Treatment C) could elucidate the contradiction between house-money and loss aversion
in dynamic settings.
Table 10 - Regression Results
Tobit regression, p-values are provided in brackets. The dependent variable is the individual bet.
“Previous Loss” is the omitted subject category, and therefore represents the baseline group. Prior Gain
(Treatment F) is the win (treatment) indicator variable that equals 1 if the subject won in the previous
round (in Treatment F), 0 otherwise. Prior Gain*Treatment F is the win indicator variable interacted with
the frequent feedback treatment variable. P-values of each regression coefficient are in brackets. The Fstatistic,
R2 and the number of observations N are also provided.
Variable
Constant 60.852 [0.000] 52.084 [0.000]
Prior Gain 0.91 [0.747] -9.68 [0.011]
Treatment F -9.88 [0.000] -1.17 [0.619]
Prior Gain*Treatment F -1.33 [0.456] 8.19 [0.048]
F stat
R2
N
Specification
F and I F and C
15.25 28.47
1276 1342
2.68% 5.58%
Observe that the coefficient for Treatment F in the specification F and C was
negative (-1.17) but not significant and the coefficient for the compounding effect of
Prior Gain*Treatment F was even positive and significant which would lead to an
opposite result. With Treatment C we could distinguish the effect of previous risk
Chapter Three: House-Money and Disposition Effect Overseas
74
decisions and outcomes over future decisions and in these cases loss aversion dominates
house-money.
3.5. Concluding Remarks
In the behavior finance literature, experiments were used to identify or give
support to the existence of behavioral biases, which could clarify market anomalies.
Nevertheless, those experiments were mainly made in unique locations: the USA or
Europe. Moreover, two of these behavioral effects - house-money and loss aversion -
lead to controversial results when we analyze the risk taking behavior of agents in a
dynamic setting.
Our results support the existence of MLA over the countries but a country effect
is also found, indicating that care should be taken when generalizing behavioral findings
over the international financial market. A slightly salient gender and age effect was
found in the sample. Young and male subjects presented a lower level of risk aversion.
Finally, and most importantly, disposition effect is found to dominate the risktaking
behavior of subjects in dynamic settings, overcoming the house-money effect.
Subjects that experienced a gain (loss) tend to assume less (more) risk in the following
period. As possible extensions we suggest to go further in the investigation of which
country characteristics are generating the effect found in this study, and also the use of
other countries’ students replicating the experimental design C, since we have found
that there seems to be a country effect as well, which could be due to cultural
differences.
75
Chapter Four
Behavior Finance and Estimation Risk in Stochastic Portfolio
Optimization
4.1. Introduction
In a standard asset allocation process, once the risk tolerance, constraints, and
financial goals are set, the output is given by a mean-variance optimization (Markowitz,
1952; Feldman and Reisman, 2002). Unfortunately this procedure is likely to fail for
individuals, who are susceptible to behavioral biases. For instance, in response to shortterm
market movements and to the detriment of the long-term investment plan, the
individual investor may require his asset allocation to be changed. Fernandes et al.
(2007) suggest that early liquidation of a long term investment may be the cause of
momentum.
Moreover, the paradigm of individual behavior in finance theory is based on
expected utility maximization and risk aversion, which has been under attack in recent
years due to its descriptive inaccuracy. Experimental psychologists have demonstrated
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
76
that people systematically deviate from the choice predictions it implies as they are
typically biased.
Behavioral biases can roughly be grouped in two categories: cognitive and
emotional, though both types yield irrational decisions. Because cognitive biases
(heuristics like anchoring, availability, and representative biases) stem from faulty
reasoning, better information and advice can often correct them. Conversely, emotional
biases, such as regret and loss aversion, originate from impulsive feelings or intuition,
rather than conscious reasoning, and are hardly possible to correct. Lo et al. (2005)
investigated several possible links between psychological factors and trading
performance, finding that subjects whose emotional reaction to monetary gains and
losses was more intense on both the positive and negative side exhibited significantly
worse trading performance.
Shefrin (2005) posits that the portfolios selected by investors whose choices
conform to prospect theory will differ in key aspects from the portfolios selected by
investors whose choices conform to expected utility theory. In this sense, an optimal
solution to the asset allocation problem should guide investors to make decisions that
serve their best interest. This could be the recommendation of an asset allocation that
suits the investor’s natural psychological preferences (emotional biases), even though it
may not maximize expected return for a given level of risk. More simply, a client’s best
practical allocation may be a slightly under-performing long-term investment program
to which the investor can comfortably adhere. From a mean-variance optimization
perspective, behavioral investors select portfolios that are stochastically dominated.
This does not mean that the individual investors are irrational in any sense: it is not
irrational for people to anticipate emotional reactions and take them into account when
making decisions that try to synchronize their choices to their preferences. However,
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
77
portfolio managers lack the guidelines necessary for incorporating these biases during
the process of determining asset allocation. We address this issue by evaluating whether
managers should moderate the way clients naturally behave to counteract the effects of
behavioral biases so that they can fit a pre-determined asset allocation or they should
create an asset allocation that adapt to clients’ biases, so that clients can comfortable
adhere to the fund.
In terms of emotional biases, several empirical studies (Tversky and Kahneman,
1992) have shown that, when dealing with gains, agents are risk averse, but when
choices involve losses, agents are risk seeking (asymmetric risk taking behavior).
Moreover, in a wide variety of domains, people are significantly more averse to losses
than they are attracted to same-sized gains (Rabin, 1998). Loss aversion (Schmidt and
Zank, 2005) is a relevant psychological concept that has been imported to financial and
economic analysis, and it represents the foundation of prospect theory.
In general terms, prospect theory and its latter version cumulative prospect
theory35 (Kahneman and Tversky, 1979, 1992) posits four novel concepts in the
framework of individuals’ risk preferences. First, investors evaluate assets according to
gains and losses and not according to final wealth (mental accounting). Second,
individuals are more averse to losses than they are attracted to gains (loss aversion).
Third, individuals are risk-seeking in the domain of losses and risk averse in the domain
of gains (asymmetric risk preference). Finally, individuals evaluate extreme
probabilities in a way that overestimates low probabilities and underestimates high
probabilities (probability weighting function). This study, as far as we know, is the first
to consider all those aspects in the framework of portfolio choice.
35 Tversky and Kahneman’s Cumulative Prospect Theory (CPT) (1992) combines the concepts of loss
aversion and a non linear rank dependent weighting of probability assessments.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
78
There are conflicting results in the finance literature on how prior outcomes
affect the risk taking behavior of investors in subsequent periods. Loss aversion would
predict that traders with profitable mornings would reduce their exposure to afternoon
risk, trying to avoid losses and thus guaranteeing the previous gains (Weber and Zuchel,
2003). Odean (1998) and Weber and Camerer (1998) have shown that investors are
more willing to sell stocks that trade above the purchase price (winners) than stocks that
trade below purchase price (losers) – a phenomenon termed the disposition effect
(Schefrin and Statman, 1985). Both works interpreted this behavior as evidence of
decreased risk aversion after a loss, and increased risk aversion after a gain. The
standard explanation for the previous behavior is based on prospect theory, and
particularly on the fact that individuals are risk-seeking in the domain of losses and risk
averse in the domain of gains (asymmetric risk preference).
However, another stream of the literature found the opposite behavior. Thaler
and Johnson (1990) name the house-money effect, the behavior of increasing risk
appetite after a gain. Barberis et al. (2001) present a model where investors are less loss
averse after a gain while they become more loss averse after prior losses.
Despite the vast literature confirming the behavioral biases associated with
prospect theory, the consideration of all those biases in an asset allocation framework is
still missing. Barberis and Huang (2001) and Barberis, Huang, and Santos (2001) use
loss aversion and mental accounting (Thaler, 1999) to explain aspects of stock price
behavior, but do not employ the full prospect theory framework and don’t examine
optimal asset allocation. Benartzi and Thaler (1995) consider prospect theory to solve
the equity premium puzzle when investors are loss averse and evaluate their portfolios
myopically with a horizon of approximately one year. They also suggest an optimal
allocation in equities from 30% to 55%. Magi (2005) uses behavioral preferences to
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
79
numerically solve a simple model of international portfolio choice, providing a possible
explanation for the equity home bias puzzle.
Davies and Satchell (2004) provide a solution for the optimal equity allocation,
and explore more thoroughly the cumulative prospect theory parameter space that is
consistent with observed equity allocations given a financial market’s returns
distributions over a one-month horizon. Shefrin (2005) considers heterogeneous
investors to see the impact of behavioral concepts in the framework of asset pricing.
The first main goal of this study is to incorporate mental accounting, loss
aversion, asymmetric risk taking, disposition effect, and probability weighting in
portfolio optimization in a multi-period setting for individual investors. We provide a
solution for the asset allocation problem, taking into account all behavioral biases
associated with prospect theory and using a utility function (suggested in Giorgi et. al.,
2004) consistent with both the experimental results of Tversky and Kahneman, and also
with the existence of equilibrium. We also shed more light on the issue of how prior
outcomes affect subsequent risk taking behavior, investigating the investor’s risk taking
behavior following a rise, or a fall, in the price of the risky asset.
In line with prospect theory, investors derive utility from fluctuations in the
value of their final wealth. In our framework, there is a financial market on which two
assets are traded. A riskless asset, also called a bond, and a risky asset, also called a
stock (under the assumption of normally distributed returns for the risky asset). As we
are modeling the decision making process of an individual investor, short-selling is not
allowed. In each period (we consider two periods), the investor chooses the weight of
his endowment to be invested in the risky asset, in order to maximize his utility
(prospect theory based). We assume that the investor acts myopically, and that the
reference point relative to which he measures his gains and losses for the first period is
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
80
his initial endowment. Although all agents solve the same maximization problem in the
first period, the second period decision depends on the reference point relative to which
the agent measures the second period outcomes (gains or losses). We consider two
possible reference points: the initial wealth or the current wealth, and analyze both
cases. St-Amour (2006) results reveal that references are strongly relevant and statedependent.
Another well-known issue in asset allocation problems, using Markowitz
optimization, is that the output is strongly driven by the risk/return estimation, which
usually generates very unstable portfolios. The most famous problem with this
technique is the substitution problem, where two assets with the same risk but slightly
different expected returns. The optimizer would give all the weight to the asset with the
higher expected return, leading to a very unstable asset allocation.
Recent literature has tried to overcome the previous problem of leading to
unfeasible portfolios. The main focus of those models is to find out how to create
realistic portfolios considering that the values used for risk and return are not
deterministic but instead just estimates (they are stochastic). It should be noted that the
misspecification of expected returns is much more critical than that of variances
(Zimmer and Niederhauser, 2003).
Jorion (1986) offers a simple empirical Bayes estimator that should outperform
the sample mean in the context of a portfolio. His main idea is to select an estimator
with average minimizing properties relative to the loss function (the loss due to
estimation risk). Instead of the sample mean, an estimator obtained by “shrinking” the
means toward a common value is proposed (the average return for the minimum
variance portfolio), which should lead to decreased estimation error. Similar to Jorion,
Kempf (2002) assumes that the prior mean is identical across all risky assets. However,
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
81
Kempf’s model considers estimation risk as a second source of risk, determined by the
heterogeneity of the market and given by the standard deviation of the expected returns
across risky assets.
Black and Litterman (1992) postulate that the consideration of the global CAPM
equilibrium can significantly improve the usefulness of asset allocation models, as it can
provide a neutral starting point for estimating the set of expected excess returns required
to drive the portfolio optimization process. Horst et al. (2002) propose a new adjustment
in mean-variance portfolio weights to incorporate the estimation risk. The adjustment
amounts to using a pseudo risk-aversion, rather than the actual risk aversion, which
depends on the sample size, the number of assets in the portfolio, and the curvature of
the mean-variance frontier. The pseudo risk aversion is always higher than the actual
one and this difference increases with the uncertainty in the expected return estimations.
Maenhout (2004) also considers an adjustment in the coefficient of risk aversion to
insure the investor against some endogenous worst case.
Finally, Michaud (1998) suggests portfolio sampling as a way to allow an
analyst to visualize the estimation error in traditional portfolio optimization methods,
and Sherer (2002) posits that sampling from a multivariate normal distribution (a
parametric method termed Monte Carlo simulation) is a way to capture the estimation
error. Markowitz and Usmen, 2003, compared the traditional approach to resampling
and their results support the latter. Fernandes et al. (2007) evaluates several asset
allocation models and suggests that resampling methods typically offer the best results.
This study presents a novel approach (BRATE – Behavioral Resampling
Adjusted Technique) to incorporate behavioral biases and estimation risk into meanvariance
portfolio selection. In a paper close to ours, Vlcek (2006) proposes a model to
evaluate portfolio choice with loss aversion, asymmetric risk-taking behavior, and
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
82
segregation of risk less opportunities. His findings suggest that the changes in portfolio
weights crucially depend on the reference point and the ratio between the reference
point and the current wealth, and thus indirectly on the performance of the risky asset.
Our work differs from his study as we explicitly consider all novel aspects of prospect
theory: mental accounting, loss aversion, asymmetric risk-taking behavior, and
probability weighting function. We also evaluate the inefficiency cost of the behavioral
biases and consider a more general form for the risky asset return process, including
estimation risk in the analysis.
Considering daily equity data from the period from 1995 to 2007, we empirically
evaluate our model in comparison to the traditional Markowitz model. Our results
support the use of BRATE as an alternative for defining optimal asset allocation and
posit that a portfolio optimization model may be adapted to the individual biases
implied in prospect theory.
The remainder of this Chapter contains the following sections. Section 2
discusses the behavioral biases considered and describes our model proposing the
behavioral resampling adjusted technique (BRATE). Section 3 presents the empirical
study, describing the data and implementation, and providing the results. Section 4
concludes the chapter by reviewing the main achievements.
4.2. The Behavioral Model
We present a two period’s model for portfolio choice in a stylized financial
market with only two assets, where the investor’s preferences are described by
cumulative prospect theory as suggested by Kahneman and Tversky (1979) and Tversky
and Kahneman (1992). In our framework, there is a financial market in which two assets
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
83
are traded. A riskless asset, also called the bond, and a risky asset, the stock. Let us
consider the return of the stock in each period given by the following process:
R = μ +σn , with n ~ N(0,1) . The risk free bond yields a sure return of f R . We assume
that the time value of the money is positive, i.e. that interest rates are non-negative.
The preferences of the investor are based on changes in wealth and are described
by prospect theory. We assume that he owns an initial endowment, 0 W (normalized to 1
monetary unit), and that he earns no other income. The agent invests a proportion θ of
his wealth in the stock and (1 - θ ) in the bond. Since we want to model the individual
investor behavior, we assume that short selling is not allowed ( 0 ≤θ ≤ 1). We also
assume that the investor acts myopically, and the reference point relative to which he
measures his gains and losses in the first period is his initial wealth. Then, the perceived
gain or loss in the end of the first period is given by:
[( ) ]
( )
(1 ) ( ) (Eq. 01)
1
1 (1 ) (1 ) 0 0 0
x R n
x R R
x W W R W R W
f
f
f
θ θ μ σ
θ θ
θ θ
∴ = − + +
∴ = − +
= Δ = − + + + −
As pointed out in Vlcek (2006) the choice process under prospect theory starts
with the editing phase, followed by the evaluation of edited prospects, and finally the
alternative with the highest value is chosen. During the editing phase, agents
discriminate gains and losses. They also perform additional mental adjustments in the
original probability function p = f (x) , defining the probability weighting function
π ( p) . Based on experimental evidence, individuals adjust the likelihood of outcomes
such that small probabilities are overweighted and large probabilities are
underweighted. We will consider the probability weighting function, as in Giorgi et al.
(2004) given by:
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
84
( γ γ )γ
γ
π 1
(1 )
( )
p p
p p
+ −
= , (Eq. 02)
where γ is the adjustment factor. The following graph compares the values of p and
π ( p) , considering γ =0.8036.
0
0,2
0,4
0,6
0,8
1
1,2
0 0,2 0,4 0,6 0,8 1 1,2
pi(p)
p
Figure 1 – Cumulative probability weighting function for γ =0.80.
In the valuing phase, the agents attach a subjective value to the gamble. Let us
assume the value function proposed by Giorgi et al. (2004), as follows:
⎩ ⎨ ⎧
− <
− ≥
= − −
+ + −
, if 0
, if 0
( )
e x
e x
v x x
x
λ λ
λ λ
α
α
(Eq. 03)
where α is the coefficient of absolute risk preference, λ− > λ+ > 0 makes the value
function steeper in the negative side (loss aversion), and x is the change in wealth or
welfare, rather than final states (mental accounting), as proposed by Kahneman and
Tversky (1979). Also, the value function is concave above the reference point and
convex below it (asymmetric risk preference). It is useful to consider the previous
36 Experiments suggest a value of γ between 0.80 and 0.90.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
85
form for the value function because of the existence of a CAPM equilibrium37 and
the ability to reach constant coefficients of risk preference (α ). The previous
formulation is also supported by the laboratory results from Bosh-Domènech and
Silvestre (2003). The following graph indicates v(x) when α = 0.88, λ− = 2.25 and
λ+ = 1 ( Kahneman and Tversky suggested values).
-1,5
-1
-0,5
0
0,5
1
-1,5 -1 -0,5 0 0,5 1 1,5
Figure 2 – Prospect theory value function for α = 0.88, λ− = 2.25 and λ+ = 1
In our two-period model for portfolio choice, the investor chooses a weight in
the risky asset to maximize his expected utility (V). His preferences are based on
changes in his wealth ( x ) and are described by prospect theory. The total expected
value he addresses to a given choice of θ is given by:
∫
∞
−∞
= f x dx
dx
V v(x) d π ( ( )) (Eq. 04)
37 Under Cumulated Prospect Theory (CPT) with Tversky and Kahneman (1992) specifications, equilibria
do not exist as at least one investor can infinitely increase his utility by infinitely leveraging the market
portfolio (the utility index is almost linear for large stakes), while the Security Market Line Theorem
holds (Giorgi et al. , 2004).
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
86
where v(x) is the prospect value of the outcome x , and π ( f (x)) is the weighted
cumulative probability associated with that outcome. Prospect theory is a descriptive
theory, postulating that, in comparing alternatives, the investor will choose the
alternative that makes V as high as possible. Let us then evaluate the investor’s problem
in each period.
4.2.1 First Period
In the first period, the agent’s problem consists of defining the allocation of his
initial wealth between the two assets traded in the financial market. He maximizes his
utility in t = 0 by allocating a fraction, 0 θ , of his initial wealth38, 0 W , in the risky asset
and (1 - 0 θ ) in the riskfree asset. We consider that the investor is a myopic optimizer in
the sense that he takes into account only the first period result. For multi-period
horizons, the choices at earlier dates impact the reference points at later dates. This
feature allows for complex modeling. However, as pointed out in Shefrin (2005),
prospect theory is a theory about investors who oversimplify, and so, assuming that
individuals are sophisticated enough to perceive the link between their current choices
and future reference points is something unreasonable. We also constrain short selling,
as it is common for individual investors’ models. Thus, his problem can be given by
∫
∞
−∞
≤ ≤
= f x dx
dx
maxV v(x) d ( ( ))
0 1
π
θ
(Eq. 05)
Let us make the following derivation: (1 ) ( ) 0 0 x R n f = −θ +θ μ +σ . Rearranging
the terms in x , we get x ( )R n f θ θ μ θ σ 0 0 0 = 1− + + . We call ( θ ) θ μ 0 0 1− + f R = B and
θ σ = C 0 . Then, x = B + Cn , and so x > 0 implies
C
n > − B . Then,
38 We will consider the investor’s initial wealth equals to 1.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
87
( ) ( )
( ) ( )
( ) ( )
( ) ( ) ( ) ( ) (Eq. 06)
( ) ( ) ( ( ))
1 ˆ( ) ( ) ( ( )) ( ( ))
( ( )) ( ( ))
( ( )) ( ( ))
( ) ( ( ))
2 2
2
1
( ) ( )
0
0
⎥⎦
⎤
⎢⎣
∴ = − + − + ⎡ − − − −
∴ = − + − + −
− + − − ⎟⎠
⎞
⎜⎝
∴ = ⎛ − −
∴ = − + + −
∴ = − + + −
=
+ + − − + −
∞
−
+ − −
∞
+ + − − −
∞
−
+ − −
−
−∞
+ − −
−
−∞
− + −
∞
−
+ − + +
−∞
− −
∞
+ − +
∞
−∞
∫ ∫
∫ ∫
∫ ∫
∫ ∫
∫
C
C
C e B
C
e e B
C
V B
e e d f n e e d f n
C
V B
e e d f n e e d f n
C
B
C
V B
V e d f n e d f n
V e d f x e d f x
f x dx
dx
V v x d
C B B
C
B
B nC
C
B
B nC
C
B
B nC
C
B
B nC
C
B
B Cn
C
B
B Cn
x x
λ λ λ π λ π α λ π α
λ λ λ π λ π λ π
λ π λ π λ π λ π
λ λ π λ λ π
λ λ π λ λ π
π
α α α
α α α α
α α α α
α α
α α
Where, for the last step, we used39:
∫ ( )
∞
− = − −
z
e xdφ x e φ ασ z
ασ α σ ˆ ( )
2 2
2
1
Observe that, if we were considering a standard utility function (risk aversion
over all possible outcomes), the value would be given by:
2 2
2
B 1 C V S e α α
λ λ
+ + − + = − (Eq. 07)
Moreover, the partial derivatives of V (Eq. 06) are:
[( ) ] [( ) ]
[( ) ] [( ) ]
[( ) ]
[( ) ] [( ) ] ( ) [( ) ]
} (Eq. 09)
1
)]
1
(
)
1
{ [ (
]} (Eq. 08)
1
ˆ
1
{ [
0
0
0 0
0
0
( 1 ) 0 0
0
0
( ) 0 0
2
1
0
2
0 0
0
1 0 0
0
0
( ) 1 0 0
2
1
0 0
2
0
2
0 0
0 0
2
0
2
θ
θ σ
θ θ μ
αθ σ α λ λ π
θ σ
θ θ μ
λ π
αθ σ
θ σ
θ θ μ
α θ σ λ π
σ
αθ σ θ
θ σ
θ θ μ
λ π
αθ σ
θ σ
θ θ μ
α λ π
μ
α θ θ μ
α θ σ α
α θ θ μ
α θ σ α θ θ μ
⋅ ⎟
⎟⎠
⎞
⎜ ⎜⎝
⎛ − +
− − −
− +
− −
− +
= −
∂
∂
⋅ ⎟
⎟⎠
⎞
⎜ ⎜⎝
⎛
−
− +
+ ⎟
⎟⎠
⎞
⎜ ⎜⎝ ⎛ −
− +
= −
∂
∂
+ − − + − +
−
+ − − +
− − +
R f f
B f
R f
R f
R R
e
R
V e e
R
e
R
V e e
f
f
f
39 It’s valid for γ=1.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
88
As a consequence, the following properties hold40,
i) > 0
∂
∂
μ
V ;
ii) = 0
∂
∂
σ
V for σ = 0 or σ = ∞;
iii) < 0
∂
∂
σ
V for σ > 0 .
Equations 06 and 07 clearly yield different weights for the risky asset,
considering the remaining parameters fixed. Thus, it is possible to evaluate the cost of
inefficiency associated with the behavioral biases as compared to the standard utility
solution.
[(1 ) ] [(1 ) ] (Eq. 10) 0 0 0 0 Cost R R R PT R
f
S PT
f
= −θ S +θ − −θ +θ
where S
0 θ is the risky asset weight given by the standard utility maximization problem,
and PT
0 θ is the stock weight as defined in our model.
Proposition 1. The optimal asset allocation in t = 0, for the risky asset *
0 θ is such that
maximizes the value function given by:
( ) ⎥⎦
⎤
⎢⎣
= + − + + − (− ) + ⎡ − (− − ) − + − ( − )
2 2
2
1
C
C
C e B
C
e e B
C
V B C B B λ λ λ π λ α π α λ α π α α
where: [( θ ) θ *μ ]
0
*
0 = 1− + f B R and θ *σ
0 C = .
If we were considering a standard utility function, the optimal allocation in t = 0,
for the risky asset would then be given by:
2
*
0 ασ
μ
θ f − R
=
40 See Appendix B for the proofs.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
89
Let us first consider standard values for the model’s parameters41. The risk free
rate equals the historical annual return of the US three-month Treasury Bill ( f R =
2.73%). The equity expected return and volatility equals the historical average of the
MSCI global equity index and its standard deviation (μ = 7.61% and σ = 12.98%). The
adjustment factor in the probability weighting function equals γ = 0.90. The coefficient
of risk aversion equals α = 3. Also, as suggested by Kahneman and Tversky, λ− = 2.25
and λ+ = 1. The individual’s values (prospect theory and standard) as a function of the
percentage of his wealth invested in the risky asset are given in Figure 3. The individual
investor is expected to choose the allocation in the risky asset which maximizes his
expected value.
0 20 40 60 80 100 120
0.08
0.09
0.1
0.11
0.12
0.13
0.14
0.15
0.16
V(θ)
θ%
Vpt
Vs
Figure 3 – Prospect value and standard utility value as function of θ
41 The riskfree rate, the expected return of the risky asset and the volatility of the risky asset were
calculated, using daily data, over the period from 1995 and 2007. The results were annualized.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
90
As can be observed from the graph, using a standard utility function, the
allocation in the risky asset approaches 100% (theta for which the value function
reaches its maximum), while using prospect theory utility, the investor should allocate
81% of his wealth in the stock42. The shapes of the graphs are different, notably for
large allocations in the stock. The value function using standard utility is equal to or
greater than the one for prospect utility.
The reason for this difference comes from the fact that in prospect theory,
negative outcomes are penalized more (as are risky portfolios) because individuals are
loss averse (λ− >λ+ ). In the loss aversion literature evidence suggests that individuals
are around twice more sensitive to losses than they are attracted to same size gains. For
small allocations in stocks, the prospect of losses becomes less likely and the value
functions tend to coincide.
Related to the effect of probability weighting, if we set γ = 1, thus canceling out
its effect, we reach the following Figure representing the value function:
42 Davies and Satchell (2004) found that the average proportion in domestic and foreign equities of large
pension funds in 1993 was 83% in the UK, which is in line with the prospect theory results.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
91
0 20 40 60 80 100 120
0.08
0.09
0.1
0.11
0.12
0.13
0.14
0.15
0.16
V(θ)
θ%
Vpt
Vs
Figure 4 – Prospect value and standard utility value as function of θ
Note that the amount optimally invested by the behavioral investor in the risky
asset decreases to 48%, and so probability weighting tends to increase the risk appetite.
Kahneman and Tversky (1979) suggest that the overweighting of low probabilities has
an ambiguous effect on risk taking, as it can induce risk aversion in the domain of
losses, and risk seeking in the domain of gains. In our case, the overestimation of the
extreme positive outcomes probabilities, shown in Figure 3, is inducing investors to
take more risk.
However, despite the effects of loss aversion and probability weighting, even if
we consider λ− =λ+ = 1 and γ = 1, keeping constant the remaining parameters, the
value functions wouldn’t coincide, as can be seen in Figure 5:
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
92
0 20 40 60 80 100 120
0.08
0.09
0.1
0.11
0.12
0.13
0.14
0.15
0.16
0.17
V(θ)
θ%
Vpt
Vs
Figure 5 – Prospect value and standard utility value as function of θ
Both models would predict that the investor should allocate 100% of his
endowments in the stock. However, the value functions are different because, in
prospect theory, individuals are risk seeking in the loss domain (asymmetric risk
preference). Thus, they would be more comfortable in allocating a greater part of their
wealth in the risky asset. The prospect value function is greater than the standard utility
function.
Observe that the effect of the asymmetric risk preference goes in the opposite
direction of loss aversion and probability weighting. When we diminish the coefficient
of risk preference (α = 0.25) in both utility functions, we reduce the effect of
asymmetry, and so the value functions are much closer, as can be seen in the following
figure.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
93
0 20 40 60 80 100 120
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0.02
V(θ)
θ%
Vpt
Vs
Figure 6 – Prospect value and standard utility value as function of θ
The effects of the behavioral biases can thus be summarized as follows: loss
aversion reduces risk taking, and asymmetric risk taking behavior induces risky
attitudes. Probability weighting has an ambiguous effect on risk. Our intuition is that, in
the long run, as the value function parameters are changing, these biases tend to cancel
out, eliminating the efficiency loss originated by each bias. That is why we argue that
human biases do not need to be moderated to reach an efficient investment strategy. The
experimental results of Blavatskyy and Pogrebna (2006) reveal that the effect of loss
aversion is largely neutralized by the overweighting of small probabilities and
underweighting of moderate and high probabilities.
In order to verify property (i), Let us evaluate V while changing μ and keeping
constant the other parameters (considering θ = 50%). Figure 7 presents the graph which
indicates that over all positive values of μ , the slope of V is positive. The value
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
94
function is increasing in μ . Thus, when the risky asset has a higher expected return,
ceteris paribus implies a higher value for the investor.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
V(μ)
μ
Figure 7 – Prospect value as function of μ
Considering properties (ii) and (iii), Let us evaluate V while changing σ and
keeping constant the other parameters (considering θ = 50%). Figure 8 presents the
graph indicating that over all positive values of σ , the slope of V is negative, while for
σ = 0, the slope is null. When σ tends to infinity, the slope tends to null. The value
function is decreasing in σ .
The intuition is that, if the volatility of the risky asset is higher, for the same
allocation, this implies a higher probability of losses reducing the value of the prospect.
In line with traditional rational investor, behavioral individuals also prefer higher return
and lower risk; mainly because they are risk averse in the gain domain and also loss
averse.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
95
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
V(σ)
σ
Figure 8 – Prospect value as function of σ
Now let us evaluate the values of 0 θ when we change the riskfree rate and the
expected return of the risky asset. Since many parameters are involved, it is not possible
to find closed form solutions for 0 θ . Therefore, we present numerical results for the
optimal allocation of wealth in t = 0. Figure 9 presents the results for 0% < μ < 15%
and 0 < < 6% f R . The remaining parameters are fixed (σ = 12.98%, α = 3, λ− = 2.25,
and λ+ = 1).
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
96
0
0.05
0.1
0.15
0
0.02
0.04
0.06
0
0.2
0.4
0.6
0.8
1
rf μ
θ
Figure 9 – Optimal equity allocation in the first period as function of μ and rf.
As expected, when the risky asset offers more attractive returns, the agent
gradually invests more in the stock. When the stock is very attractive, the investor
chooses to allocate his entire wealth in the risky asset. Thus, we observe that 0 θ is
increasing in μ and decreasing in f R . Also, when f R is higher, the changes in 0 θ due
to a variation in μ are smoother, because in these cases losses are less likely and we
approach the standard utility solution. When f R is lower, the changes in 0 θ due to a
variation in μ are more abrupt, giving rise to extreme portfolio allocations. If we
consider that μ is not known with certainty, the resulting portfolio would be very
unstable. Gomes (2003), in a model with loss-averse investors, has found that
individuals will not hold stocks unless the equity premium is quite high.
We can evaluate the expected cost of inefficiency related to the behavioral
biases associated to the prospect theory function, for the same parameters considered in
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
97
the previous analysis, using equation 10. The result is presented in Figure 10, and its
form is due to the fact that, in standard utility function, the investor is willing to take
more risk than with the loss averse prospect utility. The cost is due to the fact that the
expected return of the stock is greater than the bond, and the standard utility investor is
allocating a greater part of his wealth in the risky asset than the prospect utility
individual. Thus, the cost is increasing in μ . However, it is worth noting that the
previous cost is based on expected returns, which are stochastic in practice. The real
cost can just be observed at the end of the first period with the realization of the stock’s
return.
An important insight can be made from Figure 10 in terms of the best practice
for asset allocation. As long as the risk free rate is lower and the expected return of the
stock is higher, the optimal allocation should moderate the investor’s biases in order to
reach a better performance. On the other hand, if the risk premium is lower, the
moderation is less relevant, and the optimal allocation may adapt to the individual’s
biases.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
98
0
0.05
0.1
0
0.02
0.04
0.06
0
5
10
15
rf μ
Cost
Figure 10 – Expected cost in the first period as function of μ and rf.
We can also analyze the change in the allocation of the stock when we vary the
loss aversion in the risk taking behavior. The result is shown in Figure 11, for
2 < λ− < 4 . Observe that, as long as the investor is much more averse to losses than he
is attracted to gains, the allocation in the risky asset is lower. When λ− = 2.25 , the
allocation in the risky asset corresponds to 81%, as previously mentioned.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
99
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
θ
λ -
Figure 11 –Optimal equity allocation in the first period as function of λ− .
Dimmock(2005) has already shown that a higher level of loss-aversion leads to
lower equity exposure, and heterogeneity in the coefficient of loss aversion has the
ability to explain puzzling features of household financial behavior.
4.2.2 Second Period
In order to evaluate the second period allocation choice of the investor, Let us
keep some parameters fixed: (σ = 12.98, α = 3, λ− = 2.25 and λ+ = 1). After the
investor has made his first period decision in t = 0, the state of nature realizes in t = 1,
when he is faced with his second period problem. Again, he must allocate his wealth in
the two possible assets in the financial market, bond and stock, to maximize his utility.
Let us consider the same normal distribution for the return of the risky asset. The
investor’s wealth position at t = 2 equals his position in t = 1 plus the return of his
portfolio in the second period.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
100
While all agents solve the same maximization problem in the first period, in the
second period, it will depend on the reference point to which he measures his gains and
losses (in the framework of prospect theory). In our model, there are two candidates for
the investor’s reference point at t = 1: his initial wealth at t = 0 (W0 = 1) or his wealth at
the end of the first period, t = 1 ( 1 W ). If he measures his gains and losses relative to his
wealth at t = 1 (his current wealth), he treats each gain and loss separately. On the other
hand, if he considers his initial wealth as the reference point, he adds up the outcomes
(gains and losses), that is, he nets his positions. The previous distinction is relevant in
prospect theory. The value function is concave in the domain of gains and convex in the
loss domain (asymmetric risk behavior).
First, Let us consider as the investor’s reference point his current wealth at t = 1.
In this case, the maximization problem he will solve in the second period is the same as
the one for the first period. Thus, we can state the following proposition.
Proposition 2. The optimal asset allocation in t = 1, for the risky asset *
1 θ
, if the agent
measures his gains and losses relative to his current wealth, is such that maximizes the
same value function of the first period. *
0
*
1 θ =θ
We can observe that an individual who measures his gains and losses relative to
his current wealth is actually solving the same maximization problem in each period.
That is why the allocation in the risky asset might be the same. This is not surprising; as
he is not using past information to update his beliefs about the assets, his preferences
are similarly unaffected.
Next, let us analyze the investor’s maximization problem if he evaluates his
gains and losses relative to his initial wealth. If he has an initial wealth position of 0 W
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
101
= 100 and his wealth rises in the first period to 1 W = 110 and falls in the next period to
2 W = 105, he values his position at t = 2 as a gain of 5, and not as a gain of 10 followed
by a loss of 5.
In the second period, the agent’s problem consists of defining the allocation of
his wealth ( 1 W ) between the two assets traded in the financial market. He maximizes his
utility in t = 1 by allocating a fraction, 1
θ , of his wealth 1 W in the risky asset and 1- 1
θ
in the risk-less asset. As we did in the first period analysis, we also constrain short
selling.
∫
∞
−∞
≤ ≤
= f x dx
dx
maxV v(x) d ( ( ))
0 1
π
θ
Let us make the following derivation:
[( ) ] [( ) ] 1 1 1 0 0 0 1 x W 1 R ( n) W 1 R R f f = −θ +θ μ +σ + −θ +θ
and [ ( ) ] 1 0 0 0 1 W W 1 1 R R f = + −θ +θ , where 1 R is the return of the stock in the first
period. So [( ( ) ) (( ) ) ( ) ] 0 0 0 1 1 1 0 0 1 x W 1 1 R R 1 R ( n) 1 R R f f f = + −θ +θ ⋅ −θ +θ μ +σ + −θ +θ .
Rearranging the terms in x and considering W0 = 1, we get
( (( ) )) ( (( ) )) [( ) ]
((1 ) )]
[ 1 1 1 1 1
0 0 1
1 0 0 1 0 0 1 1 1
R R
x n R R R R R
f
f f f
θ θ
θ σ θ θ θ θ θ θ μ
+ − +
= + − + + + − + ⋅ − + +
Let us call
[(1 ((1 ) )) [(1 ) ] ((1 ) )] 0 0 1 1 1 0 0 1 B R R R R R f f f = + −θ +θ ⋅ −θ +θ μ + −θ +θ
and
( (( ) )) 1 0 0 1 C 1 1 R R f =θ σ + −θ +θ
Then, x = B + Cn , so x > 0 implies
C
n > − B . Then,
( ) ⎥⎦
⎤
⎢⎣
= + − + + − (− ) + ⎡ − (− − ) − + − ( − )
2 2
2
1
C
C
C e B
C
e e B
C
V B C B B λ λ λ π λ α π α λ α π α α
(Eq. 11)
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
102
Proposition 3. The optimal asset allocation in t = 1, for the risky asset *
1 θ
, if the agent
measures his gains and losses relative to his initial wealth, is such that it maximizes the
value function given by:
( ) ⎥⎦
⎤
⎢⎣
= + − + + − (− ) + ⎡ − (− − ) − + − ( − )
2 2
2
1
C
C
C e B
C
e e B
C
V B C B B λ λ λ π λ α π α λ α π α α
where:
[ ( θ ) θ ] [( θ ) θ *μ ]
1
*
0 0 0 1 1 = 1+ 1− + ⋅ 1− + f f B W R R R , [ ( θ ) θ ] θ *σ
0 0 0 1 1 C =W 1+ 1− R + R ⋅ f ,
0 θ is the amount allocated in the risky asset in the first period, and 1 R is the observed
return of the risky asset in the previous period.
Observe that the value function to be maximized is close to the one of the first
period, but with changes in the parameters B and C, which account for the previous
period outcome (gain or loss). As we are interested in the investor’s risk taking behavior
after realizing a gain or a loss, let us evaluate the values of 1
θ when we change the total
return obtained in the first period. Recall that the total return from t = 0 to t = 1 ( 1 Rtot ),
depends both on his allocation choice in t = 0 and on the realized return of the risky
asset 1 R .

considering the realized return of the stock in the first
period varying over the following range: μ 2σ μ 2σ 1 − < R < + . We present numerical
results for the optimal allocation of wealth, *
1 θ
, at t = 1. The remaining parameters are
fixed (μ = 7.61%, σ = 12.98, α = 3, 1 0 W = , λ− = 2.25 and λ+ = 1). Figure 12 shows the
results. Recall that the optimal allocation in the risky asset for the first period,
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
103
considering the previous parameters, is 81%. Thus, we need to verify whether the
allocation in the stock in the second period is greater or lower than 81%, indicating
greater or lower risk appetite, respectively. First, observe that, for a total return in the
first period equal to zero (no gains/losses), the situation replicates the same framework
the investor faced in the first period. Then we reach the same optimal allocation in the
risky asset (for Rtot1 = 0 implies *

Figure 12 –Optimal equity allocation in the second period as function of the total return obtained in
the first period.
Consider the surroundings of the net value ( 0 1 Rtot = ). If the investor
experiences a gain in the first period, the model predicts that he should optimally invest
less in the risky asset in the second period. This behavior prevails up to the point where
the loss aversion effect is less pronounced. On the other hand, if a loss is observed in the
first period, he should take more risk in the following period, allocating a greater part of
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
104
his wealth in the stock. This prediction is in line with several experiments, which have
shown that disposition effect dominates house-money in dynamic settings (Weber and
Zuchel 2003). When the investor experiences a gain in the first period, he tends to
reduce his risk appetite in order to guarantee the previous outcome. On the other hand,
if he experiences a loss in the first period, he will increase his bets on stocks, trying to
avoid the previous loss. In the model, the pattern holds for the whole gain domain;
however, in the loss domain, high losses in the first period induce less risk appetite in
the second period. The intuition is that if the investor is facing a huge loss, the loss
aversion effect will dominate the risk seeking behavior, inducing a reduction in the
optimal allocation in the stocks.
When we evaluate the expected cost (Eq. 10) of the behavioral inefficiency in
the second period as a function of the return of the risky asset in the first period (Figure
13), it is possible to observe that, depending on the previous outcome, the cost can be
increasing or decreasing. If the value for 1 R is such that it implies a small loss in the
first period, the cost is even negative, which means that the expected return in the
second period under prospect theory is greater than the one associated with standard
utility. This is related to a greater risk appetite of the prospect theory individual after a
loss, implying a greater allocation in the stock, which has a greater expected return. If
1 R indicates a gain in the first period, then the cost is positive once the allocation in the
stock for the standard utility investor is greater than for the prospect utility individual.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
105
-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
-0.01
0
0.01
0.02
0.03
0.04
0.05
Cost 2
R1
Figure 13 –Expected cost in the second period as function of the equity return obtained in the first
period.
We can conclude that for losses in the first period, the optimal allocation should
adapt to the individual’s biases to reach better performance as the cost comes out to be
negative in this domain. For gains in the previous outcome, the allocation should
moderate the biases (observe a positive value for the expected cost). For extreme losses
in the first period, the allocation should also moderate the investor’s biases.
If we accumulate the cost results in periods 1 and 2, we get the graph
represented in Figure 14. It indicates that, for a negative stock result in the first period,
or even a slightly positive one, the prospect theory individual outperforms the standard
utility investor. And so, the allocation strategy should be adapted to the individual
biases. The previous results should be taken with care as they refer to expected values.
In section 3, we provide a more robust comparison, taking into account the performance
of those individuals in an out-of-sample analysis.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
106
-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
Acum Cost
R1
Figure 14 –Expected cumulative cost in the second period as function of the equity return obtained
in the first period.
4.2.3 Multi-Period Analysis
If we extend the two-period analysis to a multi-period one, by analogy, if the
investor considers his current wealth as the reference to which he measures his
gains/losses, he will solve the same maximization problem for each period and the
optimal asset allocation is given as in proposition 1. In this situation, the agent acts
myopically, just considering the following period possible gain/loss. In general, this
result implies a smaller stock allocation if compared to a standard utility investor,
generating an expected cost associated to the prospect theory biases.
On the other hand, if the individual’s reference point is his initial wealth (or his
wealth in some moment in time t = t1), the allocation is defined as in proposition 3, but
now considering the previous outcome as the total return obtained by him from t = 0 (or
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
107
from t = t1) to the current time. As discussed in the two-period model, the allocation in
the risky asset will depend on the previous gains/losses, and can be greater or smaller
than the one chosen by the standard utility investor. Observe that the standard utility
investor always chooses the same allocation in the risky asset, no matter what the
reference point, as neither his decisions nor his beliefs are affected by previous
outcomes.
4.2.4 Resampling
In sections 4.2.1, 4.2.2 and 4.2.3 we already evaluated the optimal asset
allocation under prospect theory preferences and considering mental accounting, loss
aversion, asymmetric risk taking behavior, and probability weighting. However, there is
still an important issue in portfolio optimization missing: estimation error. Up to now,
when solving the investor’s problem, we considered the expected return known with
certainty, which is not the case in reality (especially in emerging markets where the
uncertainty is higher). The assumed return for the risky asset is just an estimate, and so
the real value can be different. This problem is relevant in any model of portfolio
optimization and is crucial under prospect theory, where for lower values of the riskfree
rate, a slightly increase in the risk premium of stocks can lead to extreme
allocations. If the real return of the risky asset is lower, the likelihood of facing a loss is
greater and should significantly reduce the value of that prospect.
In an attempt to overcome this estimation problem, Michaud (1998) proposed
the resampling technique. Portfolio sampling allows an analyst to visualize the
estimation error in traditional portfolio optimization methods. Suppose that we
estimated both the variance and the excess return by using N observations. It is
important to note that the point estimates are random variables and so another sample of
the same size from the same distribution would result in different estimates.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
108
Sherer (2002) suggests that sampling from a multivariate normal distribution (a
parametric method termed Monte Carlo simulation) is a way to capture the estimation
error. In this sense, return and variance would just be the expected values for a
multivariate normal distribution. If we just consider two assets, the probability density
function for a multivariate normal distribution would be given by
.
By repeating the sampling procedure n times, we get n new sets of optimization
inputs, and then a different efficient allocation. The resampled weight for a portfolio
would then be given by
Σ=
=
n
n i 1
Resamp 1
i θ
θ
The resampled portfolios should reflect a greater diversification (more assets
enter in the solution) than the classical mean-variance efficient portfolio, and should
also exhibit less sudden shifts (smooth transitions) in allocations as return requirements
change. Both characteristics are desirable for investors.
Recent literature has shown unambiguous results in favor of resampled
portfolios in out of sample analysis (Pawley, 2005; Markowitz and Usmen, 2003; Wolf,
2006; Jiao, 2003). However, Harvey et al. (2006), evaluating Bayes vs. resampling
methods, posit that the choice of risk aversion drives the results. Kohli (2005) concludes
that, despite the fact that there are no conclusive advantages or disadvantages of using
resampling as a technique to obtain better returns, resampled portfolios do seem to offer
higher stability and lower transaction costs, two crucial features for long term investors’
choices.
We then propose the BRATE (Behavior Resampling Technique) as a novel
methodology to define asset allocation, which incorporates behavioral ideas and
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
109
resampling techniques into portfolio optimization, thus adapting to the individual’s
preferences. In this case, the optimal asset allocation should be given by the previous
propositions (1 and 2 or 3, depending on the reference point), but the procedure should
be performed several times for different expected stock returns (given by a multivariate
normal distribution). The population allocation is then given by the expected risky asset
allocation. The procedure can be summarized as follows43:
Step 1: Estimate variance-covariance and return from the historical inputs.
Step 2: Resample from inputs (created in Step 1) by taking n draws from the
input distribution. The number of draws reflects the degree of uncertainty in
the inputs. Calculate new variance-covariance and return from sampled
series. Estimation error will result in estimations that are different from
those obtained in Step 1.
Step 3: Calculate the optimal allocation for inputs defined in Step 2, using
the appropriate propositions (1 and 2 or 3, depending on the reference point
considered).
Step 4: After repeating Steps 2 and 3 many times, calculate average
portfolio weights. This is the BRATE portfolio allocation.
In the next section, we provide an empirical analysis comparing the BRATE
allocation performance to a standard utility allocation.
4.3. Empirical Study
43 This methodology is an adaptation of the one proposed in Michaud (1998).
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
110
4. 3. 1. Data and Implementation
Our tests are based considering daily data from 26 countries’ MSCI stock
indices and risk free rates, plus the MSCI World Index, for the period from April 4th,
1995 to January 5th, 2007. Developed countries and emerging markets (Brazil, Chile,
South Africa, South Korea, Taiwan, Thailand, Turkey) were included in the analysis in
order to find generalizable results. The total return time series are calculated on each
country’s currency and also in US-Dollars. Thus, we are considering both hedged and
unhedged investors. Table 1 presents some descriptive statistics of each market
considered, for the whole sample period.
Table 1 – Descriptive Statistics
This Table provides descriptive statistics for the sample of world markets. For each market we present,
the average risk free rate, the mean, standard deviation, skewness, and kurtosis of stock returns, as well as
the Sharpe index (annualized values). The values are presented in the countries’ currency and also in
USD. The risk free rate used to calculate the Sharpe Index in USD is the 3 month UST Bill rate for all
markets.
Risk Free Mean Std. Skew Kurt Sharpe Index Mean Std. Skew Kurt Sharpe Index
T-Bill 3 month 2.722 2.722 0.076 -0.574 1.787 0.000 2.722 0.076 -0.574 1.787 0.000
'Australia' 3.856 8.971 13.022 -0.322 6.685 0.392 10.105 17.275 -0.125 6.389 0.428
'Austria' 1.537 11.844 15.790 -0.574 7.277 0.652 12.020 17.816 -0.319 5.626 0.522
'Belgium' 2.218 10.811 18.065 0.317 9.921 0.476 10.886 19.295 0.163 7.275 0.423
'Brazil' 16.531 23.184 30.022 0.972 23.667 0.222 17.590 34.552 0.035 8.387 0.430
'Canada' 2.722 11.516 16.686 -0.426 8.866 0.527 13.230 18.337 -0.532 8.046 0.573
'Chile' 2.470 7.636 15.309 0.166 7.188 0.337 5.670 18.370 -0.067 6.509 0.161
'Denmark' 2.218 14.767 17.256 -0.321 5.778 0.727 14.540 17.973 -0.281 5.292 0.657
'Finland' 2.293 21.269 37.629 -0.162 9.041 0.504 21.118 37.650 -0.101 9.202 0.488
'France' 2.243 11.516 20.800 -0.048 5.926 0.447 11.416 21.048 -0.012 5.332 0.413
'Germany' 2.772 10.987 23.172 -0.138 6.244 0.355 10.861 23.232 -0.097 5.337 0.350
'Ireland' 2.848 10.282 17.956 -0.528 8.877 0.414 10.458 19.676 -0.304 6.763 0.392
'Italy' 2.974 10.710 20.213 -0.064 6.000 0.383 10.660 20.726 -0.032 5.237 0.383
'Japan' 0.151 4.536 19.215 0.051 5.152 0.229 2.570 22.234 0.332 6.647 -0.008
'Netherlands' 2.092 10.156 21.489 -0.076 7.018 0.375 10.004 21.551 -0.006 6.177 0.338
'Norway' 3.326 12.121 19.635 -0.304 6.706 0.449 12.172 20.767 -0.318 7.104 0.454
'Portugal' 2.923 9.727 15.801 -0.261 8.097 0.430 9.878 17.664 -0.051 5.834 0.405
'SouthAfrica' 7.938 12.625 19.769 -0.437 9.002 0.237 8.039 24.810 -0.429 7.053 0.214
'SouthKorea' 2.318 12.676 34.524 0.271 6.664 0.300 13.709 41.728 1.336 26.151 0.263
'Spain' 2.797 16.405 21.118 -0.078 6.249 0.644 16.405 21.781 0.031 5.682 0.628
'Sweden' 2.696 15.473 24.896 0.187 6.700 0.513 16.380 26.850 0.120 6.322 0.509
'Switzerland' 1.058 12.197 18.051 -0.106 7.639 0.617 11.441 17.713 0.010 6.549 0.492
'Taiwan' 3.251 4.687 26.244 0.149 5.442 0.054 3.150 27.563 0.110 5.505 0.016
'Thailand' 3.654 0.076 32.851 1.415 17.779 -0.109 -1.865 36.002 0.984 13.281 -0.128
'Turkey' 39.514 47.804 45.171 0.324 8.017 0.184 21.521 50.887 0.219 8.094 0.369
UnitedKingdom' 3.704 6.779 16.476 -0.153 6.225 0.187 8.392 17.037 -0.100 5.213 0.332
'UnitedStates' 2.722 9.904 16.978 -0.024 6.598 0.422 9.904 16.978 -0.024 6.598 0.422
World Index 2.722 7.610 12.976 -0.144 5.763 0.376 7.610 12.976 -0.144 5.763 0.376
Currency USD
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
111
From the table, we verify a risk premium associated with the stock market, both
considering the values in each country’s currency and in USD, with the mean return of
stocks being higher than the one of the corresponding riskfree rate44.
Let us first consider the values in each country’s currency. The average
annualized return of the risk free rate varied from 0.151% (Japan) to 39.514% (Turkey),
while for the stock index, it ranges from 0.076% (Thailand) to 47.804% (Turkey). The
annualized volatility (standard deviation) of the stock market varied from 12.976%
(World Index) to 45.171% (Turkey). As expected, emerging markets tend to be more
volatile than developed markets. While in Brazil, South Korea, Thailand, and Turkey
the volatility was above 30 %, in countries like United Kingdom and United States, its
value was close to 16%. In terms of skewness and kurtosis, usual results appear,
indicating that daily stock index returns are negative skewed and have excess kurtosis
(greater than 3). Finally, Table 1 presents the annualized Sharpe Index, which was
greater in developed markets (around 0.35) than emerging markets (0.19). Our results
are in line with previous literature (Mehra, 2003) which gives 0.34 as an estimation of
the long-term Sharpe Ratio for the U.S. economy.
When we consider the values in USD, say in the perspective of a US based
international investor who doesn’t currency hedge his investments, we find similar
results. The average daily return in USD is close to the one in the country’s currency,
which is evidence of the mean reverting aspect of the foreign exchange market.
However, the standard deviation in USD is slightly greater than the one in the country’s
currency, as the former includes both stock market risk and currency risk (the volatility
of the foreign exchange rate). In terms of skewness and kurtosis, the previous results
remain. However, now the Sharpe Indexes do not present relevant differences among
44 The only exception is Thailand where the sharpe index is negative (-0.109).
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
112
emerging and developed markets (for instance it is 0.430 for Brazil and 0.422 for the
United States). Thus it seems that emerging stock markets are less interesting for
domestic investors than for foreign unhedged investors.
Next we analyze the performance of the following optimization strategies: an
investor with a standard utility preference - STU; an investor with prospect utility
preference, with reference point given by his current wealth – PTU; an investor with
prospect utility preference, with reference point given by his wealth in the previous
period – CPT; an investor with a standard utility preference (resampled) – RSTU; an
investor with prospect utility preference, with reference point given by his current
wealth (resampled) – BRATEa; and an investor with prospect utility preference, with
reference point given by his wealth in the previous period (resampled) – BRATEb. The
utility function parameters are fixed (α = 3, λ− = 2.25 and λ+ = 1). We vary the
estimation period (p) in an out of sample analysis. The parameters are estimated using
daily return observations of the past p days. We define the efficient portfolio and hold it
for the next (e) months, then re-estimate the parameters and adjust the portfolio weights.
To judge the financial performance of the strategies, we compute their average return
and empirical Sharpe-Ratios.
4. 3. 2. Results
The Sharpe Ratios of the different strategies are presented in Table 2 for the
World Index and for the total period from 1995 to 2007, considering p = 6 months, 1, 2,
and 4 years, and e varying from 2 months to 1 year. We are evaluating the different
strategies for a US based international stock investor. The risk free rate considered was
the 3 month T Bill.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
113
Table 2 – Sharpe Ratios
This Table presents the Sharpe-Ratio of the efficient portfolio generated by each estimation model. The
Sharpe-Ratio is calculated by dividing the excess return observed by the standard deviation.
STU PTU CPT RSTU BRATEa BRATEb
6m-2m 0.189 0.134 0.136 0.207 0.154 0.156
6m-6m 0.101 0.080 0.083 0.125 0.102 0.114
1y-6m 0.439 0.392 0.392 0.438 0.400 0.401
2y-6m 0.462 0.426 0.421 0.464 0.434 0.423
4y-6m -0.135 -0.023 -0.023 -0.122 -0.018 -0.019
1y-1y 0.413 0.347 0.389 0.420 0.354 0.393
2y-1y 0.456 0.428 0.431 0.461 0.444 0.465
4y-1y -0.206 -0.126 -0.126 -0.193 -0.114 -0.113
mean 0.215 0.207 0.213 0.225 0.219 0.227
In general, we can state that the resampled models offered better results for a
short selling constrained investor. It is an expected result as resampled models take into
account the estimation risk, generating a more diversified portfolio which tends to
outperform in out of sample studies. The highest Sharpe ratio was reached by the
BRATEb model for an estimation period of 2 years and evaluation period of 1 year
(0.465). On average resampled models increase the Sharpe ratio in around 0.10, when
compared to the deterministic ones. Also, while the (R)STU investor seems to
outperform (R)PTU, it doesn’t happen with (R)CPT.
If we consider just the total return obtained by each strategy, we find the results
presented in Table 3. In this case, it’s possible to infer an inefficiency cost related to the
behavioral investors, who tend to underperform the results of the standard utility
investor in around 10 bps45. However if take into account the increment in risk (a risk
adjusted measure like Shape Ratio), the inefficiency disappears.
45 1 bps = 0.01%.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
114
Table 3 – Average Total Return
This Table presents the Average Total Return of the efficient portfolio generated by each estimation
model.
STU PTU CPT RSTU BRATEa BRATEb
6m-2m 4.302 3.781 3.822 4.447 3.935 3.976
6m-6m 3.654 3.449 3.476 3.883 3.643 3.749
1y-6m 6.670 6.211 6.211 6.632 6.238 6.242
2y-6m 7.377 6.987 6.875 7.345 6.910 6.775
4y-6m 1.065 2.083 2.083 1.064 1.992 1.993
1y-1y 6.711 5.981 6.295 6.630 5.979 6.341
2y-1y 7.247 6.935 6.966 7.289 7.068 7.194
4y-1y 0.419 1.226 1.226 0.530 1.287 1.297
mean 4.681 4.582 4.619 4.727 4.631 4.696
Based on the previous results, we can state that resampled models tend to
outperform traditional models. Also, there is no clear advantage of standard utility
investors over behavioral prospect theory investors at least to the CPT investor. Levy
and Levy (2004) reached a similar result, positing that the practical differences between
prospect theory and traditional mean-variance theory are minor. In this sense,
behavioral biases should not be moderated, nor should standard models be adapted to
include behavioral biases.
When we take into account each market separately, we find the results presented
in Table 4 (in each country’s currency). Considering each country individually, there’s
no clear dominance of a single strategy. Resampled models tend to outperform
traditional models in emerging markets (observe the results for Brazil, Chile, South
Africa, South Korea, Taiwan, Thailand and Turkey), where the uncertainty over the
risk/return estimation is higher.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
115
Table 4 – Sharpe Ratios
This Table presents the Sharpe-Ratio of the efficient portfolio generated considering an estimation period
of 1 year and evaluation period of 6 months (in each country’s currency). The Sharpe-Ratio is calculated
by dividing the excess return observed by the standard deviation.
STU PTU CPT RSTU BRATEa BRATEb
'Australia' 0.309 0.352 0.353 0.309 0.346 0.345
'Austria' 0.629 0.578 0.584 0.618 0.593 0.597
'Belgium' 0.977 0.982 0.973 0.982 0.986 0.977
'Brazil' 0.323 0.326 0.304 0.335 0.333 0.317
'Canada' 0.490 0.490 0.490 0.490 0.488 0.489
'Chile' 0.729 0.726 0.721 0.735 0.740 0.736
'Denmark' 0.914 0.914 0.914 0.908 0.910 0.909
'Finland' 0.696 0.685 0.638 0.691 0.658 0.665
'France' 0.778 0.790 0.755 0.780 0.785 0.764
'Germany' 0.619 0.614 0.619 0.619 0.616 0.616
'Ireland' 0.615 0.607 0.636 0.626 0.607 0.634
'Italy' 0.737 0.769 0.740 0.733 0.753 0.726
'Japan' 0.041 0.080 0.042 0.057 0.051 0.040
'Netherlands' 0.657 0.655 0.657 0.657 0.654 0.655
'Norway' 0.389 0.368 0.368 0.402 0.398 0.398
'Portugal' 0.751 0.685 0.728 0.764 0.716 0.738
'SouthAfrica' 0.161 0.206 0.218 0.167 0.208 0.224
'SouthKorea' 0.101 0.019 0.035 0.111 0.066 0.058
'Spain' 0.932 0.949 0.954 0.930 0.936 0.936
'Sweden' 0.634 0.631 0.631 0.643 0.634 0.633
'Switzerland' 0.773 0.720 0.739 0.773 0.739 0.748
'Taiwan' -0.001 -0.003 0.000 -0.004 -0.005 -0.004
'Thailand' 0.041 -0.012 -0.048 0.055 0.033 0.018
'Turkey' 0.183 0.189 0.094 0.185 0.190 0.103
UnitedKingdom' 0.411 0.428 0.423 0.411 0.429 0.426
'UnitedStates' 0.618 0.624 0.626 0.615 0.623 0.616
World Index' 0.439 0.392 0.392 0.438 0.400 0.401
In terms of the comparison between the standard and the prospect utility
investor, generally the former doesn’t outperform the latter, indicating no clear
dominance of the traditional rational model. In this sense, there is no need for
moderating the behavioral biases as described by prospect theory, as no extra financial
efficiency is gained.
Generally speaking, an interesting finding is the fact that all previous allocation
models outperform the 100% risky strategy. The Sharpe ratio of the 100% stock
strategy was 0.383 while all resampled models reached, on average, a result above
0.50946.
46 A t-test over the Sharpe Ratio differences offered a significant result with a p-value of 0.0001.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
116
Finally, if we take into account the values in USD and so considering that the
investor is facing foreign exchange risk, we reach the results presented in Table 5.
Table 5 – Sharpe Ratios
This Table presents the Sharpe-Ratio of the efficient portfolio generated considering an estimation period
of 1 year and evaluation period of 6 months (values in USD). The Sharpe-Ratio is calculated by dividing
the excess return observed by the standard deviation.
STU PTU CPT RSTU BRATEa BRATEb
'Australia' 0.589 0.567 0.564 0.591 0.595 0.591
'Austria' 0.896 1.015 1.015 0.882 0.971 0.977
'Belgium' 0.904 0.904 0.904 0.886 0.886 0.899
'Brazil' 0.656 0.653 0.653 0.669 0.675 0.668
'Canada' 0.421 0.421 0.421 0.423 0.420 0.420
'Chile' 0.752 0.757 0.757 0.774 0.776 0.775
'Denmark' 0.781 0.754 0.820 0.776 0.773 0.803
'Finland' 0.612 0.596 0.597 0.613 0.597 0.593
'France' 0.654 0.643 0.625 0.649 0.625 0.629
'Germany' 0.533 0.509 0.521 0.537 0.502 0.516
'Ireland' 0.654 0.593 0.600 0.641 0.620 0.618
'Italy' 0.674 0.614 0.640 0.681 0.638 0.664
'Japan' 0.195 0.181 0.194 0.198 0.175 0.182
'Netherlands' 0.645 0.655 0.656 0.645 0.653 0.654
'Norway' 0.351 0.387 0.387 0.361 0.381 0.380
'Portugal' 0.666 0.641 0.641 0.669 0.647 0.660
'SouthAfrica' 0.465 0.441 0.456 0.479 0.459 0.460
'SouthKorea' 0.226 0.222 0.189 0.233 0.230 0.191
'Spain' 0.858 0.899 0.899 0.862 0.892 0.894
'Sweden' 0.562 0.558 0.566 0.563 0.554 0.565
'Switzerland' 0.599 0.531 0.552 0.596 0.607 0.612
'Taiwan' -0.090 -0.100 -0.086 -0.083 -0.095 -0.076
'Thailand' 0.104 0.040 0.030 0.114 0.108 0.096
'Turkey' 0.288 0.250 0.238 0.296 0.267 0.246
UnitedKingdom' 0.625 0.574 0.605 0.632 0.597 0.626
'UnitedStates' 0.618 0.624 0.626 0.612 0.615 0.612
World Index' 0.439 0.392 0.392 0.438 0.400 0.401
Again, the results indicate a dominance of resampled models in emerging
markets, while for developed countries, no clear dominance can be seen. The traditional
rational model does not outperform the behavioral ones. Finally, all six dynamic models
add value for the investor when compared to a 100% stock invested individual. Observe
that the Sharpe Ratio found for the different markets (both in the country’s currency and
in USD) are notably higher than the ones presented in Table 1.
Summing up, resampled models, which take into account estimation risk, tend to
outperform deterministic models, notably for emerging markets where the uncertainty
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
117
of the expected return estimation is higher. Moreover, prospect theory utility investors
don’t reach worse returns if compared to the traditional rational ones, which indicate no
need for addressing bias moderation in the portfolio allocation.
4. 4. Conclusions
This study had two objectives: first to incorporate mental accounting, loss
aversion, asymmetric risk-taking behavior, and probability weighting in portfolio
optimization for individual investors; and second to take into account the estimation risk
in the analysis.
Considering daily index stock data from 26 countries over the period from 1995
to 2007, we empirically evaluated our model (BRATE – Behavior Resample
Technique) against the traditional Markowitz. Several estimation and evaluation periods
were used and we also considered a foreign exchange hedged and an unhedged strategy.
Our results support the use of BRATE as an alternative for defining optimal
asset allocation and posit that a portfolio optimization model may be adapted to the
individual biases implied in prospect theory. Behavioral biases don’t seem to reduce
efficiency when we consider a dynamic setting. This result is robust for different
developed and emerging markets. Also, the previous optimization models add value for
the individual investor when compared to a naive 100% risky strategy.
As further extensions of the present research, we suggest the inclusion of several
risky assets in the analysis. In this case, the issue of multiple mental accounting is a
crucial issue to address the problem. An investor who evaluates every security in their
own mental account will not necessarily view additional securities as redundant, which
dramatically increases the complexity of the problem.
Chapter Four: Behavior Finance and Estimation Risk in Portfolio Optimization
118
We also leave unasked the question of how individuals arrive at the underlying
return distribution. That is the model above is a proposed mechanism for how
individuals might transform a given probability distribution (assumed to be an accurate
representation of the underlying distribution) into decision weights. Once we introduce
uncertainty, it can induce individual biases, subjectivity and error. There is evidence
that people display considerable overconfidence when asked to provide a subjective
assessment of a probability distribution47. Moreover, it is questionable whether the
weightings provided by CPT truly reflect the process by which individuals evaluate
continuous probability distributions.
The agent who measures his gains and losses always relative to his actual wealth
solves the same maximization problem each period, therefore selecting a fix-mix
strategy. An open question remains, if a fix-mix strategy can be the cause of the
disposition effect.
47 Their subjective distribution is too tightly centered on their estimated mean.
119
Chapter Five
General Conclusions, Contributions and Lines for Further
Research
In this thesis, we assumed as invariants of human behavior: loss aversion;
asymmetric risk taking behavior; probability weighting and mental accounting. Under
the previous theoretical framework, we evaluated three financial contexts: the role of
incentives in risk taking decisions (Chapter 2); the effect of prior outcomes in future
risk decisions (Chapter 3); and finally the portfolio choice problem (Chapter 4).
Our main conclusion is that absolute evaluations based on final wealth are
limited and the relativity of risk taking decisions, where the perception of gains or
losses drives the process, is a requirement to understand individual’s decisions. As a
reply to behavioral critics, we reach propositions that can be falsified and make several
predictions. Our contributions are both on the theoretical and empirical streams of
finance literature.
In Chapter 2 we formulated five propositions linking investment strategy,
compensation and risk taking in professional investor’s context. We suggested that
managers in passive managed funds tend to be rewarded without incentive fee and be
Chapter Five: General Conclusions
120
risk averse. On the other hand, in active managed funds, whether incentives will reduce
or increase the riskiness of the fund will depend on how hard is to outperform the
benchmark. If the fund is (un)likely to outperform the benchmark, incentives (increase)
reduce the manager’s risk appetite. Furthermore, the evaluative horizon influences the
trader’s risk preferences, in the sense that if traders performed poorly (well) in a period,
they tend to choose riskier (conservative) investments in the following period given the
same evaluative horizon. We tested the model in an empirical analysis over a sample of
4684 mutual equity funds (cross-section data), from developed and emerging markets,
and we reached supportive results to the propositions established in the theoretical
model.
In Chapter 3, we performed a cross-country (Brazil and Spain) experiment on
how prior outcomes affects risk taking decisions trying to elucidate the controversial
result found in house-money and loss aversion papers. Our results support the existence
of MLA over the countries but a country effect is also found, indicating that care should
be taken when generalizing behavioral findings over the international financial market.
No salient sex or age effect was found in the sample; however, some cultural
differences among countries seem to have influence on the risk taking decision.
Disposition effect is found to dominate the risk-taking behavior of subjects in dynamic
settings, overcoming the house-money effect. Subjects that experienced a gain (loss)
tend to assume less (more) risk in the following period.
Chapter 4 had two objectives: first to incorporate mental accounting, loss
aversion, asymmetric risk-taking behavior and probability weighting in portfolio
optimization for individual investors; and second to take into account the estimation risk
in the analysis. Considering daily index stock data, from 26 countries for the period
from 1995 to 2007, we empirically evaluated our model (BRATE – Behavior Resample
Chapter Five: General Conclusions
121
Technique) to the traditional Markowitz. Our results support the use of BRATE as an
alternative for defining optimal asset allocation and posit that a portfolio optimization
model should be adapted to the individual biases implied in prospect theory. Behavioral
biases don’t seem to reduce efficiency when we consider a dynamic setting. This result
is robust for different developed and emerging markets.
Several extensions to the present research can be mentioned. Overall we
considered several invariants of human behavior based on prospect theory. However,
the impact of cognitive biases in agent’s risk preference still needs to be better
understood, in order to understand the way psychological states may affect risk
preferences in this context.
In terms of the delegated portfolio management model and the influence of
incentives in the risk taking behavior of incentives, further research may include the
type of financial institution the trader works for (banks, insurance companies, pension
funds) to take into account regulatory and institutional effects. Also, delegated portfolio
management often involves more than one layer of agency: how does this feature affect
incentives? More generally, studies about general equilibrium implications and price
impact of the agency aspects of professional portfolio management should be interesting
especially for policy-makers, given the relevance of these funds in all developed
financial market.
Also, in the compensation analysis, only financial compensations were
considered and we think that including non-financial rewards like recognition and
prestige would enrich the theory and enable better predictions. Finally, inclusion of
career concerns in the model could also improve multi-period analysis. Kempf et al.
(2007) suggest that when employment risk is high, managers that lag behind tend to
decrease risk relative to leading managers in order to prevent potential job loss.
Chapter Five: General Conclusions
122
Related to the experiment performed in Chapter 3 and the influence of prior
outcomes in risk taking behavior, as possible extensions we suggest to go further in the
investigation of what country's cultural characteristics are generating the effect found in
the research, and also the use of other countries’ students replicating the experimental
design C, since we have found that there seems to be a country effect as well, which
could be due to cultural differences.
Finally, related to Chapter 4 and the portfolio choice problem, as further
extensions of the presented research, we suggest the inclusion of several risky assets in
the analysis. In this case, the issue of multiple mental accounting is a crucial issue to
address the problem. An investor who evaluates every security in its own mental
account will not necessarily view additional securities as redundant. Also a deeper
discussion of how individuals access returns’ underlying distribution is another crucial
point which can be explored in the future.
Summing up, as can be seen, behavioral finance is an open avenue where several
studies are still missing. We forecast that in the following years, behavioral models will
be the rule and not the exception, and doing so, economists will be closer to the
complexity of social reality, and their theories will explain much better what is going on
in financial markets. As mentioned by Thaler (2000), it is the evolution from homo
economicus to homo sapiens.
Appendix A: Experiment Instructions (Chapter 3)
123
Appendix A: Experiment Instructions (Chapter 3)
(Translated from Portuguese and Spanish)
Introduction [Read aloud only]
Welcome to our experimental study of decision-making. The experiment will
last about 30 minutes. The instructions for the experiment are simple, and if you follow
them carefully, you can earn a considerable amount of money. All the money you earn
is yours to keep, and will be paid to you, privately and in cash, immediately after the
experiment.
Before we start the experiment you will be asked to pick one envelope from this
pile. In the envelope you will find your Registration Form. This form will be used to
register your decisions and earnings. On the top of your Registration Form you will find
your registration number. This number indicates behind which table you are to take a
seat. When everyone is seated, we will go through the instructions of the experiment.
After that, you will have the opportunity to study the instructions on your own, and to
ask questions. If you have a question, please raise your hand, and I will come to your
table. Please do not talk or communicate with the other participants during the
experiment.
Are there any questions about what has been said until now? If not, then will the
person on my left please be the first to pick an envelope, open it, and take the
corresponding seat.
[Treatment F: Read aloud and distributed]
The experiment consists of 12 successive rounds. In each round you will start
with an amount of 100 cents. You must decide which part of this amount (between 0
cents and 100 cents) you wish to bet in the following lottery.
Appendix A: Experiment Instructions (Chapter 3)
124
You have a chance of 2/3 (67%) to lose the amount you bet and a chance of 1/3 (33%)
to win two and a half times the amount bet.
You are requested to record your choice on the Registration Form. Suppose that
you decide to bet an amount of X cents (0 ≤ X ≤ 100) in the lottery. Then you must fill
in the amount X in the column headed Amount in lottery, in the row with the number of
the present round.
Whether you win or lose in the lottery depends on your personal win color. This
color is indicated on the top of your Registration Form. Your win color can be Red,
Yellow or Green, and is the same for all twelve rounds. In any round, you win in the
lottery if your win color matches the round color that will be randomly selected and
announced, and you lose if your win color does not match the round color.
The round color is determined as follows. After you have recorded your bet in
the lottery for the round, the computer will randomly select one color that will be shown
in the screen. If the round color matches your win color, you win in the lottery;
otherwise you lose. Since there are three colors, one of which matches your win color,
the chance of winning in the lottery is 1/3 (33%) and the chance of losing is 2/3 (67%).
Hence, your earnings in the lottery are determined as follows. If you have
decided to put an amount X cents in the lottery, then your earnings in the lottery for the
round are equal to - X if the round color does not match your win color (you lose the
amount bet), and equal to (2.5X) if the round color matches your win color.
The round color will be shown in the screen and announced by the assistant. You
need to record this color in the column Round color, under win or lose, depending on
whether the round color does or does not match your win color. Also, you need to
record your earnings in the lottery in the column Earnings in lottery. Your total earnings
for the round are equal to 100 cents (your starting amount) plus your earnings in the
Appendix A: Experiment Instructions (Chapter 3)
125
lottery. These earnings are recorded in the column Total Earnings, in the row of the
corresponding round. Each time we will come by to check your Registration Form.
After that, you are requested to record your choice for the next round. Again you
start with an amount of 100 cents, a part of which you can bet in the lottery. The same
procedure as described above determines your earnings for this round. It is noted that
your private win color remains the same, but that for each round the computer selects a
new color randomly. All subsequent rounds will also proceed in the same manner. After
the last round has been completed, your earnings in all rounds will be totaled. This
amount determines your total earnings in the experiment, which you will receive in cash
after the numbers are checked.
[Treatment I: Read aloud and distributed]
The experiment consists of 12 successive rounds. In each round you will start
with an amount of 100 cents. You must decide which part of this amount (between 0
cents and 100 cents) you wish to bet in the following lottery.
You have a chance of 2/3 (67%) to lose the amount you bet and a chance of 1/3 (33%)
to win two and a half times the amount bet.
You are requested to record your choice on the Registration Form. Suppose that
you decide to bet an amount of X cents (0 ≤ X ≤ 100) in the lottery. Then you must fill
in the amount X in the column headed Amount in lottery. Please note that you fix your
bet for the next three rounds. Thus, if you decide to bet an amount X in the lottery for
round 1, then you also bet an amount X in the lottery for rounds 2 and 3. Therefore,
three consecutive rounds are joined together on the Registration Form.
Whether you win or lose in the lottery depends on your personal win color. This
color is indicated on the top of your Registration Form. Your win color can be Red,
Yellow or Green, and is the same for all twelve rounds. In any round, you win in the
Appendix A: Experiment Instructions (Chapter 3)
126
lottery if your win color matches the round color that will be randomly selected and
announced, and you lose if your win color does not match the round color.
The round color is determined as follows. After you have recorded your bet in
the lottery for the next three rounds, the computer will randomly select one color that
will be shown in the screen for each of the next three rounds. If the round color matches
your win color, you win in the lottery; otherwise you lose. Since there are three colors,
one of which matches your win color, the chance of winning in the lottery is 1/3 (33%)
and the chance of losing is 2/3 (67%).
Hence, your earnings in the lottery for the three rounds are determined as
follows. If you have decided to put an amount X cents in the lottery, then your earnings
in the lottery for the round are equal to - X if the round color does not match your win
color (you lose the amount bet), and equal to (2.5X) if the round color matches your win
color.
The three round colors will be shown in the screen and announced by the
assistant. You need to record this color in the column Round color, under win or lose,
depending on whether the round color does or does not match your win color. Also, you
need to record your earnings in the lottery in the column Earnings in lottery. Your total
earnings for the three rounds are equal to 300 cents (your starting amount) plus your
earnings in the lottery. These earnings are recorded in the column Total Earnings, in the
row of the corresponding round. Each time we will come by to check your Registration
Form.
After that, you are requested to record your choice for the next three rounds (4-
6). For each of the three rounds you again start with an amount of 100 cents, a part of
which you can bet in the lottery. The same procedure as described above determines
your earnings for these three rounds. It is noted that your private win color remains the
Appendix A: Experiment Instructions (Chapter 3)
127
same, but that for each round the computer selects a new color randomly. All
subsequent rounds will also proceed in the same manner, also grouped by three (i.e., 7-9
and 10-12). After the last round has been completed, your earnings in all rounds will be
totaled. This amount determines your total earnings in the experiment, which you will
receive in cash after the numbers being checked.
[Treatment C: Read aloud and distributed]
The experiment consists of 12 successive rounds. In each odd round (1, 3, 5, 7, 9
and 11) you will start with an amount of 100 cents. In each even round (2, 4, 6, 8, 10
and 12) you will start with an amount of 100 cents plus your Total Earnings in the
previous round. You must decide which part of this amount (between 0 cents and 100
cents for odd rounds; and between 0 cents and 100 cents plus your Total Earnings in the
previous round for even rounds) you wish to bet in the following lottery.
You have a chance of 2/3 (67%) to lose the amount you bet and a chance of 1/3 (33%)
to win two and a half times the amount bet.
You are requested to record your choice on the Registration Form. Suppose that
you decide to bet an amount of X cents in the lottery. Then you must fill in the amount
X in the column headed Amount in lottery, in the row with the number of the present
round.
Whether you win or lose in the lottery depends on your personal win color. This
color is indicated on the top of your Registration Form. Your win color can be Red,
Yellow or Green, and is the same for all twelve rounds. In any round, you win in the
lottery if your win color matches the round color that randomly selected and
announced, and you lose if your win color does not match the round color.
The round color is determined as follows. After you have recorded your bet in
the lottery for the round, the computer will randomly select one color that will be shown
Appendix A: Experiment Instructions (Chapter 3)
128
in the screen. If the round color matches your win color, you win in the lottery;
otherwise you lose. Since there are three colors, one of which matches your win color,
the chance of winning in the lottery is 1/3 (33%) and the chance of losing is 2/3 (67%).
Hence, your earnings in the lottery are determined as follows. If you have
decided to put an amount X cents in the lottery, then your earnings in the lottery for the
round are equal to - X if the round color does not match your win color (you lose the
amount bet), and equal to (2.5X) if the round color matches your win color.
The round color will be shown in the screen and announced by the assistant. You
need to record this color in the column Round color, under win or lose, depending on
whether the round color does or does not match your win color. Also, you need to
record your earnings in the lottery in the column Earnings in lottery. Your total earnings
for the round are equal to 100 cents in odd rounds; or 100 cents plus your Total
Earnings in the previous round for even rounds (your starting amount); plus your
earnings in the lottery. These earnings are recorded in the column Total Earnings, in the
row of the corresponding round. Each time we will come by to check your Registration
Form.
After that, you are requested to record your choice for the next round. Again,
you start with an amount of 100 cents in odd rounds or 100 cents plus your Total
Earnings in the previous round for even rounds, a part of which you can bet in the
lottery. The same procedure as described above determines your earnings for this round.
It is noted that your private win color remains the same, but that for each round the
computer selects a new color randomly. All subsequent rounds will also proceed in the
same manner. After the last round has been completed, your earnings in all rounds will
be totaled. This amount determines your total earnings in the experiment, which you
will receive in cash after the numbers being checked.
Appendix A: Experiment Instructions (Chapter 3)
129
Win Color: RED
REGISTRATION FORM
(Treatment F)
Name: ________________________________________________________________
Date: ___/___/______ Gender: Male / Female Age: ____
University: __________________________ Country: ______________
Course: _____________________________
Answer the following questions:
1) You have just won $30. Choose between:
(a) A 50% chance to gain $9 and a 50% chance to lose $9
(b) No further gain or loss
2) You have just lost $30. Choose between:
(a) A 50% chance to gain $9 and a 50% chance to lose $9
(b) No further gain or loss
Round Amount
in
Lottery
Round
Color
Win Lose Earnings
in
Lottery
Total
Earnings
0 (test) Win Lose
1 Win Lose
2 Win Lose
3 Win Lose
4 Win Lose
5 Win Lose
6 Win Lose
Appendix A: Experiment Instructions (Chapter 3)
130
7 Win Lose
8 Win Lose
9 Win Lose
10 Win Lose
11 Win Lose
12 Win Lose
Total
Recall:
Amount in Lottery (X): must be between 0 (zero) and 100 (one hundred) cents.
Round Color: write the color randomly selected for that round.
Win / Lose: check the corresponding box depending on if you won or lost that round.
Earnings in Lottery: equals (-X) if you lost and equals (2.5 X) if you won.
Total Earnings: equals 100 plus Earnings in Lottery.
Appendix A: Experiment Instructions (Chapter 3)
131
Win Color: RED
REGISTRATION FORM
(Treatment I)
Name: ________________________________________________________________
Date: ___/___/______ Gender: Male / Female Age: ____
University: __________________________ Country: ______________
Course: _____________________________
Answer the following questions:
1) You have just won $30. Choose between:
(a) A 50% chance to gain $9 and a 50% chance to lose $9
(b) No further gain or loss
2) You have just lost $30. Choose between:
(a) A 50% chance to gain $9 and a 50% chance to lose $9
(b) No further gain or loss
Round Amount
in
Lottery
Round
Color
Win Lose Earnings
in
Lottery
Total
Earnings
0 (test) Win Lose
1 Win Lose
2 Win Lose
3 Win Lose
4 Win Lose
5 Win Lose
6 Win Lose
Appendix A: Experiment Instructions (Chapter 3)
132
7 Win Lose
8 Win Lose
9 Win Lose
10 Win Lose
11 Win Lose
12 Win Lose
Total
Recall:
Amount in Lottery (X): must be between 0 (zero) and 100 (one hundred) cents and it
must be the same in the following 3 rounds.
Round Color: write the color randomly selected for that round.
Win / Lose: check the corresponding box depending on if you won or lost that round.
Earnings in Lottery: equals (-X) if you lost and equals (2.5 X) if you won.
Total Earnings: equals 100 plus Earnings in Lottery.
Appendix A: Experiment Instructions (Chapter 3)
133
Win Color: RED
REGISTRATION FORM
(Treatment C)
Name: ________________________________________________________________
Date: ___/___/______ Gender: Male / Female Age: ____
University: __________________________ Country: ______________
Course: _____________________________
Answer the following questions:
1) You have just won $30. Choose between:
(a) A 50% chance to gain $9 and a 50% chance to lose $9
(b) No further gain or loss
2) You have just lost $30. Choose between:
(a) A 50% chance to gain $9 and a 50% chance to lose $9
(b) No further gain or loss
Round Amount
in
Lottery
Round
Color
Win Lose Earnings
in
Lottery
Total
Earnings
0 (test) Win Lose
1 Win Lose
2 Win Lose
3 Win Lose
4 Win Lose
5 Win Lose
6 Win Lose
Appendix A: Experiment Instructions (Chapter 3)
134
7 Win Lose
8 Win Lose
9 Win Lose
10 Win Lose
11 Win Lose
12 Win Lose
Total
Recall:
Amount in Lottery (X): must decide between 0 cents and 100 cents for odd rounds (1,
3, 5, 7, 9); and between 0 cents and 100 cents plus your Total Earnings in the previous
round for even rounds (2, 4, 6, 8, 10)
Round Color: write the color randomly selected for that round.
Win / Lose: check the corresponding box depending on if you won or lost that round.
Earnings in Lottery: equals (-X) if you lost and equals (2.5 X) if you won.
Total Earnings: equals 100 plus Earnings in Lottery for odd rounds (1, 3, 5, 7, 9); and
equals Total Earnings in the previous round plus Earnings in Lottery for even rounds
(2, 4, 6, 8, 10).
Appendix B: Proofs of the Value Function Properties (Chapter 4)
135
Appendix B: Proofs of the Value Function Properties (Chapter 4)
We want to prove that the following property hold:

σ μ is the unique maximum/minimum of f (μ,⋅) and since for
σ >σ *(μ ) , ∂ (μ,σ ) > 0 σ f and for 0 σ σ *(μ ), (μ,σ ) 0,σ *(μ ) σ < < ∂ f < is a
minimum. This contradicts the existence of μ * and σ (μ*) local maxima of f (μ*,⋅)
such that f (μ*,σ (μ*)) > 0 . Hence, f (μ,σ ) < 0 and therefore ∂ (μ,σ ) < 0 σV .
Also,
lim ∂ ( , ) = 0 →∞ μ σ σ σ f for μ > 0 since
Appendix B: Proofs of the Value Function Properties (Chapter 4)

References
ACKERT, Lucy F., CHARUPAT, Narat, CHURCH, Bryan K. and Richard DEAVES,
2006, An experimental examination of the house money effect in a multi-period
setting, Experimental Economics, Vol. 9, 5-16.
ALLAIAS, Maurice, 1953, Le Comportement de l’Homme Rationnel devant le Risque:
Critique des Postulates et Axiomes de l’Ecole Americaine, Econometrica, 21,
503-546.
ARAUJO, Aloisio, Humberto MOREIRA and Marcos TSUCHIDA, 2004, The Trade-
Off Between Incentives and Endogenous Risk. Fundaçao Getulio Vargas,
Working Paper, Version: January 2004.
AVENI, Richard, 1989, Dependability and organizational bankruptcy: an application of
agency and prospect theory, Management Science. Vol. 35, No. 9, September, pp.
1120-1138.
BAKER, George, 2000, The Use of Performance Measures in Incentive Contracting,
The American Economic Review, Vol. 90, No. 2, pp. 415-420.
BANDIERA, Oriana, Iwan BARANKAY and Imran RASUL, 2004, Relative and
Absolute Incentives: Evidence on Worker Productivity, Econometric Society 2004
North American Summer Meeting, No. 277.
Bank for International Settlements, 2003, Incentive structures in institutional asset
management and their implications for financial markets, report submitted by a
Working Group established by the Committee on the Global Financial System.
BARBER, Brad and Terrance ODEAN, 2000, Trading in Hazardous to Your Wealth:
The Common Stock Investment Performance of Individual Investors, Journal of
Finance, Vol. LV, No. 2, pp. 773-806.
References
139
BARBERIS, Nicholas and Ming HUANG, 2001, Mental accounting, loss aversion, and
individual stock returns, Journal of Finance, LVI (4), pp. 1247-1295.
BARBERIS, Nicholas, HUANG, Ming and Tano SANTOS, 2001, Prospect Theory and
Asset Prices, Quarterly Journal of Economics, CXVI (1), pp. 1-53.
BARBERIS, Nicholas and Richard THALER, 2003, A Survey of Behavior Finance, in
Handbook of the Economics of Finance, George M. Constantinides, ed. Elsevie,
Chapter 18, pp 1053-1123.
BASAK, Suleyman, PAVLOVA, Anna and Alex SHAPIRO, 2003, Offsetting the
Incentives: Risk Shifting and Benefits of Benchmarking in Money Management.
MIT Sloan Working Paper No. 4303-03; NYU Finance Working Paper; EFA
2003 Annual Conference Paper No. 269; AFA 2005 Philadelphia Meetings.
Available at SSRN: http://ssrn.com/abstract=403140 or
DOI: 10.2139/ssrn.403140
BELLEMARE, Charles, KRAUSE, Michaela, KROGER, Sabine and Chendi ZHANG,
2005, Myopic Loss Aversion : Information Feedback vs Investment Flexibility,
Economics Letters, Vol. 87, 319-324.
BENARTZI, Shlomo, and Richard THALER, 1995, Myopic Loss Aversion and the
Equity Premium Puzzle, Quarterly Journal of Economics, Vol. 110, No. 1, 73-92.
BLACK, Fisher and Robert LITTERMAN, 1992, Global Portfolio Optimization,
Financial Analysts Journal, 48, 5, pp. 28-43.
BLAVATSKYY, Pavlo and Ganna POGREBNA, 2006, Myopic Loss Aversion
Revisited: the Effect of Probability Distortions in Choice Under Risk. IEW
Working Paper No. 249 Available at SSRN: http://ssrn.com/abstract=754644
References
140
BLOOM, Matt and George MILKOVICH, 1998, Relationships among risk, incentive
pay, and organizational performance. The Academy of Management Journal. Vol.
41, No. 3, June, pp. 283-297.
BOSCH-DOMÈNECH, Antoni and Joaquim SILVESTRE, 2003, Reflections on Gains
and Losses: A 2 X 2 X 7 Experiment, Working Paper, Universitat Pompeu Fabra
and University of California at Davis.
BRENNAN, Michael. J., 2001, Mental Accounting, Loss Aversion, and Individual
Stock Returns: Discussion, The Journal of Finance, Vol. 56, No. 4, 1292-1295.
CARPENTER, Jennifer, 2000, Does Option Compensation Increase Managerial Risk
Appetite ?, The Journal of Finance, Vol. 55, No. 5, October, pp. 2311-2331.
CARROLL, Carolyn, THISTLE, Paul and John WEI, 1992, The Robustness of Risk-
Return Nonlinearities to the Normality Assumption, The Journal of Financial and
Quantitative Analysis, Vol. 27, No. 3, pp. 419-435.
CHA, J.H., 1994, Aspects of individualism and collectivism in Korea. In U. Kim et al.
Individualism and collectivism: Theory, method and applications, pp. 157-174.
Thousand Oaks: Sage.
CHARNESS, Gary and Uri GNEEZY, 2003, Portfolio Choice and Risk Attitudes: An
Experiment, University of California at Santa Barbara, Economic Working Paper
Series, No. 1167.
CHEVALIER, Judith and Glenn ELLISON, 1997, Risk Taking by Mutual Funds as a
Response to Incentives, The Journal of Political Economy, Vol. 105, No. 6,
December, pp. 1167-1200.
CLARK, Jeremy, 2002, House-money Effects in Public Good Experiments,
Experimental Economics, Vol. 5, No. 3, 223-231.
References
141
COVAL, Joshua D. and Tyler SHUMWAY, 2005, Do Behavioural Biases Affect
Prices?, The Journal of Finance, Vol. 60, No. 1, February, pp. 1-34.
CUOCO, Domenico and Ron KANIEL, 2003, Equilibrium Prices in the presence of
Delegated Portfolio Management, Working Paper, Wharton School, May.
DAIDO, Kohei and Hideshi ITOH, 2005, The Pygmalion Effect: An Agency Model
with Reference Dependent Preferences, CESifo Working Paper No. 1444,
Category 10: Empirical and Theoretical Methods.
DAVIES, Greg and Stephen SATCHELL, 2004, Continuous Cumulative Prospect
Theory and Individual Asset Allocation, Cambridge Working Papers in
Economics, 467.
DEEPHOUSE, David and Robert WISEMAN, 2000, Comparing alternative
explanations for accounting risk-return relations. Journal of economic behavior &
organization. Vol. 42, No. 4, pp. 463-482.
DIMMOCK, Stephen G., 2005, Loss-Aversion and Household Portfolio Choice.
Available at SSRN: http://ssrn.com/abstract=680783
EDMANS, Alex, GARCIA, Diego and Oyvind NORLI, 2007, Sports Sentiment and
Stock Returns, The Journal of Finance, forthcoming.
EISENHARDT, Kathleen, 1985, Control: organizational and economic approach.
Management Science. Vol. 31, No. 2, February, pp. 134-149.
EISENHARDT, Kathleen., 1989, Agency Theory: An Assessment and Review. The
Academy of Management Review, Vol. 14, No. 1, Jan, pp 57-74.
ELLSBERG, D., 1961, Risk, Ambiguity, and the Savage Axioms, Quarterly Journal of
Economics, 75, 643-669.
ELTON, Edwin, GRUBER, Martin and Christopher BLAKE, 2003, Incentive Fees and
Mutual Funds, The Journal of Finance, Vol. 58, No. 2, pp. 779-804.
References
142
FELDMAN, David and Haim REISMAN, 2002, Simple Construction of the Efficient
Frontier, Forthcoming, European Financial Management.
FERNANDES, Jose L.B., 2007, Incorporating Estimation Risk Into Portfolio
Optimization for Global Long Term Investors. Available at SSRN:
http://ssrn.com/abstract=955861
FERNANDES, Jose, Augusto HASMAN and Ignacio PEÑA, 2007, Risk Premium:
Insights over the Threshold, Available at SSRN: http://ssrn.com/abstract=902812.
Forthcoming in Applied Financial Economics.
FERNANDES, José, PEÑA, Juan I., TABAK, Benjamin and José ORNELAS, 2007,
Professional Portfolio Managers, A Setting for Momentum Strategies, Available
at SSRN: http://ssrn.com/abstract=961343
FERNANDEZ, Pablo, CARABIAS, Jose Maria and de MIGUEL, Lucia, 2007a,
"Mutual Funds in Spain 1991-2006" . Available at SSRN:
http://ssrn.com/abstract=982821
FERNANDEZ, Pablo, CARABIAS, Jose Maria and de MIGUEL, Lucia, 2007b, Equity
Mutual Funds in Spain (Fondos de Inversión de Renta Variable Nacional 1991-
2006). Available at SSRN: http://ssrn.com/abstract=985120
FREEMAN, Mark A., 1997, Demographic correlates of individualism and collectivism:
A study of social values in Sri Lanka. Journal of Cross-Cultural Psychology, 28,
pp. 321-341.
GIL-BAZO, Javier and Pablo RUIZ-VERDU, 2007, Yet Another Puzzle? The Relation
between Price and Performance in the Mutual Fund Industry. Universidad Carlos
III de Madrid Business Economics Working Paper No. 06-65 Available at SSRN:
http://ssrn.com/abstract=947448
References
143
GIORGI, Enrico, HENS, Thorsten and Haim LEVY, 2004, Existence of CAPM
Equilibria with Prospect Theory Preferences, National Centre of Competence in
Research Financial Valuation and Risk Management, Working Paper No. 85, pp.
1-42.
GNEEZY, Uri, and Jan POTTERS, 1997, An Experiment on Risk Taking and
Evaluation Periods, The Quarterly Journal of Economics. Vol. 112, No. 2, May,
631-645.
GNEEZY, Uri, KAPTEYN, Arie and Jan POTTERS, 2003, Evaluation Periods and
Asset Prices in a Market Experiment, Journal of Finance, Vol. 58, pp. 821-838.
GOLEC, Joseph and Laura STARKS, 2002, Performance fee contract change and
mutual fund risk, Journal of Financial Economics, Vol. 73, No. 1, pp. 93-118.
GOMES, Francisco J., 2003, Portfolio Choice and Trading Volume with Loss Averse
Investors. EFA 0663. Available at SSRN: http://ssrn.com/abstract=249569 or
DOI: 10.2139/ssrn.249569
GOMEZ-MEJIA, Luis. and Robert WISEMAN, 1997, Reframing Executive
Compensation, Journal of Management. Vol. 23, No. 3, pp. 291-371.
GONG, James Jianxin, 2006, Financial Reporting Concerns, Framing Effects and
Employee Stock Option Grants, AAA 2007 Financial Accounting & Reporting
Section (FARS) Meeting Papers, Available at SSRN:
http://ssrn.com/abstract=930813
GOUVEIA, Valdiney and Miguel CLEMENTE, 2000, Individualism-colletivism in
Brazil and Spain: Socio-demographic correlates, Estudos de Psicologia, Vol. 5,
No. 2.
References
144
GRABKE-RUNDELL, Arden. and Luis GOMEZ-MEJIA, 2002, Power as a
determinant of executive compensation, Human Resource Management Review.
Vol. 12, No. 1, pp. 3.
GRINBLATT, Mark and Sheridan TITMAN, 1989, Adverse Risk Incentives and the
Design of Performance-Based Contracts, Management Science, Vol. 35, No. 7,
pp. 807-822.
GRUBER, Martin, 1996, Another Puzzle: The Growth in Actively Managed Mutual
Funds, The Journal of Finance. Vol. 51, No. 3, July, pp. 783-810.
HAIGH, Michael S. and John A. LIST, 2005, Do professional Traders Exhibit Myopic
Loss Aversion ? An Experimental Analysis. The Journal of Finance, Vol 60, No.
1, February, pp. 523-534.
HAN, G. and S. M. CHOE, 1994, Effects of family, region, and school network ties on
interpersonal intentions and the analysis of network activities in Korea. In U. Kim
et al, Individualism and collectivism: Theory, method and applications, pp. 213-
224. Thousand Oaks: Sage.
HARRIS, Lawrence, and Eitan GUREL, 1986, Price and volume effects associated with
changes in the S&P500: new evidence for the existence of price pressure, The
Journal of Finance. Vol. 41, No. 4, September, pp. 815-829.
HART, Oliver and Bengt HOLMSTROM, 1987, The Theory of Contracts, in Advances
in Economic Theory Fifth World Congress, Edited by Truman Bewley,
Econometric Society Monographs, pp. 71-156.
HARVEY, Campbell, LIECHTY, John and Merrill LIECHTY, 2006, Bayes vs.
Resampling: A Rematch, Available at SSRN: http://ssrn.com/abstract=898241.
HOLMSTROM, Bengt and Paul MILGROM, 1987, Aggregation and Linearity in the
Provision of Intertemporal Incentives, Econometrica. Vol. 55, No. 2, pp. 303-328.
References
145
HOLMSTROM, Bengt and Paul MILGROM, 1991, Multitask Principal-Agent
Analyses: Incentive Contracts, Asset Ownership and Job Design. Journal of Law,
Economics and Organization. Vol. 7, pp. 24-52.
HORST, Jenke, ROON, Frans and Bas WERKER, 2002, Incorporating Estimation Risk
in Portfolio Choice, CentER Working Paper No. 65. Available at SSRN:
http://ssrn.com/abstract=244695
HUANG, Lixin and Hong LIU, 2007, Rational Inattention and Portfolio Selection, The
Journal of Finance, forthcoming.
HUI, C.H. & Yee, C., 1994, The shortened individualism-collectivism scale: Its
relationship to demographic and work-related variables, Journal of Research in
Personality, 28, pp. 409-424.
JENSEN, Michael and William MECKLING, 1976, Theory of the firm: Managerial
behavior, agency costs and ownership structure. Journal of Financial Economics.
Vol. 3, October. pp. 303-360.
JIAO, Wei, 2003, Portfolio Resampling and Efficiency Issues, Master Thesis Presented
at Humboldt-University of Berlin.
JOHNSON, Eric J., GAECHTER, Simon and Andreas HERRMAN, 2006, Exploring
the Nature of Loss Aversion. IZA Discussion Paper No. 2015 Available at SSRN:
http://ssrn.com/abstract=892336.
JORION, Philippe, 1986, Bayes-Stein Estimation for Portfolio Analysis, The Journal of
Financial and Quantitative Analysis, Vol. 21, No. 3, pp. 279-292.
KAGITÇIBASI, Ç., 1994, A critical appraisal of individualism and collectivism:
Toward a new formulation. In U. Kim et al, Individualism and collectivism:
Theory, method and applications, pp. 213-224. Thousand Oaks: Sage.
References
146
KAHNEMAN, Daniel and Amos TVERSKY, 1979, Prospect Theory: An Analysis of
Decision Under Risk, Econometrica. Vol. 47, No. 2, March, pp. 263-292.
KEMPF, Alexander, KREUZBERG, Klaus and Christoph MEMMEL, 2002, How to
Incorporate Estimation Risk into Markowitz Optimization, Operations Research
Proceedings 2001, Berlin et al. 2002.
KEMPF, Alexander, RUENZI, Stefan and Tanja THIELE, 2007, Employment Risk,
Compensation Incentives and Managerial Risk Taking: Evidence from the Mutual
Fund Industry. Available at SSRN: http://ssrn.com/abstract=964923
KOHLI, Jasraj, 2005, An Empirical Analysis of Resampled Efficiency, Thesis
submitted to the Faculty of the Worcester Polytechnic Institute.
KOUWENBERG, Roy and William ZIEMBA, 2004, Incentives and Risk Taking in
Hedge Funds, Working Paper.
LANGER, Thomas and Martin WEBER, 2003, Does Binding of Feedback Influence
Myopic Loss Aversion ? An Experimental Analysis, CEPR Discussion Paper, No.
4084, October.
LANGER, Thomas, and Martin WEBER, 2005, Myopic Prospect Theory vs Myopic
Loss Aversion: How General is the Phenomenon ? Journal of Economic Behavior
& Organization, Vol. 56, 25-38.
LAZEAR, Edward, 2000a, Performance Pay and Productivity, The American Economic
Review, Vol. 90, No. 5, December, pp. 410-414.
LAZEAR, Edward, 2000b, The Power of Incentives, The American Economic Review,
Vol. 90, No. 2, May, pp. 1346-1361.
LEVINSON, Justin and Kaiping PENG, 2007, Valuing Cultural Differences in
Behavioral Economics, The ICFAI Journal of Behavioral Finance, Vol IV, No. 1,
pp. 32-47.
References
147
LEVY, H., and M. LEVY, 2004, Prospect Theory and Mean-Variance Analysis, Review
of Financial Studies, Vol. 17, No. 4, pp. 1015-1041.
LO, Andrew, REPIN, Dmitry and Brett STEENBARGER, 2005, Fear and Greed in
Financial Markets: A Clinical Study of Day-Traders, Working Paper.
LUCEY, Brian and Michael DOWLING, 2005, The Role of Feelings in Investor
Decision-Making, Journal of Economic Surveys, Vol. 19, No. 2, 211-237.
MAENHOUT, Pascal, 2004, Robust Portfolio Rules and Asset Pricing, Review of
Financial Studies, 17, 4, pp. 325-362.
MAGI, Alessandro, 2005, Financial Decision-making and Portfolio Choice under
Behavioral Preferences: Implications for the Equity Home Bias Puzzle. Available
at SSRN: http://ssrn.com/abstract=725521
MARKOWITZ, Harry and Nilufer USMEN, 2003, Resampled Frontiers versus Diffuse
Bayes: An Experiment, Journal of Investment Management, vol. 1, no. 4 (Fourth
Quarter), pp. 9-25.
MARKOWITZ, Harry. 1952. Portfolio Selection. The Journal of Finance, Vol. 7, No.
1, March, pp. 77-91.
MASSA, Massimo and Rajdeep PATGIRI, 2007, Incentives and Mutual Fund
Performance: Higher Performance Or Just Higher Risk Taking?, Available at
SSRN: http://ssrn.com/abstract=891442
MEHRA, Rajnish, and Edward PRESCOTT, 1985, The Equity Premium: A Puzzle,
Journal of Monetary Economics, Vol. 15, 145-61.
MEHRA, Rajnish, 2003, The Equity Premium: What is the Puzzle ? Financial Analyst
Journal, Jan-Feb, 59-64.
MICHAUD, Richard, 1998, Efficient Asset Management, Boston, MA: Harvard
Business School Press.
References
148
MILLER, K. and P. BROMILEY, 1990, Strategic risk and corporate performance: An
analysis of alternative risk measures. Academy of Management Journal, Vol. 33,
No. 4, December, pp. 756-779.
MOHNEN, Alwine and Kathrin POKORNY, 2004, Reference Dependency and Loss
Aversion in Employer-Employee-Relationships: A Real Effort Experiment on the
Incentive Impact of the Fixed Wage, Available at SSRN:
http://ssrn.com/abstract=641542.
ODEAN, Terrance, 1998, Are Investors Reluctant to Realize their Losses, The Journal
of Finance, Vol. 53, No. 5, October, 1775-1798.
PALOMINO, Frederic and Andrea PRAT, 2002, Risk Taking and Optimal Contracts
for Money Managers, CEPR, Working Paper, July.
PAWLEY, Mark, 2005, Resampled Mean-Variance Optimization and the Dynamic
Nature of Markets, Paper presented at the biennial 2005 conference of the
Economic Society of South Africa.
POLKOVNICHENKO, Valery, 2002, Household Portfolio Diversification, Working
Paper, University of Minnesota.
PRENDERGAST, Canice, 2002, The tenous trade-off between risk and incentives,
Journal of Political Economy, 110 (5), pp. 1071-1102.
RABIN, Mattew and Dimitri VAYANOS, 2007, The Gambler’s and Hot-Hand
Fallacies: Theory and Applications, Working Paper.
RABIN, Matthew, 1998. Psychology and economics, Journal of Economic Literature.
Vol. 36, No. 1, March, 11-46.
ROSS, Stephen, 2004, Compensation, Incentives, and the Duality of Riskness and Risk
Aversion, Journal of Finance, 59, 1, pp. 207-225.
References
149
SCHMIDT, Ulrich, and Horst ZANK, 2005, What is Loss Aversion ? The Journal of
Risk and Uncertainty, Vol. 30, No. 2, 157-167.
SENSOY, Berk A., 2006, Agency, Incentives, and Mutual Fund Benchmarks. Available
at SSRN: http://ssrn.com/abstract=890692
SHEFRIN, Hersh and M. STATMAN, 1985. The Disposition Effect to sell winners too
early and ride losers too long. Journal o Finance, Vol. 40, No. 3, pp. 777-790.
SHEFRIN. Hersh, 2005, A Behavioral Approach to Asset Pricing, Elsevier Academic
Press.
SHERER, Bernd, 2002, Portfolio Resampling: Review and Critique, Financial Analysts
Journal, Vol. 58, No. 6, pp. 98-109.
SHLEIFER, Andrei and Robert VISHNY, 1997, The Limits of Arbitrage, The Journal
of Finance. Vol. 52, No. 1, March, pp. 35-55.
SLATTERY, Jeffrey and Daniel GANSTER, 2002, Determinants of Risk Taking in a
Dynamic Uncertain Context, Journal of Management, Vol. 28, No. 1, 89-106.
St-AMOUR, Pascal, 2006, Benchmarks in Aggregate Household Portfolios. Swiss
Finance Institute Research Paper No. 07-09 Available at SSRN:
http://ssrn.com/abstract=985256.
STARKS, L. T., 1987, Performance Incentive Fees: An Agency Theoretic Approach,
Journal of Financial and Quantitative Analysis, 22, pp- 17-32.
STRACCA, Livio, 2005, Delegated Portfolio Management - A Survey of the
Theoretical Literature, European Central Bank, Working Paper Series, No. 520,
September.
SUNDARAM, Rangarajan and David YERMACK, 2006, Pay me Later: Inside Debt
and Its Role in Managerial Compensation, Journal of Finance, Forthcoming.
References
150
THALER, Richard and Eric JOHNSON, 1990, Gambling with the House Money and
Trying to Break Even: The Effects of Prior Outcomes on Risky Choice.
Management Science, pp. 643-660.
THALER, Richard H, TVERSKY, Amos, KAHNEMAN, Daniel and Alan
SCHWARTZ, 1997. The Effect of Myopia and Loss Aversion on Risk Taking:
An Experimental Test, The Quarterly Journal of Economics, MIT Press, vol.
112(2), pages 647-61, May.
THALER, Richard H., 1999, Mental accounting matters, Journal of Behavioral
Decision Making, 12, pp. 183-206.
THALER, Richard H., and Eric J. JOHNSON, 1990, Gambling with the House-money
and Trying to Break Even: The Effects of Prior Outcomes on Risky Choice,
Management Science, Vol. 36, 643-660.
THALER, Richard, 2000, From Homo Economicus to Homo Sapiens, The Journal of
Economic Perspectives, Vol. 14, No. 1, pp. 133-141.
TRIANDIS, Harry, 1995, Individualism and Collectivism, Westview Press, Oxford.
TVERSKY, Amos, and KAHNEMAN, Daniel, 1992. Advances in prospect theory:
cumulative representation of uncertainty, Journal of Risk and Uncertainty. Vol. 5,
No. 4, 297-323.
VLCEK, Martin, 2006, Portfolio Choice with Loss Aversion, Asymmetric Risk-Taking
Behavior and Segregation of Riskless Opportunities, Swiss Finance Institute,
Research Paper Series, N. 6 – 27.
Von NEUMANN, J. and O. MORGENSTERN, 1944, The Theory of Games and
Economic Behavior, Princeton University Press.
References
151
WEBER, Martin and Colin CAMERER, 1998, The Disposition Effect in Securities
Trading: An Experimental Analysis, Journal of Economic Behavior and
Organization, Vol. 33, No. 2, 167-184.
WEBER, Martin, and Heiko ZUCHEL, 2003, How do Prior Outcomes Affect Risk
Attitude ? Universität Mannheim, Working Paper, version: February.
WHITE, H. 1980, A Heteroskedasticity-Consistent Covariance Matrix Estimator and a
Direct Test for Heteroskedasticity. Econometrica, Vol. 48, No. 4, pp. 817-838.
WISEMAN, Robert, and Luis GOMEZ-MEJIA, 1998, A Behavioral Agency Model of
Managerial Risk Taking, Academy of Management Review. Vol. 23, No. 1,
January, pp. 133-152.
WOLF, Michael, 2006, Resampling vs. Shrinkage for Benchmarked Managers, Institute
for Empirical Research in Economies, University of Zurich, Working Paper
Series, No. 263.
WRIGHT, Peter, KROLL, Mark and ELENKOV, Detelin. 2002. Acquisition returns,
increase in firm size, and chief executive officer compensation: the moderating
role of monitoring. The Academy of Management Journal. Vol. 45, No. 3, June,
pp. 599-608.
ZIMMER, Christian and Beat NIEDERHAUSER, 2003, Determining an Efficient
Frontier in a Stochastic Moments Setting, Universidade de São Paulo,
Departamento de Administração, Working Paper Series Nº 03/010.

In Proper format thesis is available here!
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CDMQFjAA&url=http%3A%2F%2Fe-archivo.uc3m.es%2Fbitstream%2F10016%2F2509%2F1%2FTesis_barros_fernandes.pdf&ei=yaHiULqrL-up0AW48YHwDQ&usg=AFQjCNHw1up-p5V4i4mOcEGfaUm5irOMzQ&sig2=AwPjmJDHHTLEB9-BWIr2fg&bvm=bv.1355534169,d.d2k


Accessing rural finance
The rural financial market in Northern Vietnam
Studies on the Agricultural and Food Sector
in Central and Eastern Europe
Edited by
Leibniz Institute of Agricultural Development
in Central and Eastern Europe
IAMO
Volume 36
Accessing rural finance
The rural financial market in Northern Vietnam
by
Thomas Dufhues
IAMO
2007
Bibliografische Information Der Deutschen Bibliothek
Die Deutsche Bibliothek verzeichnet diese Publikation in der Deutschen
Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über
http://dnb.ddb.de abrufbar.
Bibliographic information published by Die Deutsche Bibliothek
Die Deutsche Bibliothek lists the publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available in the internet at: http://dnb.ddb.de.
This thesis was accepted as a doctoral dissertation in fulfillment of the requirements
for the degree "Doktor der Agrarwissenschaften" by the Faculty Agricultural
Sciences at University of Hohenheim on 29.06.2005.
Date of oral examination: 06.03.2006
Examination Committee
Dean and Head of the Committee: Prof. Dr. K. Stahr
Supervisor and Reviewer: Prof. Dr. F. Heidhues
Co-Reviewer: Prof. Dr. M. Zeller
Additional examiner: Prof. Dr. J. Zeddies
Additional examiner: Prof. Dr. G. Buchenrieder
Diese Veröffentlichung kann kostenfrei im Internet unter
<www.iamo.de/dok/sr_vol36.pdf> heruntergeladen werden.
This publication can be downloaded free from the website
<www.iamo.de/dok/sr_vol36.pdf>.
􀂤 2007
Leibniz-Institut für Agrarentwicklung in Mittel- und Osteuropa (IAMO)
Theodor-Lieser-Straße 2
06120 Halle (Saale)
Tel.: 49 (345) 2928-0
Fax: 49 (345) 2928-199
e-mail: iamo@iamo.de
http://www.iamo.de
ISSN 1436-221X
ISBN 3-938584-16-5
D 100 Diss. Universität Hohenheim
to Meike
"So close no matter how far…"
(Nothing Else Matters [Hetfield/Ulrich],
The Black Album, 1991, Metallica)

Preface
Transition economies have gone through a fundamental restructuring of their
economy. This process was not without difficulties arising from the interruption
of long lasting economic relations and the need to change institutional structures
and responsibilities. Vietnam is no exception. During the transition, the adaptation
of the financial system was one of the most challenging reforms in Vietnam.
One of the major tasks of this reform was to expand the financial system’s outreach
to the newly emerging private sector and household economies, particularly
in rural areas.
This research was carried out within the research program "Sustainable Land Use
and Rural Development in Mountainous Regions of Southeast Asia" (SFB 564,
Sub-project F2) of the University of Hohenheim. The work of Thomas Dufhues
analyzes important features of the rural financial system in Northern Vietnam.
The main research focus is on factors that impede or support access of rural
households to the formal financial system on the one hand and the outreach of
this system to rural households on the other. Access to financial products of the
formal sector has a great potential for reducing poverty which is still widely
spread in the northern uplands of Vietnam. Thus, conceptually, this research
focuses on access and access constraints of rural households to the formal financial
system.
The thesis comprises cumulatively four individual articles. Each article focuses
on different aspects of the formal financial system. The first paper reviews the
transformation process of the financial system in Vietnam, with special attention
to the rural financial market. Methodologically, an in-depth literature review is
carried out, supplemented by anecdotal evidence gathered during the field research.
In the second article, the information polices of the Vietnam Bank for the
Poor (VBP) as the main supplier of credit to the rural poor in Northern Vietnam
during the years 2001 and 2002 are investigated by means of information economics
analysis. Methodologically, secondary data from the VBP and the local
administration are combined with qualitative evidence collected at the intermediary
and household level. The third paper analyzes the poverty outreach of the
formal financial intermediaries by a Principal Component Analysis; household
access constraints to formal credit are assessed by a binary logit analysis.
IV Preface
Both analyses are based on quantitative rural household data. In the fourth article,
client-adapted financial services are developed using a Conjoint Analysis approach,
which is a marketing research tool that combines quantitative and qualitative
data in its analysis.
Major results are that the rural households are well supplied with formal credits
through the state-owned formal financial intermediaries. However, this outreach
has been bought by immense amounts of subsidies and at the expense of institutional
sustainability. Furthermore, formal savings, which are highly demanded
by the rural population, are not offered by the financial intermediaries. Whether
the enormous government subsidies channeled into rural credit contribute to
poverty reduction, remains a subject for future research. It is clear, however, that
these subsides bypass households that have no demand for credit, and these
households are often the poorest. Finally, the highly subsidized credit interest
rates imply low savings rates, which again affect the poor most, because they
cannot assemble enough savings to buy lumpy, non-financial assets, such as
land and buffaloes. The backlash of cheap credit is that the poor take a beating
on their financial savings. The transformation polices of the Vietnamese government
concerning the rural financial system are biased towards the supply of
preferential credits and this policy discriminates against the poorest households;
it hampers particularly their access to formal savings products. Thus, a paradigm
change for the rural finance sector within is called for with a change from a
‘credit only’ to a ‘full-intermediary’ approach in the rural financial system.
Prof. Dr. Dr. h.c. Franz Heidhues Prof. Dr. Gertrud Buchenrieder
Director of SFB 564 2000-06 Leader of sub-project F2, SFB 564, 2000-06
University of Hohenheim, Stuttgart IAMO, Halle
Acknowledgements
"You will never walk alone…" I am not exactly an Elvis Presley fan, but this
small song line is particularly true of my Ph.D. thesis. I have received the help
of numerous people to whom I wish to express my deepest gratitude; without
their support I would not have been able to complete my thesis. Although I cannot
list the names of all those who supported, commented on and contributed to this
work, I would like to thank them all.
My sincere gratitude goes first of all to Prof. Dr. Dr. h.c. Franz Heidhues, my
adviser, for his support and valuable criticism. I am also grateful to Prof. Dr.
Manfred Zeller for reviewing part of my work and serving on the thesis committee.
I would also like to thank Prof. Dr. Dr. h.c. mult. Jürgen Zeddies and
PD Dr. Gertrud Buchenrieder for serving on the thesis committee. I am particularly
indebted to Gertrud: She was also the project leader of the research project
within which this work has grown. Her encouragement at all stages of the study
and her understanding of imponderables and her backing during the field work
was essential for the success of my work. Her active support, enthusiasm, and
constructive criticism were greatly appreciated. I am especially grateful to her
for passing on to me her extensive knowledge about writing articles.
I want to thank the following persons for their comments on the articles or earlier
drafts of the articles presented in this thesis: PD Dr. Gertrud Buchenrieder,
Prof. Dr. Dr. h.c. Franz Heidhues, Dr. Daniel Müller, Prof. Dr. phil. Hans Dieter
Seibel, Ph.D. Manohar Sharma, Dipl. agr. Insa Theesfeld and Prof. Dr. Manfred
Zeller.
I am particularly thankful to my wife and former colleague Meike Geppert for a
beautiful year together in Vietnam, for her valuable criticism, and for her understanding
and encouragement during our eight months of unanticipated separation.
I wish to thank all my colleagues in the Department of Development Theory and
Policy in the Tropics and Subtropics at the University of Hohenheim for their
friendship, fruitful discussions and the pleasant working atmosphere. Special
thanks go to Dr. Annette Luibrand, Judith Möllers, Rainer Schwarzmaier and
Dr. Sabine Daude. I also want to express my gratitude to my colleagues from the
VI Acknowledgements
Uplands Program, Isabel Fischer and Ute Lemke. Particularly Ute and Judith
were always available to listen to the problems and the sunny sides of my research
work. Furthermore, I want to thank Dr. Harald Leisch for his invaluable work
as a coordinator in Hanoi, and Dr. Jens Pape, Nicole Flick, Anne Schmepp,
Gudrun Contag and Cornelia Schumacher for their work on administrative and
accounting matters.
During my stay in Vietnam, I was supported by the Hanoi Agricultural University
and by Thai Nguyen University of Agriculture and Forestry. Very special thanks
go to my Vietnamese advisor Prof. Dr. Pham Thi My Dung for her invaluable
support. Furthermore, I want to thank Dr. Nguyen The Dang for his organizational
help in the research region of Ba Be.
My special appreciation goes to my Vietnamese research assistants and interpreters
Mr. Hoang Dinh Quoc and Ms Nguyen Thi Thu Huyen. Quoc in particular,
who became a close friend over the years, saved me many times from embarrassing
situations during my personal clash of cultures and I am deeply indebted
to him. While he helped me to eat ‘inedible’ things, I could at least help
him to drink ‘undrinkable’ things and thus ease to some extent the debt I owe
him. In addition, I want to say thank you to all the staff involved in my research
project. I especially want to thank Le Minh Tuan and Vu Long for their outstanding
work, but I also want to express my thanks to all the farmers who were
patient enough to participate in the surveys.
Special thanks goes to my former flat-mates in Hanoi, Dr. Daniel Müller,
Steffen Rehschwamm and Jeremy Cross, for fighting my bad moods and for being
good company. Here I want to express special thanks to Daniel for his invaluable
support with the logit regressions and to Steffen Rehschwamm for his help with
setting up the database for my work. Further, I also want to thank Olaf Balding,
Tobias Barisch, and Will Conroy for being, among other things, good company
in the ‘Bia Hoi’s of Hanoi.
The research for this thesis was carried out within the framework of the German-
Thai-Vietnamese Collaborative Research Program ‘Sustainable Land Use and
Rural Development in Mountainous Regions of Southeast Asia’, also known as
Uplands Program, within subproject F2 ‘Rural Financial Market Development’.
Funding from the Deutsche Forschungsgemeinschaft (DFG) and co-funding
from the Ministry of Science, Technology, and Environment of Vietnam is
gratefully acknowledged.
Table of contents
Preface ............................................................................................................ III
Acknowledgements............................................................................................. V
Table of contents ............................................................................................ VII
List of tables ............................................................................................................. X
List of figures ................................................................................................. XI
List of abbreviations...................................................................................... XII
Abstract ................................................................................................................ XV
Deutsche Kurzfassung................................................................................... XIX
1 Introduction ............................................................................................... 1
1.1 Conceptual framework and objectives................................................................. 2
1.2 Organization of the study ...................................................................................... 5
2 The transformation of the financial system in Vietnam and its
implications for the rural financial market – An update ...................... 7
2.1 Introduction ............................................................................................................ 7
2.2 The financial landscape in Vietnam ..................................................................... 9
2.3 Reforms in the financial market ........................................................................... 12
2.3.1 Reforming steps in the banking sector.............................................................. 12
2.3.1.1 Competition.................................................................................................. 13
2.3.1.2 Interest rate policy........................................................................................ 15
2.3.1.3 Political lending ........................................................................................... 17
2.3.2 State-owned enterprises and non-performing loans......................................... 18
2.3.3 Legal environment ............................................................................................ 21
2.3.4 Regulatory environment.................................................................................... 23
2.4 Conclusions and policy recommendations ........................................................... 25
VIII Table of contents
3 Information and targeting policies and their principal-agent
relationships – The case of the Vietnam Bank for the Poor .................. 29
3.1 Introduction ............................................................................................................ 29
3.1.1 Problem statement ............................................................................................ 31
3.1.2 Objectives and methodology............................................................................. 32
3.2 New Institutional Economics and the analysis of information flows in
Vietnam’s rural credit allocation.......................................................................... 32
3.2.1 Transaction costs .............................................................................................. 33
3.2.2 Principal-agent concept.................................................................................... 34
3.3 Targeting policies of the VBP in Ba Be ................................................................ 36
3.3.1 Target group definition and poverty criteria.................................................... 36
3.3.2 Effective targeting of the VBP loans in the research communes...................... 39
3.4 Decision-making process of credit allocation for the poor in the
case of VBP ............................................................................................................. 42
3.4.1 Credit policy ..................................................................................................... 42
3.4.2 Credit allocation process.................................................................................. 44
3.4.3 Loan application procedure.............................................................................. 45
3.4.4 Backward and forward information flows between VBP,
local authorities and households ...................................................................... 47
3.4.4.1 Formalized information channels................................................................. 47
3.4.4.2 Non-formalized information channels ......................................................... 51
3.5 Conclusions and policy recommendations ........................................................... 51
4 Outreach of credit institutes and households' access
constraints to formal credit in Northern Vietnam ................................. 55
4.1 Introduction ............................................................................................................ 55
4.2 Methodology and data ........................................................................................... 57
4.2.1 Access constraints to formal rural credit – The conceptual framework........... 57
4.2.2 Measuring outreach and access: Econometric models .................................... 61
4.2.3 Regional focus and sampling procedures......................................................... 67
4.3 Outreach of and access to formal rural lenders in Vietnam.............................. 69
4.3.1 Collateral use.................................................................................................... 69
4.3.2 Effective formal credit demand......................................................................... 71
4.3.3 Credit outreach ................................................................................................. 75
4.3.4 Credit-constrained households ......................................................................... 78
4.4 Conclusions and policy recommendations ........................................................... 84
Table of contents IX
5 Participatory product design by using Conjoint Analysis
in the rural financial market of Northern Vietnam............................... 87
5.1 Introduction ............................................................................................................ 87
5.2 Material and methods ............................................................................................ 89
5.2.1 Conjoint survey ................................................................................................. 90
5.2.2 Regional focus and sampling procedures......................................................... 94
5.2.3 Econometric analysis........................................................................................ 95
5.3 Results and discussion............................................................................................ 97
5.3.1 Institutional assessment .................................................................................... 97
5.3.1.1 Structure of VBARD/VBP........................................................................... 98
5.3.1.2 Supply of savings products .......................................................................... 110
5.3.2 Potential demand .............................................................................................. 111
5.3.2.1 Potential demand for credit .......................................................................... 111
5.3.2.2 Potential demand for savings ....................................................................... 114
5.4 Conclusions and policy recommendations ........................................................... 117
6 Final conclusion ......................................................................................... 119
Reference list.................................................................................................. 124
Annex.............................................................................................................. 141
List of tables
Table 2-1: Increasing number of players in the banking sector .................................. 13
Table 3-1: Income categories of households in VND ................................................. 39
Table 3-2: Wealth status and VBP loans of HH in research communes..................... 40
Table 3-3: Percentage of HH per village with VBP credits? ...................................... 41
Table 3-4: Nominal interest rates per month of the VBP in Ba Be district................. 43
Table 4-1: Variables for the binary logistic regression model on credit access.......... 64
Table 4-2: Research areas and sample composition.................................................... 68
Table 4-3: Depth of outreach of formal lender clients ................................................ 76
Table 4-4: Outreach of formal credit by poverty group .............................................. 76
Table 4-5: Parameters influencing households’ access to formal credit –
Binary logit estimation .............................................................................. 79
Table 4-6: Classification of correctly predicted access to formal
credit-constrained households ................................................................... 80
Table 5-1: Credit attributes and their levels ................................................................ 91
Table 5-2: Deposit attributes and their levels.............................................................. 92
Table 5-3: Research areas and sample composition.................................................... 95
Table 5-4: Loan structure of VBARD/VBP in Ba Be district..................................... 104
Table 5-5: Nominal interest rates per month of VBARD/VBP in Ba Be district ....... 105
Table 5-6: Savings products at VBARD in Ba Be district.......................................... 111
Table 5-7: Logit estimation of average utility values for saving attributes................... 116
Table A-1: Descriptive statistics of the independent variables for the binary
logistic regression model on credit access................................................. 141
Table A-2: Logit estimation of average utility values for credit attributes.................. 143
Table A-3: Logit estimation of average utility values for saving attributes ................ 144
List of figures
Figure 1-1: Factors of access and outreach relating to the financial system ................ 2
Figure 1-2: The rural financial market and the levels of analysis ................................ 3
Figure 3-1: The nine-step procedure to obtain a loan from the VBP ........................... 46
Figure 3-2: Formalized information flows ................................................................... 49
Figure 4-1: Conceptual framework – The capital-collateral system ............................ 59
Figure 4-2: Percentage of households using different credit sectors............................ 72
Figure 4-3: Access-constrained households ................................................................. 72
Figure 4-4: Interest rates per month of formal and informal credits ............................ 73
Figure 4-5: Formal and informal credit terms .............................................................. 74
Figure 4-6: Loan size of formal and informal credits in VND millions....................... 74
Figure 4-7: Loan term of formal credits in years.......................................................... 78
Figure 4-8: Loan amount of formal credits .................................................................. 78
Figure 5-1: Real credit interest rate of VBP in Ba Be district...................................... 105
Figure 5-2: The six-step procedure to obtain an individual loan from the VBARD
in Ba Be district ......................................................................................... 109
Figure A-1: Decision tree of the effective credit demand in the formal sector ............. 142
Figure A-2: Principal Component Indicators ................................................................ 142
List of abbreviations
ADB Asian Development Bank
AusAID Australian Agency for International Development
BRI Bank Rakyat Indonesia
CA Conjoint Analysis
CBC Choice-Based Conjoint Analysis
CECARDE Center for Consultation on Agriculture and Rural Development
CGL Credit Group Leader
CIA Central Intelligence Agency
CIEM Central Institute of Economic Management
CIRAD Centre de Coopération Internationale en Recherche Agronomique pour
le Développement
Danida Danish Ministry of Foreign Affairs
DFG Deutsche Forschungsgemeinschaft
DFID Department for International Development
DSA Development Studies Association
DSI Development Strategy Institute
EIU The Economist Intelligence Unit Limited
FAME International Center for Financial Asset Management and Engineering
FCND Food Consumption and Nutrition Division
FFI Formal Financial Intermediary
FIA Forschungsstelle für Internationale Wirtschafts- und Agrarentwicklung
GDP Gross Domestic Product
GSO General Statistical Office
GTZ Deutsche Gesellschaft für Technische Zusammenarbeit
HAU Hanoi Agriculture University No. 1
HH Household
IAAE International Conference of Agricultural Economists
List of abbreviations XIII
IAES International Atlantic Economic Conference
IFAD International Fund for Agricultural Development
IFPRI International Food and Policy Research Institute
IRIS Institutional Reform and the Informal Sector
JSB Joint Stock Bank
MARD Ministry of Agriculture and Rural Development
MFI Microfinance Institute
MIP Ministry of Planning and Investment
MIT Massachusetts Institute of Technology
MOLISA Ministry of Labor, Invalids and Social Affairs
NBER National Bureau of Economic Research
NIAS Nordic Institute of Asian Studies
NIE New Institutional Economics
OECD Organization for Economic Co-operation and Development
PCA Principal Component Analysis
PCF People's Credit Fund
PR Property Right
PRA Participatory Rural Appraisal
PRB Poverty Reduction Board
SBVN State Bank of Vietnam
SGT Saigon Times
Sida Sweden International Development Agency
SME Small and Medium Enterprise
SOCB State-Owned Commercial Bank
SOE State-Owned Enterprise
TC Transaction Cost
UN United Nations
UNDP United Nations Development Program
UNIDO United Nations Industrial Development Organization
USAID United States Agency for International Development
VBARD Vietnam Bank for Agriculture and Rural Development
VBP Vietnam Bank for the Poor
VBSP Vietnam Bank for Social Policies
XIV List of abbreviations
VIR Vietnam Investment Review
VND Vietnamese Dong
VOV Voice of Vietnam
VPSC Vietnam Postal Savings Company
Abstract
During the transition of the Vietnamese economy, adaptation of the financial
system was one of the most challenging reforms. A major task of this reform
was to expand the financial system’s outreach to the newly emerging private
sector and household economies, especially in rural areas. Therefore, state-owned
financial intermediaries such as the Vietnam Bank for Agriculture and Rural
Development (VBARD) and the Vietnam Bank for the Poor (VBP) have been
established. Despite general successes in terms of credit outreach, certain population
groups, particularly the poorest, which are often identical to ethnic minorities,
seem to have been bypassed by both banks. Furthermore, the strategy pursued by
national rural financial policy has focused mainly on credit supply. Other financial
services that potentially have a deeper outreach, e.g. savings products, have
been neglected in the development efforts of the government and its development
banks. The overall objective of this research study was to create knowledge
on the factors that impede or support access of rural households in Northern
Vietnam to the formal financial system. Access can be hampered at different
levels of the financial system, namely macro/meso level, intermediary level and
household level. Joint analysis of the three levels is therefore appropriate. This
implies different methodologies and data collection methods. The data collection
took place between March 2001 and 2002 in two provinces of Northern Vietnam.
In total, 260 households were surveyed. In addition, several participatory methods
were applied at all levels to collect qualitative data. Furthermore, secondary data
was collected from relevant financial intermediaries and administrations. This
cumulative thesis is divided into four main sections investigating different levels
of the system and applying different methodologies. The second section reviews
the transformation process of the financial system in Vietnam with special attention
to the rural financial market. Methodologically, an in-depth literature review
is carried out, supplemented by anecdotal evidence gathered during the field research.
In the third section, the information polices of the VBP as the main
supplier of credit to the rural poor in Northern Vietnam are investigated by
means of an information economics analysis. Methodologically, secondary data
from the VBP and the local administration are combined with qualitative evidence
from the intermediary and household level. The fourth section analyzes the
Abstract
XVI
poverty outreach of the formal financial intermediaries using a Principal Component
Analysis, and it analyzes household access constraints to formal credit
using a binary logit analysis. Both analyses are based on quantitative household
data. In section fifth, client-adapted financial services are developed using a
Conjoint Analysis approach, which is a marketing research tool and combines
quantitative and qualitative data in its analysis. The rural financial market in
Vietnam is still dominated by the aforementioned highly subsidized state-owned
financial intermediaries, impeding the establishment of any viable financial
services and hampering innovations. Through the creation of the Vietnam Bank
for Social Policies (VBSP) (the successor of the dissolved VBP since 2003) the
Vietnamese government has separated political lending from commercial lending.
Evidence from development banks in other countries suggests that the VBARD,
now freed from political lending, is likely to dismiss its peasant clientele and
concentrate on wealthier farmers. The question is how long the Government can
finance the VBSP, and who will serve the rural poor after the government stops
the subsidies and the VBPS cannot carry on its operations? The sustainability of
the financial system is still threatened by an accumulation of non-performing
loans amassed by state-owned enterprises. In addition, the problem of nonperforming
loans is spreading to the private sector – including rural households.
Apart from representing macro-economic threats to the financial system, this
moral hazard behavior is hindering the establishment of any viable rural financial
intermediation. The breadth of outreach of the formal rural lenders is immense.
However, the poorest households are seldom clients. But general poverty (as
captured in the poverty index) does not significantly influence access to formal
credit. This means that the poorest households simply have much less demand
for formal credit products. Improving credit products or offering new credit lines
would only slightly improve the credit coverage of poorer households. A more
promising approach would be to introduce a specialized pro-poor extension service
to widen the scope of their investment ideas, combined with a general improvement
in the infrastructure. A good market connection serves credit outreach
in a twofold manner: First, households have better access to creditrelevant
information; and second, through better market access they may find
new investment opportunities. Nevertheless, the number of access-constrained
households is surprisingly low, at 16%. One explanation may be the eradication of
former access constraints through locally disbursed group credits. However, considering
the anecdotal reports of very low repayment rates, the price of eradicating
these access constraints has likely been a decrease in financial sustainability of
the formal lenders. Nevertheless, some barriers to access continue to exist, particularly
for ethnic minorities or female-headed households. To reduce these
Abstract
XVII
access barriers, locally-oriented actions should be taken, catering to the specific
needs of those households which lack access. The establishment of the VBSP
represents an attempt to broaden access in general. But it is questionable
whether households that do not have access today, or do not demand the existing
products, will demand loans from the VBSP. A more sustainable way to promote
outreach would be to improve the knowledge of fringe groups, such as ethnic
minorities or female-led households, about credit application procedures. The
supply of essential credit information to these groups is impeded by the supplyoriented
flow of information and by relying for the dissemination process on local
authorities, which favor the ethnic majority. Ethnic or gender diversification of
bank staff could broaden the information networks available and could create
more awareness of those groups inside the institution. In contrast to the enormous
credit outreach, formal savings are rarely used by rural households. However,
this low effective demand for savings is due to inappropriate services and
not to lack of willingness of the rural population to save. Thus, the most appropriate
tool to incorporate poorer households into the formal financial system
would be mobilization of savings by providing adapted services.

Deutsche Kurzfassung
Während der Transformation der vietnamesischen Wirtschaft war die Anpassung
des Finanzsektors eine der anspruchvollsten Reformen. Eines der wichtigsten
Ziele dieser Reform war es, den neu entstandenen privaten Betrieben und den
privaten Haushalten Zugang zu formalen Finanzdienstleistungen zu verschaffen,
wobei hierbei der ländliche Raum spezielle Förderung erfuhr. Für die Erreichung
dieses Zieles wurden staatliche Entwicklungsbanken wie die Vietnamesische
Bank für Landwirtschaft und Ländliche Entwicklung (VBARD) und die Vietnamesische
Bank für die Armen (VBP) gegründet. Diese beiden Institutionen haben
den Zugang der ländlichen Bevölkerung zu formalen Krediten erheblich verbessert.
Allerdings haben bestimmte Randgruppen wie ethnische Minderheiten,
welche häufig identisch sind mit den ärmeren Bevölkerungsschichten, oft keinen
Zugang zu diesen staatlichen Krediten. Ein weiteres Problem ist die Fokussierung
der Regierung und der staatlichen Entwicklungsbanken auf Kredite als primäre
Dienstleistung. Die Entwicklung anderer Finanzdienstleistungen wie unterschiedliche
Spar- und Versicherungsprodukte, welche einen potentiell besseren Zugang
zu marginalisierten Gruppen haben wurde, im ländlichen Finanzsystem völlig
vernachlässigt.
Das Ziel dieser Arbeit ist es, herauszuarbeiten, welche Faktoren den Zugang
ländlicher Haushalte zu formalen Finanzdienstleitungen behindern oder fördern.
Der Zugang zu Finanzdienstleistungen kann an verschieden Stellen des Finanzsystems
behindert werden, zum Beispiel auf der Haushaltsebene, der Ebene der
Finanzdienstleister oder durch gesetzliche Rahmenbedingungen. Deswegen ist
eine gemeinsame Betrachtung dieser verschieden Ebenen notwendig. Dies impliziert
wiederum unterschiedliche Analyse- und Datenerhebungsmethoden. Die
Datenerhebung, die der Datenbasis zu Grunde, liegt fand zwischen März 2001
und März 2002 in zwei verschiedenen Provinzen in Nordvietnam statt. Insgesamt
wurden 260 Haushalte mit einem standardisierten Fragebogen interviewt.
Zusätzlich wurden zahlreiche partizipatorische Erhebungstechniken angewandt,
um weitergehende Einsichten in die Haushalte, die Finanzdienstleister und andere
relevante Institutionen zu bekommen. Des Weiteren wurden Sekundärdaten
von den Finanzdienstleistern und den lokalen Verwaltungen gesammelt.
Deutsche Kurzfassung
XX
Diese kumulative Doktorarbeit ist in vier Hauptkapitel (Analysekapitel) unterteilt.
In jedem dieser Kapitel liegt der Schwerpunkt auf einer anderen Ebene des
Finanzsystems, entsprechend werden schwerpunktmäßig unterschiedliche Analysemethoden
angewandt. Das erste Kapitel beschäftigt mit dem ländlichen
Finanzsystem in seiner Gesamtheit. Hier wird der Transformationsprozess des
ländlichen Finanzsystems untersucht. Methodisch wird eine Literaturanalyse mit
auf Fallstudien beruhenden Aussagen, die während der Feldphase erhoben wurden,
verknüpft. Das zweite Kapitel beschäftigt sich mit dem größten Kreditgeber
an arme ländliche Haushalte in Nordvietnam, der VBP. In diesem Kapitel werden
Sekundärdaten der VBP und der lokalen Administration in Kombination mit
qualitativen Erhebungen und Experteninterviews gemeinsam unter Gesichtspunkten
der Neuen Institutionenökonomie analysiert. Hierbei liegt der Schwerpunkt
auf Informationsfluss, Transaktionskosten und Prinzipal-Agenten-Systeme.
Das dritte und vierte Analysekapitel behandelt hauptsächlich die Haushaltsebene.
Im dritten Kapitel werden die Breitenwirkung der staatlichen Finanzdienstleister
anhand einer Prinzipal-Komponenten-Analyse untersucht und Zugangsbeschränkungen
der Haushalte zu formalen Krediten durch eine Logit-Regression analysiert.
Beide Analysemethoden basieren weitestgehend auf quantitativen Haushaltsdaten.
Das letzte Analysekapitel entwickelt mit Hilfe eines Marktforschungsinstrumentes,
der Conjoint Analyse, angepasste hypothetische Finanzprodukte.
Hier werden quantitative Haushaltsdaten mit partizipativ erhobenen
qualitativen Daten in der Analyse kombiniert.
Der ländliche Finanzmarkt in Nordvietnam ist immer noch stark dominiert durch
hoch subventionierte staatliche Finanzdienstleister. Dadurch wird jeder Wettbewerb
und das Entstehen von nachhaltigen Finanzprodukten sowie die Entstehung
von Innovationen im Bereich Finanzdienstleistungen stark behindert.
Durch die Gründung der Vietnamesischen Bank für Sozialpolitik (VBSP), welche
die legale Nachfolgerin der VBP ist, wurden die kommerziell ausgerichteten
staatlichen Banken, insbesondere die VBARD, von ihren sozialen Aufgaben
befreit. Erkenntnisse aus anderen Entwicklungs- und Transformationsländern haben
gezeigt, dass Banken sich aus Kostengründen auf größere und reichere
Kunden konzentrieren. Daher wird vermutlich die VBARD ihren Kleinkundenanteil
stark zurückfahren. Diese werden nun von der VBSP bedient werden
müssen. Hier kommt die Frage auf, wie lange kann die vietnamesische Regierung
die defizitäre VBSP finanzieren und werden arme Kleinbauern, falls die
VBSB ihren Betrieb einstellen muss, keinen Zugang mehr zum formalen ländlichen
Finanzmarkt haben? Die Nachhaltigkeit des gesamten formalen Finanzsystems
ist bedroht durch ein ungelöstes Problem von schlechten Krediten, die
hauptsächlich von Staatsunternehmen gehalten werden. Es sind jedoch Hinweise
Deutsche Kurzfassung
XXI
zu finden, dass dieses Problem langsam auch auf den privaten Sektor übergreift,
insbesondere auf ländliche Haushalte. Abgesehen von den makroökonomischen
Folgen der schlechten Kredite, verhindert dieses staatlich tolerierte Verhalten
von "Moral Hazard" das Entstehen von nachhaltigen Finanzinstitutionen.
Die staatlichen Finanzinstitutionen im ländlichen Raum haben eine außerordentliche
Breitenwirkung. Jedoch nehmen die ärmsten Haushalte nur selten einen
Kredit bei diesen Institutionen auf. Nichtsdestotrotz hat Armut, gemessen an
einem Index (zusammengesetzt aus unterschiedlichen Armutsindikatoren), keinen
signifikanten Einfluss auf den Zugang zu Kredit. Das bedeutet, dass die
ärmsten Haushalte wahrscheinlich eine geringere Nachfrage nach formalen Kreditprodukten
haben. Eine einfache Verbesserung der Kreditprodukte oder die
Bereitstellung von komplett neuen Kreditlinien würde also nur geringfügig den
Zugang ärmerer Haushalte zu formellen Krediten verbessern. Ein vernünftigerer
Ansatz wäre die Schaffung eines auf ärmere Haushalte spezialisierten Beratungsdienstes
kombiniert mit einer verbesserten Infrastruktur bzw. Marktanbindung.
Eine verbesserte Marktanbindung kreiert nicht nur mehr Investitionsmöglichkeiten,
sie verbessert auch gleichzeitig den Zugang der Haushalte zu kreditrelevanten
Information und somit auch deren Zugang.
Insgesamt ist die Zahl der zu kreditzugangsbeschränkten Haushalte jedoch sehr
gering und die Anzahl der Haushalte mit einem formalen Kredit sehr hoch. Dies
ist zurückzuführen auf die Eliminierung von früheren Zugangsbeschränkungen,
z.B. durch eine breite Vergabe von Landnutzungstiteln, durch die Einführung
von Gruppenkrediten mit Gruppenhaftung und durch die hoch subventionierten
Zinssätze. Einerseits hat die Verringerung der zugangsbeschränkten Haushalte
aber auch dazu geführt, dass eine große Anzahl von nicht kreditwürdigen Haushalten
Zugang zu Kredit erhalten. Andererseits hat sich durch den sozialen
Charakter der staatlichen Finanzintermediäre eine schlechte Rückzahlungsmoral
verbreitet. Das deutet daraufhin, dass der Preis für die enorme Breitenwirkung
der staatlichen Finanzintermediäre eine verringerte Nachhaltigkeit dieser Institutionen
ist.
Trotz dieser enormen Breitenwirkung haben vor allem ethnische Minderheiten
und von Frauen geführte Haushalte immer noch Schwierigkeiten, Zugang zu
Kredit zu erlangen. Um den Zugang für solche Randgruppen zu verbessern, sollten
allerdings keine generellen Schritte unternommen werden, sondern speziell
auf diese Haushalte abgestimmte Maßnahmen durchgeführt werden. Die Gründung
der VBSP geht jedoch genau in die entgegengesetzte Richtung. Hier wurde
versucht, die schon ohnehin sehr große Breitenwirkung, durch generelle Maßnahmen
noch zu vergrößern. Allerdings ist es fraglich, ob die Haushalte, die bis
Deutsche Kurzfassung
XXII
heute die Kredite nicht nachgefragt haben bzw. zugangsbeschränkt waren, die
Kredite von der VBSP nachfragen werden. Eine nachhaltigere Vorgehensweise,
um den Kreditzugang dieser Randgruppen zu erhöhen, könnte ein verbesserter
Zugang zu kreditrelevanten Informationen sein. Es hat sich gezeigt, dass Randgruppen
oftmals einfach dadurch zugangsbeschränkt sind, dass sie entsprechende
Informationen nicht oder nicht rechtzeitig erhalten haben. Die staatlichen Finanzintermediäre
verlassen sich bei der Verbreitung dieser Informationen zu sehr auf
ein Netz aus lokalen Beamten, die allerdings über keine Bankausbildung verfügen
und dazu tendieren, bestimmte Bevölkerungsschichten zu übergehen. Hier könnten
zum Beispiel speziell ausgebildete und bei den Banken angestellte Frauen bzw.
Angehörige von ethnischen Minderheiten einen besseren Informationszugang
bzw. eine besseres Bewusstsein für die Probleme dieser Randgruppen in den
Institutionen schaffen.
Im Gegensatz zu der enormen Breitenwirkung von formellen Krediten sind formale
Sparprodukte fast nicht verbreitet. Dies ist jedoch nicht in der Unfähigkeit
der ländlichen Bevölkerung zu sparen begründet, sondern darin, dass die angebotenen
Sparprodukte nicht den Bedürfnissen der ländlichen Bevölkerung entsprechen.
Lokal angebotene Sparprodukte besitzen das größte Potential und sind
die nachhaltigste Maßnahme, um Haushalte, die derzeit keine formale Kredite
nachfragen, in das formale Finanzsystem zu integrieren.
1 Introduction
Vietnam has been engaged in a transition process enhancing the role of market
forces in the economy since the mid-1980s. Within this process, known as
‘Doi Moi’ (renovation), a change took place from a centrally planned economy
to a market-oriented economy with a socialist face (BRYANT 1998). During this
transition, the adaptation of the financial system was one of the most challenging
reforms Vietnam had to undertake (TEUFEL 1997; TRAN and DANG 1996). One
of the major tasks of this reform was to expand the financial system’s outreach
to the newly emerging private sector and household economies, not only in urban
but also in rural areas (WORLD BANK 1995; WORLD BANK 1998).
As various scholars have pointed out, broad access to appropriate and sustainable
financial services has vast potential for poverty reduction (BUCHENRIEDER and
HEIDHUES 2003; HEIDHUES 1995; KANBUR and SQUIRE 2001; SCHRIEDER 1996;
SCHRIEDER and SHARMA 1999; SHARMA 2001; ZELLER et al. 1997; ZELLER 1999).
Thus, the Vietnamese government proposes financial services as a powerful tool
for poverty reduction, too (SIDA-MARD 1998). Therefore, state-owned and administered
formal financial intermediaries (FFIs) such as the Vietnam Bank for
Agriculture and Rural Development (VBARD), the Vietnam Bank for the Poor
(VBP), and the People’s Credit Funds (PCFs) were established and promoted to
provide credit to the rural population.1 Evidence from other developing countries
suggests that the poorest population groups are excluded from formal financial services,
especially credit.2 This observation appears to apply to Vietnam, too.
Despite general successes in terms of credit outreach, certain population
groups, particularly the poorest, which are often identical to ethnic minorities,
1 On March 11 2003, the VBP and the PCFs were replaced by the Vietnam Bank for Social
Policies (VBSP) (VIETNAM ECONOMY 2003; WORLD BANK 2003).
2 For instance, AMIN et al. (2003) find that microcredit programs in Bangladesh do not reach
the credit-constrained.
Chapter 1
2
seem to be bypassed by the above mentioned financial institutions (NEEFJES 2001;
WORLD BANK 2003). Furthermore, the strategy of the national rural financial policy
has focused mainly on credit. Other financial services that potentially have a broader
outreach to the poorer market segment, such as different savings or insurance
products, have been neglected in the development efforts of the government.
1.1 Conceptual framework and objectives
The overall objective of this research work is to create knowledge about the factors
that impede or support access of rural households in Northern Vietnam to the formal
financial system on the one hand, and the outreach of this system to rural
households on the other. Thus, the conceptual framework of this research centers
on access and access constraints of rural households to the formal financial system.
Access can be hampered at different levels of the financial system. For instance,
rural households may be actively (e.g. due to particular policies or due to selfexclusion
because of a low risk-bearing capacity) or passively (e.g. due to unsuitable
financial services) excluded from access to the formal financial system.
Equally, an inadequate infrastructure hampers access and outreach, as transaction
costs (TCs) rise for households and lenders alike. Figure 1-1 presents the conceptual
framework of access and outreach constraints analyzed in this research. In
this context, the focus is more on the household side (access) as previous research
focused mainly on the supply side (outreach). However, knowledge about the
formal financial system and its institutions is also preeminent for the analysis.
Figure 1-1: Factors of access and outreach relating to the financial system
Outreach constraints
FFI
Policies Attitudes
Hard/
soft skills
Structure/
products
Access constraints
Human
capital
Social
capital
HH
Physical
capital
Infrastructure
Source: Own Figure, based on considerations made by EVANS et al. (1999) and VAESSEN (2001).
Notes: HH = household; FFI = formal financial intermediary.
Introduction
3
Figure 1-2 illustrates the formal financial system and the different research levels
that were suggested by the conceptual framework. The upper part of Figure 1-2
depicts the rural financial system in general with its various actors, while the
lower part presents the different research levels and the corresponding data collection
methods. As Figure 1-1 has demonstrated, access/outreach barriers exist
at different locations in the formal financial system, and so a joint view/analysis
of the three levels depicted in the lower part of Figure 1-2 is necessary.3
Figure 1-2: The rural financial market and the levels of analysis
Community level*
• Rural/community
infrastructure
• Secondary data of
rural communities
• Expert interviews
• Semi-structured
interviews with
community key
persons
Intermediary level
• Secondary data
• Semi structured
interviews with bank
staff of formal
Village/household
level
• PRAs
• Semi structured
interviews with
village key persons
• Conjoint survey
• Standardized
household survey
Government,
Central Bank
Financial market policy
Monetary policy
Trade policy
Exchange rate relations
Rural infrastructure
Agricultural sector
Extension system
Off-farm activities
Formal
Semi-formal
Informal
Insurant
Saver Borrower
Financial
intermediaries
HHs Financial
intermediaries
Rural financial
market
SMEs
Government
agencies
The rural financial system
The level of analysis
Source: Own figure.
Notes: * The community level is defined as all institutions, policies, and infrastructure above village
level. PRA = Participatory rural appraisal, HH = Household, SME = Small and medium
enterprise.
3 As the formal insurance sector plays only an insignificant role in rural Vietnam
(DUFHUES et al. 2004b), this research work covers mostly the credit and savings sector.
Chapter 1
4
The overall condition of the Vietnamese financial system is of particular importance
as the transformation of the system is still going on. Failures within the
transformation process are likely to weaken the sustainability of the financial
system. A sustainable financial system with likewise sustainable financial intermediaries
supports the long-term access of households to the formal financial market.
As pointed out above, access constraints can also be found on the side of the
financial intermediary. The VBP has been the main supplier of credit to the rural
poor in Northern Vietnam (VBP 1999b). However, it is reported that households
are often access-constrained simply because they lack sufficient information
about the credit application procedure or about the availability of funds.
KANBUR and SQUIRE (2001) state that information deficits undermine the access
of marginalized groups to credit. It appears that this information deficiency
contributes to resource misallocation in the sense that potential borrowers with
profitable investments are excluded due to a lack of information.
From the point of view of the household level, it is crucial to know who is actually
being access-constrained and why? For instance, HULME and MOSLEY (1996)
point to increasing evidence from other developing countries that the poorest
20% of the population are effectively excluded from microcredit programs.
While the Vietnamese government has so far failed to create sustainable rural
financial institutions, it has succeeded in providing a huge part of the population
with formal credit. Nevertheless, the questions remain: Did they succeed in
reaching the poor and do groups of people exist who are still accessconstrained?
There is a universal tendency by government officials in developing countries
to assume that they know what the poor need, and to seek control over how
these needs are met, which very often results in services with very limited outreach
as these services often fail to meet the true needs of the rural population
(GIBBONS et al. 2000). Thus, there is a pressing need to examine participatory
ways of designing and introducing new financial services into microfinance institutes
(MFIs) to improve their outreach and access to them (RUTHERFORD 2000;
WRIGHT 1999). This is especially true for Vietnam, where almost all decisions
are made using an top-down approach and local inhabitants have no impact on
decision making or on policy (DU 2001; FFORDE 2004).
The central hypothesis of this study is that the transformation polices of the
Vietnamese government concerning the rural financial system are biased towards
the supply of preferential credits, and that this policy discriminates against the
poorest households. Derived from the central hypothesis, the specific objectives
of this thesis are:
Introduction
5
- To describe and review the current transformation of the financial system
in Vietnam in terms of its sustainability, with a special focus on the
implications of the financial transformation for the rural financial market.
- To reveal information deficiencies in the credit procedures within the
VBP by identifying crucial gatekeepers of important credit information
and by describing the formalized and non-formalized information channels.4
- To determine the poverty outreach of the FFIs in Northern Vietnam and to
analyze which population groups are still access-constrained as regards
the formal system.
- To analyze the supply of formal financial services, and to design clientoriented
financial services by means of participatory research.
1.2 Organization of the study
The thesis is organized in a cumulative style based on four individual articles published
in different journals. Each of the main sections represents one article. Section
two reviews the transformation process of the financial system in Vietnam,
with special attention to the rural financial market. Methodologically, an in-depth
literature review is carried out, supplemented by anecdotal evidence gathered
during the field research. In the third section, the information polices of the VBP
as the main supplier of credit to the rural poor in Northern Vietnam are investigated
by means of information economics analysis. Methodologically, secondary
data from the VBP and the local administration are combined with qualitative
evidence from the intermediary and household level. The fourth section analyzes
the poverty outreach of the FFIs by a Principal Component Analysis (PCA) and
household access constraints to formal credit by a binary logit analysis. Both
analyses are based on quantitative household data. In the fifth section, clientadapted
financial services are developed using a Conjoint Analysis approach,
which is a marketing research tool and combines quantitative and qualitative data
in its analysis. Finally, the thesis ends by drawing general conclusions.
4 As already mentioned above, the VBP has been replaced by the VBSP. However, results can
be easily extended to the VBARD, which were closely related to the VBP, and to the new
VBSP as well, as main structures seem to be similar compared to the old VBP/VBARD
concept.

2 The transformation of the financial system in Vietnam
and its implications for the rural financial market –
An update5
The following section describes the transformation of the financial system in
Northern Vietnam. Special emphasis is given to the rural financial market.
2.1 Introduction
The transition process of Vietnam is well known in the economic world as
‘Doi Moi’. Doi Moi literally means ‘change’ and ‘newness’ and is the Vietnamese
term for reform and renovation. It is aimed at restructuring Vietnam’s
legal, regulatory, administrative, investment and foreign trade apparatus and
policies to transform its centrally planned economic system into a market economy
with ‘socialist characteristics’ (BRYANT 1998).
The change from a centrally planned to a market-oriented economy in the 1990s
proved to be a difficult process and one that was anything but self-implementing.
In contrast to other regions in transition, e.g. the former Soviet Union, Vietnam
emphasized gradualism over radical change, with economic restructuring to
come before privatization (CHIN and GUAN 1996; MONTES 2001). Political pluralism
or democratization were not promoted and the Communist party has still
retained its political monopoly of power (LUIBRAND 2002). However, the reforms
have achieved impressive results. The most remarkable feature of the transition
in Vietnam is the avoidance of a decline in output and immediate significant
growth in agriculture, industry, and services resulting in average growth rates in
gross domestic product (GDP) of 8.1% between 1990 and 1999, a decline in
5 This section is based on the following article: "The transformation of the financial system
in Vietnam and its implications for the rural financial market – An update", written by
Thomas Dufhues and published 2003 in the Journal for Institutional Innovation, Development
and Transition 7: 335-362. I would like to thank Dr. Gertrud Buchenrieder for her
helpful comments on this article.
Chapter 2
8
inflation to the single-digit range, and foreign direct investment flooding into the
country (WORLD BANK 2000a).
Looking at the impressive economic accomplishments of the last decade, one
might be persuaded to believe that Vietnam has already laid the foundation for a
dynamic, prosperous market-oriented economy. However, this would be a mistake
(RIEDEL 1999). Critical voices concerning the speed of transition in Vietnam
started to emerge in the last few years. The reforms that were launched have
slowed down, or in some cases have even been discontinued (LUIBRAND 2002).
Besides, doubts have been raised as to the government’s general commitment to
push economic reforms any further (RONDINELLI and LITVACK 1999). Vietnam’s
current party leaders see a ‘New’ Doi Moi as potentially more threatening than
the status quo (THAYER 2000). Nevertheless, it is well recognized, both inside
and outside the government, that without further reforms the strong performance
of the past cannot be sustained (RIEDEL 1999).
The structural features of transition economies differ in some important respects,
but they share similar fundamental conditions that give rise to a similar set of
reform issues. Worldwide experience since the early 1980s indicates that a variety
of institutions must be developed for transition (LUIBRAND 2002). This is particularly
true for the financial sector or, in the words of WOLFF, (1999:59): "The
reform of the financial sector in a transforming economy primarily means institution-
building."
In the transition process, the reform of the national financial system plays a
key role because the efficiency and the speed of the reforms have a decisive
influence on the success of the transformation process in the other sectors
(SCHRIEDER and HEIDHUES 1998). Adaptation of the financial sector was one of
the most challenging reforms that Vietnam had to undertake. After a decade of
financial sector reforms, Vietnam expected to see the emergence of an effective
and efficient banking system (TEUFEL 1997; TRAN and DANG 1996). However,
the financial sector is still weak, legislative issues are incomplete and it suffers
from a substantial amount of bad debt (LUIBRAND 2002).
This paper aims at describing and reviewing the transformation process and
status of the financial system in Vietnam. Moreover, possible future developments
are presented and evaluated. Special attention is given to the implications
of the financial transition for the rural financial market. Methodologically, this
research is based on an in-depth literature review combined with anecdotal evidence
gained during field research in Vietnam from 2000-2002.
Transformation of the financial system
9
2.2 The financial landscape in Vietnam
In 1990, the former mono-bank system was changed to a two-tier banking system
consisting of the State Bank of Vietnam (SBVN) as central bank and supervisory
institution (tier 1), and an operating system (tier 2).
The central bank, SBVN: Until the early 1990s, independent central banks were
assumed as important for growth and particularly for ensuring low inflation
(TOMPSON 1998). Few institutional recommendations by economists have gained
such rapid and wide acceptance as that of granting central banks independence
from short-term political control.6 However, while central bank independence has
a clear correlation with low rates of inflation, there is no causal link between the
two. Independence of a central bank alone does not make it an effective institution.
Its independence must receive the support of the political system (POSEN 1993).
The policy implementation for developing countries is that, in order to ensure low
inflation, the best course of action is to undertake financial sector reforms (e.g.,
liberalization and privatization) rather than instituting easily reversible and practically
meaningless changes in their legal/institutional structures (MAS 1995).7
KOVSTED et al. (2003) assess the SBVN as politically and operationally dependent
upon support from government agencies. In some cases, the SBVN has not made
decisions even on issues that were within its power, but has instead relied on
guidance from the government. This indicates that the SBVN still does not have
the status to independently execute national monetary policies. Obviously the
binding constraint is not the lack of legislative independence. Therefore, an indirect
route to obtain central banks independence by gradually building and
strengthening a non-governmental lobby for an independent central bank could be
more promising than relying solely on the legislative transformation. The preliminary
plans for equitization of the state-owned banks represent an important step in
this direction, as does the continued presence of foreign banks in Vietnam.
6 BOWLES and WHITE (1994), however, note that scholarly proponents of central banks independence
are less likely to take the case as proved than policy-makers and financial journalists.
Moreover, among academics the degree of independence of a central bank is still
the subject of much controversy (STIGLITZ 2002).
7 This might explain why the SBVN has been very successful in fighting inflation in the past
decade, despite its strong dependence on and close relations with the government (KLUMP
and SPITZENPFEIL 1998).
Chapter 2
10
State-owned commercial banks (SOCBs): There are four large SOCBs in
Vietnam: The Foreign Trade Bank of Vietnam (Vietcombank), the VBARD, the
Industry and Commerce Bank of Vietnam (Incombank), and the Vietnamese
Bank for Investment and Development. Before the financial reform (1988-1990),
the SOCBs were departments of the SBVN. These financial institutions are the
leading banks of the banking system and have a total of more than 1,200 branches
in Vietnam. Altogether, the number of staff working for the SOCBs is around
40,000 (KOVSTED et al. 2003). In the past, these banks solely served their specified
sectors. Today, these strict sector constraints have been abolished.
Policy Bank: The VBP was established in 1995 as the poor people’s lending
outlet of the VBARD.8 The VBP is called a policy bank, because the government-
mandated purpose of the VBP is not to maximize profit but to reduce
poverty (VBP 1999b). The VBP specializes in lending to poor households, primarily
the rural poor.9 The bank basically consists of a head office and has no
physical structures of its own below this level. The VBP uses the operational
facilities and staff of the VBARD and of local authorities in extending its services.
Nevertheless, due to its highly subsidized interest rates, many international
agencies consider the VBP to be financially unsustainable.10
Joint Stock Banks (JSBs): At present, 36 commercial banks exist in the form of
JSBs (VIET NAM NEWS 2002a). The first JSBs were founded in 1989. The majority
of JSBs were quickly established in the years following the initial liberalization of
the financial sector. Some of them are owned jointly by the state and private
groups and individuals; others are completely in private hands. In the beginning
the majority of stock was held by state-owned enterprises (SOE), which were
trying to diversify their business activities (KLUMP and SPITZENPFEIL 1998).
Foreign Banks and Joint Venture Banks: There are 15 branches of foreign
banks and four joint venture banks. In addition, 62 representative offices from
20 nations operate in Vietnam; these include major international banks such as
8 As mentioned before, the VBP was replaced by the VBSP.
9 Only a certain portion of the population is eligible to obtain a loan, namely the rural poor.
The Vietnamese government classifies every household into one of five classes according
to its living standard: Hungry, poor, medium, better-off, or rich (DUFHUES et al. 2002).
10 The VBP has been recognized as loss-making since the late 1990s (VBARD and DANIDA
1999).
Transformation of the financial system
11
Deutsche Bank, or Bank of America and CitiBank. The existing legislation,
however, still blocks foreign banks from becoming fully fledged participants on
the Vietnamese financial market. Despite recent progress, the Vietnamese govern
ment has yet to establish a level playing field for competition between foreign
and domestic banks (KOVSTED et al. 2003).
People Credit Funds (PCFs): After the collapse of the rural credit cooperatives in
the early 1990s, it was quickly realized that the VBARD could not fill the void left.
The system of PCFs was established in 1993 to fill the gap (WOLZ 1999). In June
2001, a total of 947 PCFs existed, supervised by the SBVN, and having a total of
about 714,000 members (GTZ 2002). The intention underlying the creation of the
PCF system was to create a three-layer organization in order to achieve a combination
of close local contacts and connections while minimizing the risks associated
with seasonality and regional shocks. As a consequence, all local credit funds were
handled and directed by regional funds, which in turn would be supervised by a
central credit fund handling the supply and balancing liquidity among the regional
funds. The deposit and credit market shares of the PCF system are relatively small,
amounting to a mere 1-2% of total market volumes (KOVSTED et al. 2003).
Insurance and leasing companies: Up to the present, 18 insurance companies
(state-owned, joint ventures and foreign companies) exist in Vietnam (DUNG 2002).
A new Law on Insurance Business became effective in April 2001. Although the
private sector’s market share is on the rise, the state-owned insurer remains the
dominant player, with non-life and life market shares of approximately 47% and
55% respectively. Coverage is low, with annual insurance premiums amounting
to 1% of GDP. But premiums have grown at an average rate of 30% per year,
reaching an estimated 321 million US$ in 2001. Vietnam also has eight finance
leasing companies, three of which are either joint ventures with foreign investors
or wholly foreign-owned. Five are subsidiaries of the SOCBs. In 2001, the value
of leased assets amounted to 131 million US$ (WORLD BANK 2002b).
Postal savings service: The Vietnam Postal Savings Company (VPSC) was established
in 1999 and operates under the authority of Vietnam Post and Telecom. Its
main function is to provide a savings product, thus mobilizing savings for government
development investments. The operating costs of the VSPC are de facto subsidized
by the Vietnam Post and Telecom through the use of its staff, as the VSPC
itself employs only 100 persons. The VPSC sees computerization as its biggest
obstacle and opportunity. Moving from paper-based to computerized operations
would be an expensive endeavor. This has, however, been identified as a priority to
increase the value of existing services and would facilitate the expansion of services
to include money transfers between accounts, payment of utilities, etc.
Chapter 2
12
A money transfer service between savings accounts within the VSPC has been
launched in Hanoi and extended to northern provinces. Recently, expansion to
Ho-Chi Minh City was started, and the service is to be extended nationwide by the
end of the year (WORLD BANK 2002a).
Stock exchange: Vietnam has been endeavoring to establish a stock exchange
since 1992. However, the differences between reform-oriented and conservative
politicians have repeatedly delayed important decisions concerning the stock
market. An essential step towards the creation of a stock exchange was the
building of the State Securities Commission in 1996, which is responsible for
the planning, implementation and performance of the stock exchange. After the
Asian Crisis in 1997, the conservative politicians gained the upper hand and the
establishment of the stock exchange was postponed again. In 2000, however, a
stock trading center was opened in Ho-Chi-Minh-City, which is a strong signal of
change in the direction of a market economy. Nevertheless, this stock trading center
is highly regulated by the government (NGUYEN 2002). Currently 19 companies
and 18 bonds are listed, with a total market capitalization of 105 million US$.
All companies listed are former SOEs that have been transformed into jointstock
companies. Trading was upgraded in December 2001, and a clearing and
settlement system consistent with best international practices will be introduced
by 2005 (WORLD BANK 2002b).
2.3 Reforms in the financial market
Prudential reforms of the financial sector are most successful in countries with
economies similar to Vietnam’s economy today. A banking system with some
problems, but without disasters, is most capable of pursuing reforms. Vietnam
therefore offers a good starting position for further reforms in the financial sector
(ARDREY et al. 1999).
2.3.1 Reforming steps in the banking sector
The unwillingness of the authorities to undertake thorough reform of the financial
sector in the past is a result of their continuing desire for a ‘market economy
with a socialist orientation’, in which the state occupies the ‘commanding
heights’ of the economy. Although private banks have been allowed to operate,
reforms so far have been careful not to threaten the dominant position of the
state-owned banks. There appears, however, to be a growing recognition on the
part of state authorities that the financial sector cannot be managed the same
way as other sectors because mistakes reverberate throughout the entire
economy, threatening not only financial institutions, but the economy as a whole
Transformation of the financial system
13
(RIEDEL and TURLEY 1999). Despite recent reforms (e.g. deregulating the interest
rate regime or putting in place restructuring plans for SOCBs), the
WORLD BANK (2002b) states that the efforts to transform the banking sector
into a commercially-based operation are still in their early stages.
2.3.1.1 Competition
The previous banking reforms implemented in Vietnam during 1988-92 were
substantial and contributed to solid macroeconomic performance. By 1994 (see
Table 2-1), the banking system had changed considerably, in the sense that other
banks began to take a growing portion of banking assets (WORLD BANK 2000c).
But the speed of introduction and implementation of reforms declined at the
end of the 1990s, leaving the banking sector to be dominated by SOCBs
(LUIBRAND 2002).
Table 2-1: Increasing number of players in the banking sector
1990 1994 1999
SOCBs1 4 4 5
JSBs 0 36 48
Joint venture banks 0 3 4
Branches and rep. offices of foreign banks 0 41 103
Source: WORLD BANK (2000c).
Notes: 1 Including the policy bank, VBP.
The competitiveness of the Vietnamese financial system is ranked very low
(UNIDO et al. 2003). SOCBs dominate the financial sector in Vietnam more
than they do in other developing countries, where governments usually control
only around half of the banking system (THAI 1998). According to OH (1999)
and WORLD BANK (2000c), the major SOCBs hold approximately 75-80% of
total assets in the banking sector. In addition, the SOCBs are large shareholders
in the JSBs and, as markets for SOCBs and JSBs are completely separate in
terms of depositors and borrowers, the share of assets under complete or partial
government control could be much higher. Due to the segmented financial markets,
competition is not of great concern to SOCBs. After all, the Vietnamese domestic
financial market remains highly distorted and segmented (MCCARTY 2001).
Given that the concentration in the Vietnamese financial market coincides with
substantial state ownership, one option to enhance competition would be
privatization of the four major SOCBs. This, however, is a lengthy process,
starting with the recently completed international audits and subsequent attempts to
address the non-performing loan problem in the SOCBs, which must precede
any attempt to recapitalize the major SOCBs (see section 2.3.2). Only after this
Chapter 2
14
has taken place would it be realistic to consider any form of privatization of the
SOCBs. Another approach to making financial markets more contestable is
liberalizing the comparatively strict entry requirements. Nevertheless, due to
potential risks associated with foreign bank entry, introducing foreign competitors
may require an initial transition period, to allow time for efficiency adjustments
in the domestic sector and for improvements in prudential regulation and
supervision. Because bank failures will almost inevitably occur during this process,
the government needs to establish transparent rules for bank exit. As a consequence,
any liberalization of the entry process must be both managed over time
and transparent (KOVSTED et al. 2003). The development of a competitive banking
sector in Vietnam will not only promote increased efficiency of the banks and
improved banking services but also promote the development of a competitive
private business sector, because borrowers will not be limited to a few banks
that serve only selected clients (THAI 1998).
The lack of competition in the rural financial market is even more severe than in
the urban financial market. Here, the state-owned banks (VBARD/VBP) are the
main (only) supplier of financial services.11 The VBP is completely dependent
on the VBARD as they use the same staff below headquarters level. The
VBARD serves the better-off clientele in the rural market, while the VBP serves
the poorer market segment.12 There is thus no real competition between the two
(DUFHUES et al. 2004a). It is not possible for other financial organizations
(e.g. non-governmental organizations, PCFs) to compete with the VBARD/VBP,
which are recipients of huge state subsidies covering both operating and financial
costs. Under such circumstances, financial innovations are unlikely to occur and
no market inducements are given to the quasi-monopoly market leaders,
VBARD/VBP, to improve their performance. But as LLANTO (2000) states, the
subsidies enable the VBARD and VBP to penetrate poor areas. However, they
11 PCFs do not exist in every province and have very limited outreach compared to the VBP/
VBARD.
12 The orientation to a special market segment reduces portfolio diversification, which can
endanger the long-term sustainability of a bank. The situation is aggravated by the fact that
the rural banks in Vietnam mostly financed investments related to agricultural production.
Possibilities for diversifying the portfolio of rural financial institutions in post-socialist
economies are generally limited as only a few small and medium enterprises existed during
the centrally planned era (SCHRIEDER and HEIDHUES 1998).
Transformation of the financial system
15
are doing so at great expense to the taxpayer and at great cost to the development
and implementation of sustainable financial services.
2.3.1.2 Interest rate policy
Countries that do not possess highly developed financial sectors may use interestrate
ceilings as a means of protecting emerging institutions. As financial sectors
mature, restrictions on competition can be removed provided that adequate systems
of prudential supervision and control of monetary growth are in place
(COLLIER and MAYER 1989). In making this decision, it is important to consider
how far advanced the country is in reforming the state enterprise sector and in
establishing a ‘credit culture’ – that is, the extent to which banks have become
accustomed to using market principles in assessing credit risks. Countries with
inadequate regulatory and supervisory frameworks, or whose financial institutions
are insolvent, are likely to run into serious problems if they liberalize interest
rates too early or too rapidly. For interest rate liberalization to succeed, the main
economic players need to be subject to hard budget constraints so that they will
avoid borrowing and lending unwisely. Otherwise, credit could be directed to socalled
pathological borrowers – those who would like to take the greatest risks and
who would borrow no matter how high the cost (MEHRAN and LAURENS 1997).
When financial deregulation is implemented – and especially where non-performing
loans are inherited from the pre-reform era – interest rate liberalization should
be accompanied by structural reforms, including restructuring bank balance sheets
to remove bad debt, privatizing state-owned banks, and introducing measures to
promote competition in the banking sector (PILL and PRADHAN 1997).
During the last decade, the Vietnamese Government has set an interest rate ceiling
through the SBVN. In recent years, greater flexibility in the management of loan
interest rates has been introduced, e.g. the interest ceiling on foreign loans has
been abolished. For lending in dong, banks have for some time been allowed to
offer interest rates up to a new ceiling rate, defined as the base rate plus 0.3% per
month for short-term loans and 0.5% for medium-term loans (WORLD BANK 2001).
However, the real interest rate was lowered every year since they became positive
(SENANAYAKE and HO 2001). While the standard interest rate was relatively
high in real terms, the various preferential rates were close to the actual inflation
rate and often below (DIEHL 1998). Finally, after a decade of pre-fixed interest rate
ceilings, the SBVN has removed the lending rate ceiling (WORLD BANK 2002a).
There is little experience as yet on what the true impact of the recently implemented
rate liberalization will be on banks’ rate-setting behavior, and thus on
their future profitability. For now, differentiation between banks and types of
borrowers remains limited (WORLD BANK 2002b). Despite the possibility of raising
Chapter 2
16
lending rates under the past interest rate cap regime, the rates stayed quite inelastic
and unresponsive to increased demand. The large SOCBs were and are
clearly the rate-setters. The SOCBs explain this inelasticity as a fear of competition
and a perceived duty to advance their clients’ business prospects through
providing good rates. This could imply quite an unusual type of pricing by the
large SOCBs, maintaining lending rates at concessional levels despite their own
interest margins possibly being insufficient to cover their costs. As interest rate
liberalization is a very recent phenomena, it is understandable that the banks are
wary of taking aggressive steps towards a more market-driven competitive approach
(WORLD BANK 2002a). Nevertheless, similarly to the central bank, the
decision makers within the SOCBs do not fully exploit their opportunities. The
transformation to a market-oriented economy obviously needs more than an
enabling legal or regulatory framework; it also needs business actors who are
willing and able to carry out these reforms.
Interest rates of rural credit institutes are still highly subsidized and subject to
close government control.13 The provision of subsidized interest rates to the rural
population has for a long time been the subject of debate between international
donors and the government of Vietnam. The donors point to international experience
with subsidized rural credit programs having resulted in massive loan
losses, low savings mobilization, and bank failures.14 In the past, subsidized interest
rates led to a rush for loans from VBP and VBARD, resulting in the common
impression that subsidized loans are a right rather than a contractual
agreement. In addition, the perception that loans at a higher rate are exploitative
makes a shift to market-based rates increasingly difficult (KOVSTED et al. 2003).
The cap on lending rates during the last decade and the view of decision makers
that savings are not important for rural development have led to a situation
where credit-oriented rural financial institutes, VBARD/VBP and now the Vietnam
Bank for Social Policies (VBSP), leave the rural population without access to
saving products. The main challenge will be to implement a safe, attractive and
cost-covering deposit collection system at the local level. However, before this
can happen, a paradigm change is called for. The official policies in Vietnam
13 The VBARD was only recently freed from lending at preferential interest rates.
14 A detailed overview of the negative effects of subsidized credits on rural development can
be found in (ADAMS et al. 1984).
Transformation of the financial system
17
need to recognize the capability and the demand of the rural population to save.
The implementation of the VPSC is a step in the right direction and it has, at
least potentially, the possibility to achieve very deep outreach.15 Nevertheless,
the policies of VBARD and VBSP have not changed so far in this regard.
2.3.1.3 Political lending
Historically, the state-owned bank sector was used as an instrument of public
policy, and during the 1990s much of its lending was still influenced by social and
political rather than commercial objectives (GOTTWALD and KLUMP 1999).
Nevertheless, over the past decade the four large SOCBs have slowly started to
evolve from specialized political lending vehicles to more commerciallyoriented
financial intermediaries, with the greatest progress seen in 2001 and
2002 (WORLD BANK 2001). Recent steps to separate political lending from commercial
lending at SOCBs resulted in the establishment of the VBSP, the successor
of the VBP. The VBSP has a branch network in all provinces. It enjoys privileges
such as having its liquidity and solvency guaranteed by the government, and being
exempt from deposit insurance and tax regimes (WORLD BANK 2002b). The VBSP
will take over all political lending from the VBARD. Thus, the VBARD, now freed
from political lending, will concent-rate purely on market-oriented lending. This
separation aims at enhancing financial transparency in the country’s banking system
(SGT DAILY 2003). Through the VBSP, the Vietnamese government will continue
its policy of providing preferential credits to ‘disadvantaged’ groups. As the
interest rates are not cost-covering, it is also quite reasonable to assume that the
VBSP will be one of the future holes into which public funds will drain.
KOVSTED et al. (2003) suggest taking the influence of the government on credit
allocation as an indicator for the progress of the transition from a centrally planned
economy to more market-based economy in Vietnam. The question is how
many funds will be channeled through the VBSP and to whom?
However, there is still no clear distinction between profit-orientation and the
social duties of banking, and the uniqueness of banks as financial intermediaries
is not well recognized in Vietnam. A bank is regarded as either a
15 In a country where 80% of the population lives in rural areas, even the branch network
of the VBARD cannot reach widely or deeply enough among the population at large.
The postal savings system offers the use of a structure that is obliged to have equal
Chapter 2
18
safe or a broker, or a distributor of government funds. This also affects the
general perception of banks and banking functions, especially among rural
households (KOVSTED et al. 2003; OH 1999). The founding of the VBSP may
be a step in the right direction. As political lending will be separate from commercial
lending, bank staff will at least have a clear view of each operation. As
the target group of the VBSP has been widened enormously, it seems fair to
assume that the VBSP will become the main supplier of credit to rural households
and may even outdo the VBARD.16 It is therefore also reasonable to suggest
that the rural population is unlikely to change its perception of rural financial
institutes as social welfare institutions.
All policy interventions of the Vietnamese government in the rural financial
market have the aim of extending credit outreach to the population. The reason
behind this policy objective is a centrally fixed production target. For instance,
the Vietnamese government has set as its target to provide 90% of poor households
with credit, regardless of whether there is a demand for it. This is reminiscent
of old socialist production planning. In a recent study, BUCHENRIEDER et al. (2003)
found that most rural households in Vietnam are well served with credit facilities.
Less than 15% of rural households are access constrained. Attaining the
aforementioned target will therefore be an immense drain on public resources. As it
is connected to political lending, it will probably also create a large number of
non-performing loans and a business culture that will likely hamper the creation
of any sustainable rural financial institution for years.
2.3.2 State-owned enterprises and non-performing loans
SOEs: The most difficult reforms that now confront Vietnam relate to the reform of
SOEs and their dependence on state credit (MONTES 2001; SHIMOMOTO 2003). The
explicit subsidies from the budget of the SOEs were stopped during the 1990s but
were partly replaced by increased credit from the banks (WORLD BANK 2000c).
After the establishment of the two-tier banking system, all enterprises were able
to borrow at the same interest rate. However, SOEs still continued to soak up the
bulk of domestic lending (MONTES 2001). The former interest rate policy of
geographic distribution (WORLD BANK 2002a).
16 Instead of serving only the ‘poor’, other disadvantaged groups, e.g. students, are now also
eligible for a loan.
Transformation of the financial system
19
granting preferential credits to SOEs squeezed banks’ profits (see Section 2.3.1.2 ).
This encouraged SOEs to accumulate excess liquidity beyond proper investment
opportunities, as capital costs were low. At the same time, it led banks to lose
motivation for profit-taking businesses (OH 1999; SHIMOMOTO 2003). Nevertheless,
the share of credits to SOEs has been lowered continuously from more than
90% in 1990 to 25% in 2002 (WORLD BANK 2002b). The rationalization of interest
rates is also an important step in paving the way for increased lending to private
small and medium enterprises (SMEs), which may not have an adequate credit
history or collateral to secure loans at administratively set low interest rates
(WORLD BANK 2001).
While efforts to transform and restructure SOEs have stayed on track so far,
the process appears to be slowing down. At present, the reform mechanisms in
place amount to an option, rather than a mandate, for enterprises to divest
(WORLD BANK 2002b). It is necessary to prevent badly performing enterprises
from obtaining new credit and to collect outstanding loans. So far, however, this
measure has been considered more from a political than an economic perspective,
as to do so might press SOEs into bankruptcy. Only few liquidations have so
far taken place (WORLD BANK 2000c). According to MCCARTY (1999), the reluctance
to close down SOEs may be influenced not only by fear of creating an
unemployment problem or by resistance on the part of enterprise management
due to fear of losing power; it may also be influenced by the Vietnamese vision
that can be traced back to the economics of Marx and Lenin, which is one of
steady accumulation: Step-by-step, more of everything. This is in contrast to
more contemporary arguments based on an understanding of scarcity, opportunity
costs, and the pace of the technological change. The crucial ingredient for a
successful modern economy is the ability to change – to open new businesses,
and to close down old ones.
Non-performing loans: Past emphasis on policy and direct lending has led to a
high share of non-performing loans among the SOCBs. An inappropriate regulatory
and supervisory system, one that did not focus sufficiently on risk and loan quality,
permitted JSBs, for example, to conduct related-party and high-risk lending
(WORLD BANK 2000b). According to the SBVN (2001), 13% of all outstanding
bank loans are non-performing, though some experts suggest the figure could be
Chapter 2
20
as high as 30% if international accounting standards were applied.17 A significant
portion of these non-performing loans was given to SOEs. Many of these
loans may never be recovered. While the exact value of non-performing loans is
difficult to estimate, the current situation amounts to a claim by the state sector
on future capital accumulation (WORLD BANK 2002b). The heavy reliance of the
JSBs on SOEs as a source of deposits and for lending and the strong interdependence
of the SOCBs and the SOEs may put the whole financial system in
jeopardy (LEUNG and RIEDEL 2000, WARNER 2001). The SBVN took some
initial steps towards improving the situation and closed several JSBs. Progress
with resolving old non-performing loans has been slow but steady. The main
difficulty is with unsecured loans, mainly to SOEs. SOCBs find it difficult to
meet their resolution targets, owing to delays in SOE reform. Although progress
with the resolution of non-performing loans varies across the four large SOCBs,
a first round of re-capitalization provided resources to each of them. It is not
clear, however, that the portfolio of the banks is seeing significant improvement.
The lack of a timely resolution process for non-performing loans hampers the
creation of a credit culture of repayment (WORLD BANK 2002b). However, in
some areas restructuring efforts have already begun. The government has, for
example, already developed a detailed restructuring plan for the Vietcombank.
During the next couple of years, a detailed restructuring plan for the VBARD,
Incombank and the BIDV will be developed on the basis of the recently completed
international accounting standard audits.18 Asset management companies will be
established for each SOCB to solve the problem of non-performing loans
(KOVSTED et al. 2003). Despite such instances of progress, the large quantity of
non-performing loans still threatens the whole financial system.
Most of the credit expansion in the last years has come from the SOCBs. In
view of their already high levels of non-performing loans and generally poor
expertise in credit risk assessment, such rapid expansion will further weaken the
17 Vietnam uses its own definition to classify loans as non-performing. Interest accrual on
non-performing loans is allowed for up to 180 days for unsecured loans and 360 days for
secured loans. Loans overdue for 90 days are not adequately classified. The classification
system, moreover, does not reflect the credit risk based on the borrower’s repayment capacity,
collateral coverage, and other factors (KOVSTED et al. 2003).
18 The restructuring of the VBARD has already begun with the separation of the VBSP from
VBARD.
Transformation of the financial system
21
asset quality of these banks (WORLD BANK 2001).19 In 2000 and 2001, a substantial
part of the credit allocation to non-state sectors reflects the government’s
decision to ask SOCBs to extend loans for rural development purposes
(WORLD BANK 2001; WORLD BANK 2002b).20 These loans were usually soft
loans. One can assume that many of these recently disbursed loans to rural
households will end up as non-performing loans. Nevertheless, as stated above,
the VBSP will take over all policy loans from VBARD and VBP. Therefore, the
VBSP will start out with a huge burden of bad debts, casting great doubts on its
financial viability and on the statement made by Prime Minister Phan Van Khai
to the VIET NAM NEWS (2003), saying that the VBSP should not operate like a
subsidized financial body.
Due to the small profit margins, the VBSP will probably seek, as did the VBP
and VBARD, to shift obligations (training, screening, monitoring, etc.) to clients
or local authorities in order to save costs, leading to a less focused development
of skills and knowledge within the financial institutions (DUFHUES et al. 2002;
KOVSTED et al. 2003). One problem resulting from this practice is the lack of
contact between professional bank staff and the borrower. Hence, many households
that are not creditworthy are assessed as creditworthy, resulting in massive
rescheduling of loans and ultimately adding to the amount of non-performing
loans (BUCHENRIEDER et al. 2003; IZUMIDA and DUONG 2001).
2.3.3 Legal environment
The transformation process includes the reshaping of the legal environment. So far,
the changes in the legal system of Vietnam have been unable to keep up with the
developments of the economic reforms (DIEHL 1998; DUCKETT 2001). The findings
of a number of studies show that countries with legal systems that enforce contracts
more effectively have more highly developed financial systems. In Vietnam, the
19 As a source of overdue loans, SOEs and non-SOEs have shown a similar trend. The increase
in the proportion of overdue loans partly reflects the loss of impetus in the reform
process. The private sector’s share of overdue credits rose to 67% in 1997 from 41% in
1994, with the bulk concentrated in VBARD (OH 1999).
20 Of the SOCBs, it is mainly the Vietcombank and the VBARD that lend to private enterprises.
The reason for this is that these two banks have an extensive branch network, and
that their respective areas of specialization are in sectors that have seen rapid growth
following the initiation of the Doi-Moi reforms (KOVSTED et al. 2003).
Chapter 2
22
financial system is relatively weak. This weakness becomes evident in the difficulty
borrowers have in giving and lenders have in enforcing pledges and mortgages
(RIEDEL 2000; UNDP 1999). Thus, the underdeveloped legal framework
does not prove adequate for the use of risk management tools such as collateral
(GOTTWALD and KLUMP 1999). From a banking perspective, the focus of legal reforms
should be on the clarification of legal concepts. These include ownership and
transfer of land use rights, collateral registration procedures, mortgage laws and
title deeds (OH 1999).21 Banks are not allowed to seize land from defaulting farmers.
It is more or less impossible to evict farmers and auction their land (WOLZ 1997).
However, the VBARD still insists on land use rights as collateral, but only few
cases exist where land has been liquidated in the event of a farmer’s collapse
(DUONG and IZUMIDA 2002). The already high share of non-performing loans will
be pushed up further if the assets of defaulters cannot be liquidated. The idea that
almost all loans of the VBARD are in fact unsecured is quite alarming.
The government has recognized private lending as a legal business and has
given private lenders the legal right to sue a borrower in the event of default
(NGUYEN 1998). The practice has shown that private lenders rarely go to the
court for dispute settlement. Particularly in rural areas (and especially at village
level), the Civil Code is not well understood and it is time-consuming and
costly to go through court settlement procedures, particularly in the case of
small loans (HUNG and GIAP 1999). The major goal in improving the legal
framework is to clarify contractual rights and to disseminate this information to
all necessary levels.
A Registry Center has recently been set up under the Ministry of Justice, which
should in principle reduce future disputes among creditors regarding claims on
secured transactions with movable assets. The registration process has the potential
to increase the transparency of lending transactions and of ownership of the
underlying collateral. However, the Registry Center does not yet have an electronic
filing system capable of tracking, recording and providing information on
a timely basis. A Credit Information Center was also established to provide
creditors with information on borrowers. This Center, too, is at an early stage of
21 Foreign banks have recently been allowed to accept land use rights as collateral. This is a
vast improvement over the situation during the 1990s, where regulations explicitly prohibited
foreign banks from accepting land use rights as collateral (KOVSTED et al. 2003).
Transformation of the financial system
23
operation, but it could become a crucial tool for banks to reduce risk in their
future lending. The related rights and duties of the parties involved should be
clearly backed by legislation with adequate sanctions for breaches, thus protecting
the data collected from unauthorized use (WORLD BANK 2002b).
Enactment of a bank secrecy law would very positively influence the use of deposit
accounts (RIEDEL 2000). The current law, however, does little to protect
bank secrecy. First, it is the credit institution and not the customer that has the
right to maintain confidentiality about the customer’s account. Therefore, the
credit institution can give out information about the customer if the institution so
chooses. Second, the credit institution is obliged to give out the information on
the request of a ‘competent state authority’. Since the term ‘competent state
authority’ is not defined, the door is open potentially to pass confidential banking
information to hundreds of state agencies at any level (UNDP 1999).
2.3.4 Regulatory environment
The degree of government involvement in banking regulation is still the subject
of vigorous debate among academics. However, there is broad agreement
that liberalization of financial markets without adequate banking regulation
will most likely lead to macroeconomic instability. Unregulated financial markets
are dangerous, as events in Russia and some Asian countries have indicated
(PILL and PRADHAN 1997; STIGLITZ 2002). Safeguarding financial markets
and institutions from shocks that might pose a systemic risk is the prime
objective of financial regulation. The failure of one non-bank firm often improves
business prospects for the remaining firms in the industry. In contrast, a
shock that seriously damages one bank can spread to other banks (HERRING and
SANTOMERO 2000).
One regulatory measure often recommended by academics to encourage development
of the banking system is the adoption of a deposit insurance scheme
(RIEDEL 2000). In Vietnam, the Deposit Insurance Agency began operating in
2000 (WORLD BANK 2002a). However, the implementation of a deposit insurance
scheme is a matter of controversy among academics. BARTH et al. (2002)
found a strong link between the generosity of the deposit insurance system and
bank sector fragility. This result is consistent with the view that deposit insurance
Chapter 2
24
may not only substantially aggravate moral hazard but can also produce deleterious
effects on bank fragility. The results suggest that the reverse effects from
deposit insurance overwhelm any stabilizing effects that these safety nets may
also have.22 DEMIRGÜC-KUNT and KANE (2001) showed that where effective
bank regulation is lacking (as it is in Vietnam), deposit insurance can do more
harm than good.23 In general, if the government is too willing to help insolvent
banks, this will create the impression that it will continue to do likewise in the
future (SCHRIEDER and HEIDHUES 1998). It is not clear, therefore, that implementation
of a deposit insurance scheme in Vietnam will have beneficial effects
on bank development (KOVSTED et al. 2003).
Vietnam is still a cash economy, with cash accounting for about 50% of
the M3 money supply.24 This may reflect a popular reluctance to use the
banking system. Experience elsewhere has shown that a sound banking system
in which business and ordinary people have confidence is an essential mechanism
for mobilizing domestic savings for productive investment (UNDP 1999;
WOLFF 1999). One of the major tasks ahead is to strengthen popular faith in the
financial system. The very low degree of financial deepening, as well as the
prevalence of US dollar holdings in Vietnam, indicate that the degree of
information opaqueness is well below the intermediate range. The government
therefore needs to promote regulations relating to the dissemination of
financial information. This would include laws relating to the regular issuance,
on the part of publicly traded companies, of financial information on a standardized
basis using internationally accepted accounting and auditing practices
(LEUNG and RIEDEL 2000). Accounting and auditing standards in Vietnam,
which differ from international practice, are another element that undermines
confidence in the banking system (UNDP 1999). Up to now, SBVN banking
supervisors fail to act in line with international standards because they lack
22 Banks do have a reason for taking on more risk than they should. The reason, paradoxically,
is the safety net that governments put in place to prevent bank failures. By trying to
make banks safer, governments give banks the means and motive to behave recklessly
(THE ECONOMIST 2003).
23 The biggest problem with Vietnamese regulations is the lack of enforcement. To effectively
implement the banking laws and regulations, a much stricter and standardized penalty
system must be established (TAM 2000).
24 M3 = Currency in circulation + Short-term deposits + Time deposits.
Transformation of the financial system
25
the power and competence to change bank accountancy standards accordingly.25
Providing greater transparency and reliable information is essential to
strengthen the faith of the population in the financial system.
Rural finance in Vietnam basically means microfinance. So far, there is no special
law regulating microfinance activity in Vietnam. MCGUIRE et al. (1998)
state that a number of prudential banking standards applied to normal banks may
not be appropriate for MFIs. This was confirmed by an SBVN official, who acknowledged
that the existing regulations are too strict and narrow for the microfinance
sector (NHGIA 2001). Governments should ensure that capital requirements
for establishing financial intermediaries are realistic for small institutes
operating at the local level, and that there are no other restrictions affecting the
establishment of such entities (MCGUIRE et al. 1998). Like all types of regulation,
banking regulation can do more harm than good if it is not well designed.
The policy of more restrictive banking regulation impedes the most promising
initiatives in microfinance (SCHMIDT 1999). In general, the question is not so
much whether or not MFIs should be included in formal banking regulation, but
rather when and how (CHRISTEN and ROSENBERG 2000). There is thus room for
improvement and, as mentioned before, particularly as regards reducing market
entry barriers for MFIs to promote competition in the rural financial market.
However, it is very unlikely that any viable financial intermediary will emerge
as long as the government continues to supply highly subsidized loans to the rural
population through the VBSP.
2.4 Conclusions and policy recommendations
After the slow-down in financial sector reforms in recent years, the WORLD BANK
(2001) states that Vietnam’s banking system reform is back on schedule. The
government has adopted a comprehensive banking reform program focusing on
the restructuring of banks and on improvements in the regulatory and supervisory
framework. The past decade has witnessed rapid deepening in the level of
25 In Vietnam, the bank accounting system is determined by the Ministry of Finance. The
process of changing the overall accounting principles to bring them into line with international
standards has been initiated and is currently under way. It is a process that will take
time, however, as it involves formulating a chart of accounts, developing an entire
accounting methodology, and training staff (KOVSTED et al. 2003).
Chapter 2
26
monetization of the Vietnamese economy. The ratio of bank credit to GDP increased
from 13% in 1990 to 27% in 1995, with a further increase to 44% in the
year 2000. The growth of the non-banking financial sector, and especially of the
insurance business, has also been remarkable, even if the size of the sector
remains small in absolute terms (WORLD BANK 2002b). The liberalization of interest
rates is an important step in the continuing transformation of the financial
market in Vietnam. Nevertheless, it is not yet clear within the regulatory environment
whether this step will have positive effects, or whether it was premature
and will thus destabilize the whole financial system.
The government will continue to supply huge sections of the population, mainly
rural, with subsidized credits through the VBSP. While the liberalization of interest
rates offers the potential for financial intermediaries to offer cost-covering
services, it is unlikely that, except for some local non-governmental organizations,
any viable services will be provided in the rural financial market that will
serve a market not reached by the VBSP.
Another area of progress is the recognition of rural savers, including women and
the poor, by the government of Vietnam. The VPSC was funded to develop the
potential of the rural deposit market. However, the network of the VPSC is still
small in comparison to the VBARD. Moreover, the VBARD has never reached
deep enough into the country to attract rural savings. Furthermore, the VBSP
does not demonstrate any intention to offer savings to its customers. Despite
this, the founding of the VPSC may be the beginning of a paradigm change,
namely acceptance of the demand of the rural population for deposit facilities.
The huge quantity of non-performing loans in SOCBs mostly owned by SOEs is
not new and they threaten to destabilize the banking system. However, there are
strong indications that this problem is spreading to the private sector, and particularly
to rural households. This behavior of moral hazard is leaving behind
burnt soil for the establishment of any viable rural financial intermediation.
Freed from political lending, the VBARD is likely to throw overboard its peasant
clientele and concentrate on more lucrative business with bigger and wealthier
farmers. The VBSP will offer subsidized loans on a broad scale and will obviously
be a drain on public resources. The question is how long the Vietnamese Government
can finance the VBSP, and who will serve the rural poor the VBSP collapse?
Finally, the transformation of the financial institutions has clearly proceeded
more quickly than that of its decision makers or employees. Often, actors within
the financial system have not fully exploited the competence margin assigned to
them by the government and continue to wait for additional approval from the
Transformation of the financial system
27
government. Further training of decision makers within the financial system thus
appears to be a key issue. Most desirable would be a reforming lobby within the
Vietnamese financial system. However, this would require more than just training
officials.

3 Information and targeting policies and their principalagent
relationships – The case of the Vietnam Bank for
the Poor26
In the following section, the targeting policies of the VBP are discussed in the
light of the principles of the New Institutional Economics. In this context, special
attention is given to TCs and agency relationships.
3.1 Introduction
A broad access to appropriate and sustainable financial services has been pointed
out repeatedly to be important for poverty reduction. It contributes to higher income
and better food security (ADB 2000; BUCHENRIEDER and THEESFELD 2000;
SCHRIEDER 1996; ZELLER et al. 1997). Many poor households are faced with transitory
food insecurity, even though their incomes seem to provide a sufficient
livelihood base over several years. Thus, there is a potential demand for savings,
credit, and insurance services to stabilize consumption more effectively and to
enhance the ability to escape chronic poverty (KANBUR and SQUIRE 2001; ZELLER
1999).
Hunger eradication and poverty alleviation, particular in the northern uplands, is
of enormous concern to the Vietnamese government (AHMED and GOLETTI 1997;
VBP 1999b). The Ministry of Agriculture and Rural Development (MARD) proposes
financial services as a powerful tool for poverty reduction (SIDA-MARD
1998). From the early 1990s onwards, the Vietnamese government has begun to
establish and promote FFIs such as the VBARD, the VBP and the PCFs to
26 This section is based on the following article: "Information and targeting policies and their
principal-agent relationships – The Case of the Vietnam Bank for the Poor", written by
Thomas Dufhues, Pham Thi My Dung, Ha Thi Hanh, and Gertrud Buchenrieder and published
2002 in the Quarterly Journal of International Agriculture 41 (4): 335-362. I would
like to thank Prof. Dr. Hans Dieter Seibel and Insa Theesfeld for their helpful comments on
an earlier version of the article.
Chapter 3
30
provide the rural poor with loans.27 In addition, the State Treasury has implemented
special sector credit programs, e.g. the 120-program for promoting employment,
which is headed by the Ministry of Labor, Invalids and Social Affairs
(MOLISA). This loan policy is based on the assumption that (1) the rural population
is too poor to repay credits at market interest rates and that (2) capital is a
scarce resource for increasing production and thus raising food security.
The Vietnam Living Standard Survey 1997-98 shows that FFIs have been very
successful in increasing their level of outreach; the state-owned VBARD and
the VBP, for example, have extended credit facilities to almost 4.5 million and
2.5 million rural households respectively (BAC 2001; HANH 2001). These figures
represent more than 58% of all rural households in Vietnam. The majority of
the loans to the rural poor are provided by the VBP through a group lending
scheme, whereby the highest lending activity of the bank is found in the northern
highlands and the lowest is observed in the southeast and central highlands
(VBP 1999b).
The state-owned VBARD and VBP collaborate with the local governments at
commune level, particularly the People’s Committees, as well as mass organizations
such as the Women’s Union and the Farmers’ Union. This operational arrangement
has proven quite successful in expanding the availability of credit to
subsistence farm households (VBARD and DANIDA 1999). Nevertheless, there
exist rural financial market failures because credit is made available not to those
rural households with the most profitable investments, but to those that are best
informed.
27 In 1988, the VBARD emerged from the mono-bank system. At that time, it was called
‘Agricultural Development Bank of Vietnam’. Its name was changed in 1990 into Vietnam
Bank for Agriculture (SEIBEL 1992). The VBP was only established in 1995 as its lending
outlet for the poor of the Vietnam Bank for Agriculture. The PCFs emerged from the cooperative
banking system that was originally part of Vietnam’s mono-banking system.
While most of the original cooperative banks went bankrupt after the introduction of a twotier
banking system, 80 survived. 78 out of them are today part of the more than 800 PCFs,
which are legally based on the decree No. 390/Tg of 1993 (SEIBEL 1997).
Information and targeting policies of the VBP
31
3.1.1 Problem statement
The VBP focuses on the poorer segment of the rural households. It faces particularly
high TCs because of the remoteness of the majority of its borrowers, the small
credit denominations, and the extensive administrative procedure in the credit
allocation process. Some of the operational steps within the screening and monitoring
process are managed free of charge (e.g. by village leaders), but most are
carried out by paid professionals (e.g. by credit officers). Currently, the average
operational costs of the VBP, expressed as interest rate spread, are 0.45% per
month (HANH 2001). The monthly interest rates charged by the VBP vary
between 0.6% and 0.7%. This means that between 64% and 75% of the interest
revenues are used to cover operational costs.
The allocation process of VBP loans for the rural poor is in many respects topdown.
District, commune and village officials are heavily involved in this process.
Apart from this, the commune and village authorities are involved in dissemination
of credit information, screening, and monitoring of rural households and
especially in exerting pressure on households with overdue loans. This ought to
reduce information asymmetry (VBARD and DANIDA 1999), but this approach
has created typical principle-agent problems at the commune and village level.
Empirical evidence suggest that (1) the heads of the communes do not necessarily
sufficiently inform all villages in their communes about the availability of credit,
and that (2) village officials do not always make certain that all households in
their village get the news. It appears that these two administrative steps contribute
to resource misallocation, in the sense that potential borrowers with profitable investments
are excluded due to a lack of information. Information deficits therefore
undermine the access of the poor to credit (KANBUR and SQUIRE 2001).
Furthermore, interviews with key persons have shown that (3) local authorities
are also responsible for the poverty assessment of households and might assess
HHs of relatives, friends, or neighbors as rather poorer than in reality in order to
give them easier access to VBP loans.
Based on these problems, one could claim that access to more timely and complete
information could contribute not only to a more efficient allocation of funds, but
also to an allocation that more effectively benefits the true target group, which is
the rural poor. Information availability in the credit allocation process is therefore
crucial for the empowerment of the rural poor and consequently for increasing
their welfare.
Chapter 3
32
3.1.2 Objectives and methodology
Besides the effect of information deficiencies, the effect of information asymmetry
on the national poverty alleviation policy is evaluated. Key objectives of
this contribution are (1) to identify crucial gatekeepers of important credit information
and (2) to describe the formalized and non-formalized information
channels and propose policy measures to improve their outreach to the target
group.
Methodologically, this contribution combines secondary data from the FFIs and
empirical evidence from the commune and household levels. The qualitative and
quantitative data were collected between March and October 2001 in the Ba Be
district of Bac Kan province in Northern Vietnam. The analysis of information
flows of FFIs at district and provincial level produced insights into policy objectives
and formalized procedures of credit fund allocation. The survey comprised
118 households in four villages and three communes of Ba Be district. By means
of ‘social mapping’28, the relationships of members in credit groups was made
transparent. Semi-structured interviews with key persons such as officials of
mass organization, credit officers, credit group leaders (CGL) and members at
village and commune level provided data on credit allocation.
3.2 New Institutional Economics and the analysis of information flows
in Vietnam’s rural credit allocation29
The term ‘New Institutional Economics’ (NIE) delineates a branch of economic
theory which focuses on institutions and organizations while maintaining the
basic principles of neo-classical economics such as the utility maximization assumption,
the marginal benefit principle, the efficiency assumption, the equilibrium
thinking (SCHMIDT and TERBERGER 1996), and the rational choice model,
or the homo economicus with bounded rationality (RICHTER 1990). Nevertheless,
the efficiency direction of NIE modifies the protective belt of neoclassical
economics by including positive TCs and the constraints of property
28 Villagers drew a map of their village, including all households. Then the households of the
CGL and their relatives, credit group members, and all key persons and their relatives were
marked.
29 The following section draws on (BUCHENRIEDER 2002).
Information and targeting policies of the VBP
33
rights (PRs) (EGGERTSSON 1990).30 The NIE defines an ‘institution’ as a commonly
accepted set of formal rules and informal constraints that shape human
interaction. Institutions also consist of the enforcement characteristics of the
rules and constraints (NORTH 1989).
If a situation does have TCs or information problems, then it does matter how
wealth and PRs are distributed. Modern economic theory emphasizes that TCs
depend on institutions, that institutions are endogenous, and that the distribution
of wealth affects economic efficiency both directly and through its effect on
institutions. Distribution of wealth, institutions, and efficiency are inseparable.
TCs caused by information asymmetry and risk are therefore of particular importance
in developing countries (HOFF and STIGLITZ 2001).
3.2.1 Transaction costs
TCs decrease the efficiency of exchange relationships (GRINDLE 2001). The
regulatory framework of institutions should reduce TCs and raise the efficiency
of exchange. Although a standard definition has yet to be agreed upon,
EGERTSSON (1990: 14) defines TCs as: "…the costs that arise when individuals
exchange ownership rights to economic assets and enforce their exclusive
rights." TCs include all expenses and opportunity costs, fixed and variable,
which arise in the exchange of PRs, except the price of the PR itself. The costliness
of information is the key to the costs of a transaction. Thus, reducing the
cost of information means reducing TC (STIGLITZ 1986). In New Development
Economics, TCs are considered to be of substantial interest for development
finance (FIEBIG et al. 1999).
One way to classify TCs is: (i) prior to the transaction, information costs related
mainly to searching for and screening potential trading partners and obtaining
price information; (ii) during the transaction, negotiation costs including
costs of arranging the transactions, physically transferring the product or
service, and drawing up contracts; and (iii) after the transaction, enforcement
30 Three directions can be delineated within NIE, all of which contribute to the understanding
of institutional change: (1) the efficiency direction that concentrates on minimizing TCs,
analyzes path-dependent institutional change etc., (2) the public choice direction, and
(3) the distribution direction that analyzes the impact of access to power and resources on
institutional change (SCHLÜTER 2001).
Chapter 3
34
costs related to monitoring the terms of the transaction and enforcing liability
(RANDOLPH and NDUNG'U 2000).31
As HOFF and STIGLITZ (2001) pointed out, the structure and the volume of TCs
depend on the institution and on the institutional environment respectively. For
instance, recent research in Thailand has shown that the type of lending policy
strongly influences the level of TCs incurred by borrowers (ERHARDT 2000).
In the case of lending, the TCs of the borrower and lender include all explicit
and implicit expenses that occur in the process of disbursing and obtaining a
loan, respectively. These costs, in case of the borrower, are associated with
transportation, paperwork, lodging and meals, gifts and the opportunity cost of
time, for example, and in the case of the lender, with wages, depreciation of
equipment and loan arrears (HEIDHUES et al. 1997). The remoter the area and the
more marginalized a group of people is, the higher the TCs usually are.
3.2.2 Principal-agent concept
The theory of agency relationships is a branch of the economics of TCs. Information
is not always readily available or easily traded. Information may be
communicated from one person to another, but it may be difficult to verify
whether it is strictly true. This provides the basis for ‘opportunism’, meaning the
incomplete or distorted disclosure of information in order to hide the true facts
(WILLIAMSON 1985). In agency relationships, uncertainty and asymmetric information
are the consequence (RICHTER and FURUBOTN 1996). Decision making
in such circumstances is represented by the agency relationship in which two
parties embark on a mutually beneficial hierarchical relationship, one being better
informed than the other. The uninformed party is called the ‘principal’ while the
other, who has all the information, is the ‘agent’. The identification of which
party to a transaction is the principal and which is the agent depends on the
circumstances. Each individual in a hierarchical structure, except at the ultimate
31 In this paper credit is defined as a search good/service. On the one hand, credit can be seen as
a search good/service from the viewpoint of the borrower. On the other hand, from the viewpoint
of the lender, credit is a mixture of a search and an experience good/service. Search
goods/services can be inspected to make a quality assessment before purchase, whereas the
quality of experience goods/services is learned after a purchase (BIJL 1995). One of the first to
make this distinction between experience and search goods/services was (NELSON 1970).
Information and targeting policies of the VBP
35
level, is simultaneously a principal and an agent when rights are transferred
down the organizational ladder (EGGERTSSON 1990).
In an agency relationship, the principal delegates PRs to an agent, who is bound
by formal contracts and informal conventions to act in the principal’s interest in
return for benefits of some kind. Thus, agency problems are characterized by
the fact that an agent’s decision not only influences his own utility, but also that
of the principal (JENSEN and MECKLING 1976). Due to the conflicting interests
between principals and agents, and to the asymmetrical information, behavioral
risks may emerge. Usually, the principal does not observe the agent’s actions
but rather the result of his actions. Agents may be evaluated by measuring their
behavior and attributes in a costly process applying proxies such as university
degree and number of hours worked. If the proxies are limited too much,
asymmetric information is the result, and this may give rise to opportunistic
behavior on the part of the agent or, in other words, agents are likely to make
sub-optimal decisions from a principal’s viewpoint unless they are effectively
constrained. Market competition can, in various ways, reduce the agency
costs to principals and raise the costs to agents of opportunistic behavior
(EGGERTSSON 1990).
JENSEN and MECKLING (1976) differentiate between categories of agency costs:
(1) monitoring costs (costs which arise for the principal to achieve more transparency
with regard to the activities of the agent), (2) guarantee costs (costs
which arise for the agent to show his reliability to the principal), and (3) residual
costs (costs that result from a lower than agreed performance of the agent). As
pointed out earlier, the actions of an individual are not easily observable. For
instance, a bank entrusts resources to a borrower but cannot perfectly monitor
his investments and initiative; a landlord entrusts land to a tenant but cannot easily
monitor his effort and care. The task of the principal is to design an incentive
scheme to try to align the agent’s incentives with his own. The principal-agent
literature focuses on the design of contracts to motivate the agent to act in the
principal’s interest. Contract provisions that can achieve this are collateral,
bonds, and provisions that shift the risk of poor output onto the agent. The
greater the agent’s ability to post collateral, put up a bond, pay rent in advance,
or absorb risk, the greater the agent’s incentives to take appropriate actions
(HOFF and STIGLITZ 2001). The problems of delegation and commitment are
thus central in a principal-agent framework (GRINDLE 2001).
In a typical development finance approach, credit officers (agents) possess
vastly more detailed and accurate information about the local environment and
the clients than does central management (principal). In the presence of such
Chapter 3
36
‘information asymmetries’ and high monitoring costs, managers are well advised
to align the objectives of the organization (which usually include a mix of
outreach and profitability indicators) with those of the agents (credit officers,
head of communes) who actually make the vast majority of operational decisions.
But these contracts need to be handled with care, as badly designed contracts
can have many unwanted side effects. In fact, it is better not to set incentives at
all than to set badly adjusted ones (HOLTMANN 2001).
3.3 Targeting policies of the VBP in Ba Be
First this section describes the poverty criteria and eligibility of rural households
for VBP loans, which will describe which types of borrowers can access the
loans of the VBP in the research area. Then we discuss the aspects of targeting
polices of the VBP to the rural poor in the research area.
3.3.1 Target group definition and poverty criteria
The majority of poverty-focused development finance programs incorporate
client selection criteria that attempt to limit participation to poorer households.
However, despite attempts to simplify and objectify these procedures, most
programs fail to maintain a focus on the poor and have very little outreach to
the extremely poor. Targeting failure occurs when non-target households are
included and when target households are excluded. The latter is of great concern
to those concerned with poverty alleviation. Exclusion of target households
may result from simple failure to identify them but very often is the result
of deliberate efforts by program officials close to the target group to bar
poor households, which are perceived as credit risks (HICKSON 2001). While
this behavior reduces the TCs of the development finance program, it also reduces
quality outreach. Another aspect that needs consideration in poverty
lending is whether it is necessary to measure poverty or whether it is sufficient
to rely on the self-selection mechanism of appropriately designed povertyoriented
financial services. Such financial services are to be designed in such a
way that the better off would not care to access them (SEIBEL 1996). Relatively
small loan volumes and group lending schemes go in this direction. While
these characteristics apply to the VBP’s lending conditions to the poor, it
explicitly monitors the poverty level of its borrowers and tries to exclude the better
off and the hungry households, clearly this contributes to the operation
costs of VBP.
Under the VBP, poverty is determined, in accordance with the MOLISA criterion,
as the inability to provide for basic nutritional needs. The mandate of
Information and targeting policies of the VBP
37
lending was formerly characterized as the following, first to the starving, then to
the hungry (VBARD and DANIDA 1999). The principle that the poorest of the
poor have priority in receiving credit has been attenuated. The households classified
in Vietnam as hungry, which should have priority with the VBP, are
nowadays generally excluded. The director of the VBP has stated that other
national programs are supposed to take care of them (BAO 2001).32 Obviously, a
change of paradigm has taken place. Nevertheless, MOLISA has recently
broadened the poverty criteria. The poverty line for rural people in the uplands
is now less then 80,000 VND per person per month (VIR 2001).33 This change
implies that households that were formerly considered as medium-income
households are now assessed as poor. For the VBP this has created a difficult
situation, as the bank does not have sufficient funds to satisfy the credit entitlements
of the expanded group of potential borrowers (THU 2001).
The Poverty Reduction Board (PRB) is responsible for the wealth ranking of
households. It exists in each commune. The members are usually the head of
the commune as the most senior official (in most cases he is also the chairman
of the board) and the heads of the respective mass organization at commune
level.34 The village heads are so called ‘temporary consultant’ members of the
board. The wealth ranking of the households is carried out together with the
village heads according to nationally defined poverty criteria (Table 3-1 represents
the wealth ranking in the survey communes referred to in this paper).
This ranking is done once per year. Many households try to be ranked as poor
to improve their chance of accessing a VBP loan. The rather large number of
people responsible for the ranking ought to reduce the danger of nepotism, a
typical principal-agent problem.
32 It is assumed that hungry households are not capable of using a loan productively or of
repaying the principal and interest. Two very important aspects are not considered in this
way of thinking: First, a consumption loan which is used for food can also be seen as an
investment, as it is invested in the most important production factor of the poor, namely labor.
Second, with this approach all hungry households would have to be assessed as not
creditworthy. While this is probably true for many hungry households, it is not true for all.
33 Before, the rural poverty line was 55,000 VND. One USD is equal to 15,709 VND
(Sunday, October 14, 2001, <http://www.oanda.com/convert/classic>.
34 The most important mass organizations in rural areas are the Farmers’ Union and the Women’s Union.
Chapter 3
38
For each household in the commune, the members of the PRB roughly estimate
income according to the given standards (see Table 3-1). The ranking is done by
applying participatory instruments (ANH 1997). Factors other than income play a
much greater role in determining the socio-economic standing of a household,
such as area of paddy rice land, education, food security, off-farm activities
(GEPPERT and DUFHUES 2003).
The VBP has delegated responsibility to the heads of the communes (and PRB)
to verify that only households which are classified as poor according to the national
poverty criteria benefit from the VBP credit line. For his duty in the
screening process, the head of commune apparently receives an incentive, which
is based on the credit volume and the repayment rate in the commune.35 He is
also heavily involved in the reinforcement of the repayment overdue loans. This
is of course an unpleasant additional duty, which the commune head tends to try
to avoid. The information asymmetry between the principal (VBP) and agent
(head of commune) combined with the adverse incentive causes opportunistic
behavior on the agent’s side.
He therefore has an interest in excluding those households that he assumes to
have a low repayment capacity. Clearly, this has a positive effect for the bank
because the repayment rate rises, and for the household as well, as it reduces the
negative effects of being indebted beyond its repayment capacity. At the same
time, he will more likely select households that may not be poor according to the
national criteria and have a higher credit demand. As the head of the commune
has a very powerful position within the local administration, he also has the
power to influence the poverty assessment of households. Hence, a better-off
household might be assessed as poor due to his influence and thus be eligible to
receive a loan from the VBP instead of a more expensive one from the VBARD.36
35 According to BAO (2001), the head of the commune receives 80,000 VND per month for
his collaboration, if: (1) the lending volume exceeds 100 million VND, (2) more than 50
households receive a loan, (3) the repayment rate is over 80%, and (4) the volume of bad
debts is less than 4% of total credit volume. If one of these conditions is not fulfilled, he
receives only 50,000 VND, thus the incentive is 30,000 VND. The monthly salary of the
head of the commune is 360,000 VND. The general rule is that an incentive of less than
20% of total remuneration does not create significant stimulus to improve performance
(HOLTMANN 2001).
36 The giant part of the loans of VBARD/VBP is used for few purposes, e.g. raising buffaloes
Information and targeting policies of the VBP
39
The national policy objective, namely that only truly poor households should
gain access to loans from the VBP, is thus attenuated.
Table 3-1: Income categories of households in VND/month and person
Commune Dia Linh Nghien Loan Xuan La
Hungry HH < 55,000 < 60,000 < 60,000
Poor HH 55,000 – 79,000 60,000 – 79,000 60,000 – 79,000
Medium HH 80,000 – 169,000 80,000 – 99,000 80,000 – 99,000
Better-off HH 170,000 – 219,000 100,000 – 119,000 100,000 – 119,000
Rich HH ≥ 220,000 ≥ 120,000 ≥ 120,000
Source: Administration book DIA LINH (2001); STATISTICS BOOK NGHIEN LOAN (2001); and
STATISTICS BOOK XUAN LA (2001).
Notes: The classification of hungry and poor households is defined by MOLISA. The communes
can vary the boundaries only slightly depending on the local situation. The band for variation
for the definition of the boundaries for medium, better off, and rich households is more flexible
(DUNG 2001).
3.3.2 Effective targeting of the VBP loans in the research communes
The outreach of the VBP differs greatly from commune to commune (see Table 3-2).
In the Xuan La commune, 65% of the households received a credit. In the other
two research communes, only about 20% of the households accessed a loan
from the VBP. At first sight this is surprising because the Xuan La commune is
the remotest of the three. But many of the credit groups in the Xuan La commune
are inter-village groups. The creation of inter-village groups has the advantage
that farmers from villages that are not able to build their own group can
take part in groups of other villages. This widens outreach, but it may reduce
social coherence among the members and thus induce moral hazard among
group members.
or pigs or buying land. The profitability of these investments (e.g. pig raising) varies only
little between villages in one commune.
Chapter 3
40
Table 3-2: Wealth status and VBP loans of HH in research communes
Dia Linh
(N = 602*)
Nghien Loan
(N = 848)
Xuan La
(N = 426**)
N % N % N %
HH with VBP credit 123 20 152 18 275 65
Hungry HH 54 9 111 13 75 18
Poor HH 73 12 300 35 118 28
Medium HH 184 31 397 47 112 26
Better-off HH 205 34 30 4 66 15
Rich HH 84 14 10 1 32 8
Source: ADMINISTRATION BOOK DIA LINH (2001); CREDIT BOOK DIA LINH (2001); STATISTICS BOOK
NGHIEN LOAN (2001); and STATISTICS BOOK XUAN LA (2001).
Notes: * Two HH are not included in the wealth ranking. ** 23 HH are not included in the wealth
ranking. Therefore, the sum of the HH in percent does not equal 100%.
N = number of households in one commune; HH = household.
Whether or not credit groups are created in a village depends heavily on the ability of
enough persons who are able and willing to be a CGL.37 In two of the four research
villages, credit group creation failed because no suitable CGLs could be found. Clearly
there is an unsatisfied demand for CGLs. The CGL must fulfill the following features:
- S/he must be mobile, as the person has to travel to the district branch of
the VBP and to the commune during the application procedure of the
group members. In many groups the CGL also has to collect the interest
payments and often s/he must carry the money to the district branch.
- S/he must have a minimum level of education, as the CGL has to negotiate
with local authorities and bank staff. S/he also has to support the other
members in completing the forms and in planning the investment.
- S/he needs to have ‘good’ contacts with the local authority network to obtain
all the necessary credit information at the right time and to have the
application approved by them.
37 Every credit group must have a CGL. The CGL is usually the one who builds the group. S/he
gets a one-day training course carried out either by the head of the district VBP or by the
credit officer. S/he has the obligation to support the group members in all credit matters. As
recompense for this task, s/he receives the difference between the interest rate charged to the
credit group members (0.7%/month) and the interest rate to be paid to the VBP (0.6%/month).
Information and targeting policies of the VBP
41
- S/he needs a certain level of wealth, because in the event of delayed repayment
of interest by members, the CGL normally advances the money.
The absences of persons with the required abilities can lead to the creation of
inter-village credit groups. Less remote villages, usually inhabited by the Tay38
minority, normally have also a greater number of potentially suitable CGLs and
thus, often have a higher number of credit groups than more remote villages.
Friends or relatives from other villages may be allowed to participate in these
credit groups. Although no empirical evidence was found, this arrangement
leaves scope for bribery and thus may increase the TCs for the credit group
members. As pointed out by SCHLÜTER (2001), human capacity and wealth distribution
also play a significant role in institutional development, and not only
efficiency considerations.
Table 3-3: Percentage of HH per village with VBP credits?
Dia Linh
N = 9 villages)
Nghien Loan
(N = 12 villages)
Xuan La
(N = 10 villages)
Name of village % Name of village % Name of village %
Coc Pai (13 HH) 46 Phia Deng (111 HH) 86 Thom Meo (85 HH) 80
Ban Vang 1 (73 HH) 38 Ban Dinh (86 HH) 22 Khoui Khi (40 HH) 37
Na Duc 2 (86 HH) 36 Khuoi Tuon (29 HH) 21 Nam Nha n/a
Pac Nghe 1 (76 HH) 34 Khuoi Thao (40 HH) 13 Khuoi Boc n/a
Na Duc 1 (101 HH) 16 Khuoi Un (70 HH) 11 Co Luong n/a
Ban Vang 2 (75 HH) 13 Pac Gia (144 HH) 7 Luong Muong n/a
Pac Nghe 2 (50 HH) 12 Ban Na (129 HH) 6
Source: Own calculations.
Notes: In Dia Linh commune two villages (22%), in Nghien Loan commune five villages (42%),
and in Xuan La commune four villages (40%) have no credit group. In the Nghien Loan
commune, inter-village credit groups are also found. Therefore, the percentage share does
not necessarily represent the actual number of HHs with a VBP credit in one village.
n/a = not available, HH = household; N = number of villages in the commune.
Table 3-3 shows which villages in three research communes have access to VBP
loans. Is also illustrates the percentage of households per village with loans. One
might assume that the villages with a high percentage of VBP loans are villages
38 The Tay minority represents the majority of the population in this area. Tay villages are
usually less remote and located in valleys. They can usually speak Vietnamese and all important
local authorities are part of this ethnic group.
Chapter 3
42
with a high number of target group households, namely the poor. Primary data
do not confirm this statement. For example, in the village Thom Meo, 33% of
the households fall below the poverty line, but 80% of all households received a
VBP loan (gap: 47%). The gap is even wider, namely 61%, since hungry households
are considered to be too poor to invest loans efficiently. Clearly, for the
reasons discussed above, non-target group households have gained access to
VBP loans. This is also confirmed by Table 3-2 (see above), as this table shows
that, in the Dia Linh and Xuan La communes, a greater share of households hold
VBP credits than just the share of poor households in the communes. This is
may be because the gatekeeper (credit officer) has family ties in this village.
3.4 Decision-making process of credit allocation for the poor in the case
of VBP
Section 3.4 very briefly describes the credit policy of VBP before discussing the
credit allocation process applied by the VBP to the communes. After this, a brief
explanation is given of how farmers gain access to loans. The section concludes
with a description of the information channels between the target group and VBP.
3.4.1 Credit policy
The VBP was established in 1995 as the poor people’s lending outlet of the
VBARD. The VBARD concentrates on the better-off market segment and provides
individual loans and also offers savings services.39 As regards the VBP, it
is a state-owned bank, specialized in lending to poor households (see Table 3-1
above); savings mobilization plays a very minor role in the service array of the
VBP.40 The purpose of the VBP is not to maximize profit but to reduce poverty
39 Recent credit policy developments, however, show that the credit products of the VBARD
and VBP, and thus also their market segments, are becoming increasingly similar. For instance,
the VBARD now tries to offer a credit product that does not require collateral for
loans below ten million VND (VBARD 2001b). At the same time, the VBP has broadened
its credit term to five years and raised the maximum amount to five million VND
(VBP 2001b). It is not clear to potential customers whether this credit is a VBP product or
a VBARD product offered by the VBP (VBARD 2001b).
40 There is a national consensus, which is adopted by the local staff of the VBARD/VBP, that
poor households are not capable of saving (CAT 2001; CHAN 2001). This is despite the fact
that many non-governmental organizations in Vietnam have proven otherwise.
Information and targeting policies of the VBP
43
(HANH 1999; VBP 1999b). The VBP is allowed to use the operational facilities
and staff of the VBARD and of mass organizations at commune and village
level in extending its services to the target group. From the monthly interest rate
charged to the clients, 0.1% is paid to the local mass organizations and 0.25% to
the VBARD for its services (NGUYEN 1998). Currently, the average operational
costs of the VBP, expressed as interest rate spread, are 0.45% per month
(HANH 2001). The monthly interest rates charged by the VBP vary between
0.6% and 0.7%. As mentioned earlier, this means that between 64% and 75%
of the interest revenues are used to cover operational costs. This implies that
between 25% and 36% of the interest revenues remain to cover refinancing
costs. The repayment performance of 98% is very good (HANH 2001). But this
is only a poor measure of good performance, as re-scheduling of loans in VBP
is extremely high; it is reported to be as high as 70% in some provinces
(VBARD and DANIDA 1999).
The poverty focus of the VBP and the high operational costs do not mean that
the bank consequently operates at a loss. Due to the highly subsidized interest
rates, however, many international agencies consider VBP to be financially unsustainable.
41 Vietnamese policy makers have realized this issue and declare the
subsidized interest rate policy to be a temporary strategy. In the long term, market
rates are to be implemented (VBP 1999b). However, this objective has not
yet been achieved (see Table 3-4).42
Table 3-4: Nominal interest rates per month of the VBP in Ba Be district
Year 1997 1998 1999 2000 2001
Interest rate 1.2 1 0.8 0.6 0.6/0.5
Source: VBP (1997, 1998a, 1999a, 2000, 2001a).
41 The VBP is now recognized as loss-making (VBARD and DANIDA 1999). If the government
does not have sufficient funds to finance the soft loans of the VBP, the Vietnam State
Bank must refinance the VBP without interest on a long-term basis and without a pre-fixed
repayment date (NHGIA 2001).
42 The law on credit institutions approved by the National Assembly in 1997 stipulates that
the state shall establish banks that operate on a non-profit basis (Article 10). This implies
that the state will continue to provide cheap credit to rural areas and the poor (MINOT and
GOLETTI 2001).
Chapter 3
44
The VBP uses joint-liability group lending schemes to reach out to the poor
(VBP 2001b). But in daily practice, the group members are not kept liable for
each other. If one group member defaults, the only consequence for the other
members is that this particular group no longer receives a credit. On the other
hand, individual members who have repaid on time can join a new group.
In March 1999, the maximum loan size per group member had been set at three
million VND (around 190 US$) with a maximum term of three years. In 2001,
the loan size was raised to five million VND and the term to five years. The
interest rate payments are collected every month, or every three or six months
and the principal is usually paid at the end of the term.
3.4.2 Credit allocation process
For its credit allocation estimation, the VBP relies on information from several
public organizations and subordinated branches at the district and provincial level.
First, the PRB compiles a list of all poor households in their commune. This list is
passed on to the MOLISA representative in the district. Second, the district-level
VBP estimates the necessary credit amount for the district and delivers this figure
to the provincial level.43 There, the same procedure is repeated. Based on the figures
of the provincial branches, VBP headquarters in Hanoi determines the nationwide
credit demand. Based on the available funds provided by the government, the
VBP in Hanoi makes a decision on how much of the fund to allocate to each district
(THEESFELD 2000). The district branch of the VBP is informed about the decision.
Thereafter, a body of experts decides how much of the credit fund granted
is allocated to each commune.44 The decision of the body is based on a recommendation
of the head of the VBP at district level (CHAN 2001). The recommendation
for the allocation at commune level depends on the number of poor (hungry)
households and the prior repayment discipline in these communes (BAO 2001).
43 The guidelines of the VBP for the credit estimation procedure propose that the PRB of
each commune compiles a quarterly credit demand plan (VBP 1998b). These plans are
handed to the VBP at district level, where the provincial demand is calculated. However, in
the Ba Be district, the PRBs of the communes do not support the VBP in compiling such a
demand plan (THEESFELD 2000).
44 The body consists of three experts from district level: (1) Vice-head of the People’s Committee,
(2) head of the PRB, and (3) head of VBARD.
Information and targeting policies of the VBP
45
3.4.3 Loan application procedure
Figure 3-1 describes the nine-step procedure to acquire a loan from the VBP.
While the head of the district branch of the VBP in Ba Be states that the credit
officer has to screen and monitor each household during several visits
(BAO 2001), the reality is that the credit officer does not normally visit the
households. In reality, he visits the village and the CGL during the credit application
process to gather secondary information about the applicants. The credit officer
relies totally on the commune and village officials to assess creditworthiness.
After the disbursement of the credit to the group members, the
credit officer may verify with the CGL whether the group members use the credit
as stated in the loan application.45 Normally he trusts in the report of the CGL
(giving him scope for opportunistic behavior) and does not monitor the credit
directly. Basically, it is not bank staff but persons outside the bank that are involved
in loan monitoring: The head of the commune and CGL (see Figure 3-1).
During discussions with the commune heads and the CGL, however, they admitted
that this is not a regular activity. Officially, the loan ought to be repaid immediately
if it is not used as stated in the credit contract. Nevertheless, farmers
pointed out that as long as they pay the interest and the principal in due time, the
credit officer and the other local authorities do not bother about the actual use of
the credit. Credit officers in Ba Be district also confirmed this. The principalagency
relationship and the delegation of tasks have provided wide scope for
moral hazard in terms of the fungible use of the loan.
45 The eligible ways of using the loans are determined by the VBP. Only productive investments
are financed. Loans are usually invested in livestock production. In the survey, 56%
of all loans (including VBARD and others) were invested in livestock. DUONG and
IZUMIDA (2002), in a survey carried out in 1997, found that 54% of the formal loans were
used for livestock, followed by 28% for cultivation. This finding accurately reflects the
fact that investment in livestock has been a popular trend in the Vietnamese rural communities
in recent years and still continues.
Chapter 3
46
Figure 3-1: The nine-step procedure to obtain a loan from the VBP
1. Build a group of at least five people (in rare exceptions four people may be sufficient).
2. Elect a CGL (usually the one who started the group will become the leader).
3. Go (CGL) to the district VBP in order to buy the application forms from the credit
officer who is responsible for the commune.
4. Fill in (group members) the credit application forms.
5. Submit (CGL) the forms to the commune and get the approving signature and stamp of
the head of the commune and the chairman of the PRB (part of the screening process)1.
6. Bring (CGL) the forms back to the bank.2
7. Go (CGL) to the bank and take the loan booklets for each member (the loan booklets
summarize the loan terms and are used by the credit officer to record the interest and
principal payments).
8. Bring (CGL) the loan booklets back to the commune and receive another recognizing
stamp and signature from the head of the commune and the chairman of the PRB
(submits them to the monitoring process).
9. Bring (CGL) the booklets back to the bank and wait for disbursement of the loans
(credit officer disburses the loans to the credit group members during a meeting in
the commune or village).
Source: Own figure.
Notes: 1 The signature also confirms that the applicants belong to the target group. The head of the
commune my decide to exclude an individual household from the credit application if he
knows of a disagreement between married couples relating to the credit application, in cases
of drug abuse by the applicant, or so-called social evil in the households.
2 The approval procedure of the application is as follows: (i) the credit officer signs the credit
application, which is synonymous with his approval, and (ii) the head of the VBP co-signs
the con-tract. Normally, the head of the VBP follows the decision of the credit officer. The
timeframe for this procedure varies between two weeks and nine months in the case of the
research villages.
The fungibility of money has been noticed by VBARD’s headquarter and their
policy on eligible investments has been changed to better reflect the reality. Rural
households are now allowed to spend their loan for any purpose, even for consumption.
Due to the close connections between VBARD and VBP, it is very
likely that the headquarters of VBP will soon adapt its policies, too.
Step 5 of the loan application process (see Figure 3-1) relates to the role of the head
of the commune in this process. If the head of the commune approves the households
in the application, the credit officer will almost always follow his recommendation.
Since the head of the VBP complies with the credit officer’s proposal, this
implies that the head of the commune is actually the decisive power in the application
process. The powerful position of the head of the commune in the VBP credit
application process of households was also pointed out in earlier research on Ba Be
district (THEESFELD 2000). According to VAESSEN (2001), since physical collateral
is not required for VBP loans, the approval of the head of the commune can be
considered social collateral. Local information sources in the form of key informants
can be an important low-cost mechanism to reduce screening and enforcement
Information and targeting policies of the VBP
47
costs. Excessive dependence of a bank on local networks as sources of information
and recommendation, however, can have negative effects in terms of social exclusion
of those households not included in the social and political network in the territory
(KARLAN 2001; VAESSEN 2001). In general, important information tends to be
segmented and to circulate within specific groups or networks (ROBINSON 2001).
Particularly the very poor households often find themselves in this position. Moreover,
they lack access to fruitful relationships with powerful allies (HICKSON 2001).
3.4.4 Backward and forward information flows between VBP, local
authorities and households
Preliminary findings indicate that information on financial services is insufficient in
rural Vietnam. The population is often not aware of the eligibility conditions for loans,
interest rates, and other loan terms. The VIETNAM-CANADA RURAL FINANCE
OUTREACH PROJECT (1999b) states that information about credit is mainly conveyed
via credit officers and is usually inconsistent. In this study it was found that the
credit officers play only a minor role in the dissemination of credit information.
According to VBP staff and local authorities in the research area Ba Be, sufficient
credit funds are available to cover the effective demand of the poor target
group. Therefore, knowledge of how to exploit these funds seems to be more
important than knowledge about their availability. As discussed above, the VBP
lends only to credit groups. Thus, the greater the awareness of the target group
regarding how to set up and manage a credit group, the greater the probability of
obtaining timely loan approval. However, villagers are still reporting that loans
are not always available at the VBP. Hence, timely knowledge regarding availability
of funds is still crucial for access.
3.4.4.1 Formalized information channels
Flows to the household. In one line of information, the head of the VBP informs
the head of the communes and the chairman of the PRB at commune level about
the credit terms, but not about the total amount of available funds allocated to their
commune. Consequently, the head of the commune decides which mass organization,
or rather its members, will benefit from the allocated credit fund. The head of
the commune conveys this information to the village leaders at a commune
meeting, or sometimes directly to the villagers during a village meeting. Figure 3-2
displays the structure of the formalized information flux regarding credit availability
from the VBP to the target group. The head of the village is responsible for inviting
all villagers to this meeting. But not every household receives an invitation.
There may be several reasons for this: (1) the village head has a personal dislike of a
certain household, (2) the household is remote and the head is reluctant to go so far
Chapter 3
48
every time, or (3) he simply forgot to inform a particular household. In one village,
some households even complained that they were never invited. These households
lack access to important information. Reasons (1) and (2) are clearly related to the
principles of NIE as discussed in Section 2. Reason (1) has to do with opportunistic
behavior on the side of a principal vis-à-vis a potential agent. Distance to a potential
agent is closely related to the reduction of TCs argument.
In the other line of information, the chairman of the PRB of the commune informs
the members of the board.46 The members of the Board, who are usually
heads of mass organizations, convey the news to the village-based heads of
their respective mass organizations during regular meetings (see Figure 3-2).47
While the meetings of the local authorities at the commune level are relatively
regular, this does not apply to the meetings at the village level.48 Clearly, the
irregularity of the meetings could cause a loss of information, thus increasing
TCs. However, most farmers turn to the heads of the mass organizations at the
village level when they are planning to apply for a loan. The quality and the
speed of the information flow from commune to household level thus crucially
depend on the leadership and management capabilities of the local authorities
at the commune and village level. This human capacity could be improved
through training. The information flow from the mass media plays only a minor
role in the dissemination process.
It is the credit officer’s responsibility to visit the communes two or three times
per month on a fixed date to collect the principal and the interest of the credit
groups. Ideally, these dates should coincide with the market days in the commune
(BAO 2001). The credit officer should inform the commune head and the
PRB about his next visit. However, there is normally no direct information flow
from the credit officer to poten-tial borrowers since, in reality, the dates of these
visits are either not known by the population or the visits do not follow a time
schedule. Publicity relating to financial ser-vices is not efficient. Communication
is mainly via credit officers and is often inconsistent.
46 Often, the head of the commune and the chairman of the PRB are one and the same person.
47 The role of the mass organizations is to exert permanent pressure on their indebted members,
i.e. the borrowers are regularly reminded to pay their debt.
48 It should also be borne in mind that each mass organization organizes its own meeting and
the Farmers’ and Women’s Union exist in every village.
Information and targeting policies of the VBP
49
The VIETNAM-CANADA RURAL FINANCE OUTREACH PROJECT (1999b) also found
that the farmers did not know the time schedule of the credit officers’ visits.
However, female credit officers tended to do a better job as far as this issue is
concerned. Obviously, the VBP relies on the trickle-down effect of information
from local authorities and CGLs to the households. This trickle-down process is
prone to bottlenecks, or in other words gatekeepers, as described above, and
therefore crucial information may be held up, diverted, or lost.
After having submitted a credit application, the credit group has to wait for the
decision. In the event of the head of the commune declining the application, he
is not obliged to inform the applicants regarding the reasons. THEESFELD (2000),
for instance, reports that the head of a commune uses a coded signature for the
application form. Depending on the code, the credit officer knows whether or
not he agrees to the application. This lack of transparency has two negative effects:
First, there is no scope for improvement of the credit group’s application (TCs
remain high on the side of the credit group). Second, the group members may
become frustrated and hesitate to apply a second time, because they already invested
time and effort without any benefit and without knowing what went
wrong (TCs for a transaction that did not occur on the side of the potential credit
group, and opportunistic behavior on the side of the head of the commune).
Figure 3-2: Formalized information flows
HH HH HH HH HH HH HH HH
Passive information
channels:
- TV
- Radio
- Newspaper
- Leaflets, etc.
Credit officer
Meeting of
village heads
Head of commune
Vice Head of VBARD/Head of VBP
Meeting of mass
organizations at
village
Village Meeting
Chairman of PRB
Meeting of heads of
mass organizations
at village level
Members of PRB
Source: Own figure.
Notes: The arrows indicate the direction of information flow. Broken lines indicate irregular information
flows. Often the head of commune and chairman of the PRB are one and the same person.
Chapter 3
50
Flows to the VBP. The only direct formalized information flow from the
household to the VBP is the submission of the credit application form. The fact
that the bank receives the application form demonstrates that the application
has already successfully passed through two screening levels. The first screening
takes place when the credit group is formed and the second when the application
is signed by the head of the commune. The application forms also contain
a section on the planned investment to evaluate its creditworthiness. The farmers
normally know which loan-financed investments are likely to be approved.
Only a limited range of ‘obvious’ or ‘fashionable’ purposes related to agricultural
production (e.g. rice cultivation, pig raising) will be approved by the bank.
There is not much room for innovative or idiosyncratic proposals (UN 1998;
VIETNAM-CANADA RURAL FINANCE OUTREACH PROJECT 1999b). As discussed
earlier, while the credit purpose is screened, the bank staff does not monitor this
later on.
Besides the information provided in the credit application form, the VBP uses
commune authorities to assess the creditworthiness of potential clients. This approach
of external screening is a reaction to the information asymmetry that exists
between the credit officer and the local population. As the local authorities have
more intimate contact with the local population, they are better able to judge
their creditworthiness. This procedure ought to reduce the information gap between
credit officers and borrowers and thus reduce TCs and improve the repayment
rate (GRINDLE 2001; HOFF and STIGLITZ 2001). However, this approach can
cause principle-agent problems between the credit officer (principal) and the
head of the commune (agent). As was discussed earlier, the head of the commune
has a strong influence on the outcome of the screening process. Good contacts
with the head of the commune can bias his decision in the loan screening process
towards a particular household, even if the latter does not in fact fulfill the eligibility
criteria of the VBP.
While the VBP has already incorporated several measures to reduce information,
negotiation and enforcement costs (RANDOLPH and NDUNG'U 2000) in the
rural credit intermediation process, these measures still leave scope for opportunistic
behavior by the principals involved and moral hazard on the part of the
agents. Better rules for credit officers and more appropriate incentives for nonstaff
resource persons involved in the credit allocation process could reduce TCs
and lead to more equity in the allocation of credit.
Information and targeting policies of the VBP
51
3.4.4.2 Non-formalized information channels
Flows to the household. Local authorities, such as the commune head, that possess
insider information regarding when one of the credit funds is made available
by VBP, function as so called ‘gatekeepers’. Households that have information
regarding when, where and how the loan funds are made available have a major
advantage in exploiting this source by appropriate signaling. Clearly, this information
tends to be given to households that belong to the social and political
network of one of these gatekeepers (BAO 2001).
The village heads confirmed that one could always find households that cannot
translate their demand into effective access to credit. These households may not
be creditworthy and/or are excluded from the intra-village information-networks
(THEESFELD 2000). Nevertheless, the 15 CGLs involved in this research claim
that all villagers are able to participate in their credit groups. They point out that
they informed the whole village about their intention to set up a credit group.
This would imply that the groups are heterogeneous in their membership. Nevertheless,
in thirteen of the 15 credit groups in the research area, the members
were neighbors or relatives of the CGL. As stated by HOFF and STIGLITZ (2001)
social ties are crucial for credit access.
Flows to the VBP. The head of the commune does not know all households in
his commune personally. He makes enquiries of the CGLs, the village heads,
heads of mass organizations, the security chiefs and other villagers regarding the
individual households of a potential credit group. The credit officer interviews
the same key persons. Again, this may cause agency problems. In this case, the
head of the commune and the credit officer are the principals and the village key
persons are the agents. Households that maintain good relations with these key
persons have a better reputation than those who do not. One result of this may be
that some households gain access to two loans from the VBP (e.g. father and
son), which is not allowed. This, of course, only works if the village authorities
tolerate it (BAO 2001).
3.5 Conclusions and policy recommendations
The analysis has shown that the financial information flows of the VBP are
fuzzy (in the sense of a credit application having different outcomes depending
on the intermediaries of information involved), despite extensive formalization
of the channels. In principle, the use of the aforementioned local authorities in
client screening and in the monitoring process is a good strategy for reducing
information asymmetry and thus TCs. Nevertheless, the involvement of these
crucial information gatekeepers at different administrative levels, namely the
Chapter 3
52
credit officer, head of the commune, members of the PRB, the village head, and
the CGL, may impede, divert, and attenuate the dissemination of information.
These gatekeepers normally belong to the same or related social and political
networks. The information collected during group discussions in the research
communes confirms that the key informants of a FFI in a certain region should
not all belong to the same network. VAESSEN (2001) points out that, this way,
the formal lender can avoid falling into the trap of lending solely to a certain
network of people, while the households outside this network are excluded. Key
informants from other networks could be from non-governmental organizations
or self-help groups. To date, however, these groups are not found in the research
area. One attempt to open up new networks might be to employ credit officers
from different ethnic minorities. So far, 95% of the credit officers in Ba Be belong
to the Tay minority. The remaining 5% are Kinh (Vietnamese). Increasing ethnic
diversity of employees could also diversify the networks of information.
However, as stated by HUNG and GIAP (1999), the efforts of formal lenders are
hampered by the difficulty in recruiting staff from ethnic minorities, especially
in remote and mountainous districts.
According to the principal-agent concept, the VBP should design incentives
schemes for its gatekeepers to improve the information flow to marginalized
groups. Another way to make the available information more transparent to
potential borrowers could be the establishment of local information centers. This
could be the communal post office, where the VBP could arrange an information
corner with leaflets. This could be combined with the introduction of a fixed day
every week (or bi-weekly), which all commune members should be made aware
of, when the credit officer is available for questions and discussions. This would
allow farmers to seek information and advice directly from a professional. Officially,
the credit officers already follow such a fixed visiting schedule. However,
they do not stick to it, as discussed above. If the VBP succeeded in asserting this
policy, it would improve the service without increasing the TCs.
The passive information flow from the VBP to the households could be improved
by using the mass media, particularly radio. Today, 95% of the country’s
population and 98% of the national territory have access to the radio station
Voice of Vietnam (VOV). In the past, most of the radio programs were broadcast
in Vietnamese, which excluded many of the ethnic minorities from the information.
As a result of recent efforts, however, the VOV has bolstered its
broadcasting network to include eight ethnic minority languages, e.g. Thai, Mong
(VIET NAM NEWS 2002b).
Information and targeting policies of the VBP
53
As the above analysis has shown, the use of a complicated poverty measurement
system that (1) involves the same local authorities that decide on credit access
for the poor and (2) sets out the framework for eligibility of households to VBP
loans, creates agency problems and is open to nepotism. Instead of using the
current system to measure poverty, one could use the land register as a much
more objective set of criteria for target group identification. Households that
possess documented land-use rights should automatically be excluded from the
target group of the VBP, as they can get loans from the VBARD too. Of course
this should go hand in hand with equalizing the interest rate of both banks,
otherwise it could provide an incentive not to register the land in order to maintain
the opportunity to apply for VBP loans.
By way of summary, improving the top-down and bottom-up information flow
in the credit market for the rural poor might not only significantly reduce the
operating costs of the FFIs concerned, but also increase market access for the
poor target group with profitable investments.

4 Outreach of credit institutes and households' access
constraints to formal credit in Northern Vietnam49
In the following section, the outreach of formal credit institutes and the access of
rural households to them are discussed. A principal component analysis (PCA) is
applied to measure the poverty outreach of the institutes and a binary logit
regression to assess access constraints of rural households to formal credits.
4.1 Introduction
Most policy and research interest regarding rural credit markets revolves around
the perception that poor households in developing countries lack adequate access
to credit, which is believed to have significant negative consequences for various
household-level outcomes. During the past 40 years, most developing countries
and donors have set up credit programs aiming at improving rural households’
access to credit. The vast majority of these programs, especially the so-called
‘agricultural development banks’, have failed both to achieve their objectives to
serve the poor and to be financially sustainable institutions. An important feature
of the rural credit market is that access to credit is far easier for some groups
than for others. Meanwhile, most rural households in developing countries continue
to rely on the informal market for their inter-temporal transfer of resources.
Therefore, outreach of the financial institution and access to financial services
have become a major issue in microfinance, and particularly in rural finance
(DIAGNE et al. 2000; SARAP 1990).50
49 This section is based on the following article: "Outreach of credit institutes and households'
access constraints to formal credit in Northern Vietnam", written by Thomas Dufhues and
Gertrud Buchenrieder and published 2005 in Research in Development Economics and
Policy Discussion Paper, No1/2005 Stuttgart: Grauer Verlag.
50 The performance of financial intermediaries in terms of breadth and depth of poverty outreach
is partly a function of a range of internal factors such as the type of services provided,
the use of screening methods to identify the poor, the financial scope of the program, and the
marketing strategy of the program (ZELLER et al. 2003). Although there are more dimensions
Chapter 4
56
Formal rural credit is considered by Vietnamese government agencies to be a
powerful tool for poverty reduction (SIDA-MARD 1998). During the early
1990s, informal credit accounted for almost 80% of total outstanding loans in
Vietnam (GSO 1995). The Vietnamese government tried to break the dominance
of the informal sector and push development by supplying credit on preferential
terms, particularly to rural households. The preferential credit was delivered
mainly by state-owned financial intermediaries such as the VBARD, the VBP)
and the PCFs.51 In addition, the State Treasury implemented special sector credit
programs, e.g. the 120 Program to promote employment (DUFHUES et al. 2004a).
Initial signs of success of formal credit outreach were reported by the Vietnamese
Living Standard Surveys, which stated that the share of the informal sector had
been considerably reduced from 78% (1992/93) of all outstanding loans to 54%
(1997/98) in favor to the formal sector (GSO 1995; GSO 2000). There is evidence
from other developing countries that credit constraints persist despite the
expansion of microfinance. For instance, AMIN et al. (2003) find that microfinance
institutes (MFIs) in Bangladesh do not really reach those who are creditconstrained.
Recent research suggest that, in Vietnam, the success of pro-poor
policies will depend on easing structural constraints such as access to credit
(GLEWWE et al. 2002; LIVINGSTONE 2000). However, deeper outreach usually
increases not only social value but also social cost. As income and wealth
decrease, it becomes more costly to a lender to judge the risk of a loan. This
happens because, compared with the rich, the poor are more heterogeneous and
less able to signal their ability and willingness to repay (CONNING 1999). Moreover,
the provision of income-generating credit leads to a bias in favor of the
less poor, because they have better opportunities to use the loan profitably
(HULME and MOSLEY 1996). While modern microcredit programs are definitely
more successful at reaching the poor than their predecessors, they are less
than depth and breadth of outreach (for a detailed overview see NAVAJAS et al. 2000; and
SCHREINER 2002), this research will focus mainly on depth and breadth of outreach. Since
society places more weight on the poor than on the rich, poverty is a good proxy for depth
(NAVAJAS et al. 2000). The poorer the clients reached by the financial institute, the deeper
the outreach (ZELLER et al. 2003).
51 On March 11, 2003, the VBP and the PCFs were replaced by the Vietnam Bank for Social
Policies (VBSP) (VIETNAM ECONOMY 2003; WORLD BANK 2003).
Outreach and access
57
successful at reaching the vulnerable poor.52 HULME and MOSLEY (1996) state
that there is increasing evidence that the poorest 20% of the population are effectively
excluded from microcredit programs.
While the Vietnamese government has so far failed to create sustainable rural
financial institutions, it has succeeded in providing a huge share of the population
with formal credit. However, despite the immense formal outreach, the distribution
of formal credit in the northern provinces of Vietnam is very heterogeneous.
In some villages, over 90% of households are served by formal credit,
while in others just a few or none at all (DUFHUES et al. 2002). The question thus
remains: Did the Vietnamese Government succeed in reaching the poor, or do
groups of people still exist who are access-constrained?
The second section describes the conceptual framework, the analytical methods
used and the sample composition. Section two starts with a discussion of collateral
use in Vietnam and then continues with a description of the effective credit demand
of the sample households. The last two chapters of this section discuss the
outreach of rural lenders and access to formal loans by the households in this
sample. The paper concludes with policy recommendations for improving outreach
to access-constrained households.
4.2 Methodology and data
The following section describes the conceptual framework, followed by a description
of the methodology used for the data analysis. Finally, the sample composition
is presented.
4.2.1 Access constraints to formal rural credit – The conceptual
framework
While the term outreach refers to the perspective of the financial intermediary
and access refers to the point of view of the household, they both relate to the
same thing: Who is getting the credit (VAESSEN 2001). Access constraints at the
household level are mostly related to a lack of collateral (physical, human and/or
52 Even with the use of group lending schemes, which are believed to have a good poverty
outreach, the evidence suggests that the poorest people are excluded (MONTGOMERY 1996).
Chapter 4
58
social capital).53 The capital endowment of a household is of enormous importance
for the household’s access to formal credit and to the outreach of rural
lenders. Capital can be classified into three different categories, physical, human
and social capital. The term physical capital refers to any non-human, infrastructural,
financial, or natural asset needed to support livelihoods. Human
capital represents the skills, knowledge, ability to labor and good health of individuals
or households. Social capital is defined here according to COLEMAN
(1999), who states that social capital is not a single entity but a variety of different
entities, with two elements in common: They all consist of some aspect
of social structures, and they facilitate certain actions of individuals who are
within the structure. The characteristics of human and social capital are identified
as ‘capital’ in order to underline the need for continuous investment and to
emphasize the importance of these factors in generating future income, particularly
for the poor.
Figure 4-1 shows the different kinds of capital/collateral. Different kinds of
capital/collateral can substitute each other, but only to a certain degree. Nevertheless,
the capital endowment of a household represents the basis for the collateral
used by the lender. Informal lenders, for instance, in the absence of physical collateral,
have always used human and social collateral. Formal lenders usually
rely on physical collateral that can be easily sold, are not moveable, are of sizeable
value, and ideally carry a legal title, such as land and buildings.54
53 Access constraints can also be intermediary-based, in other words influenced for example
by business policies or staff attitudes, hard and soft skills. A conceptual framework of clientrelated
barriers versus program-related barriers is presented in EVANS et al. (1999) and
VAESSEN (2001). However, this analysis focuses on the household level and rather neglects
intermediary-based issues as they are reflected in the households’ access constraints.
54 Physical collateral has several functions, e.g. signaling credit worthiness. However its two
main functions are: First, it insures the lenders’ loan portfolio in case of default by borrowers.
Second, it represents an incentive, enhancing the borrower’s willingness to repay his loan
(BESTER 1987; STIGLITZ and WEISS 1981).
Outreach and access
59
Figure 4-1: Conceptual framework – The capital-collateral system
Rural
households
Rural financial
intermediaries
Outreach of
Access to
Social
capital
Human
capital
Physical
capital
Social
collateral
Human
collateral
Physical
collateral
Owned by Use of
Source: Own figure.
Physical capital and collateral: Collateral in the form of physical capital plays a
key role in lending practices. It shifts a portion of the potential capital loss from
the lender to the borrower (BINSWANGER and SILLERS 1983). Lending institutions
typically resort to legal options, such as seizing the property of the borrower or
garnishing wages directly from the employer to enforce contracts. Regardless of
the actual value of the asset owned by the borrower, the act of pledging assets and
the consequent realization that they can be lost causes the client to repay the loan
if possible (LEDGERWOOD 1999). Even if the collateral is almost never collected,
this does not signal its lack of importance as an incentive device. If the threat is
believable, there should be few instances when collateral is actually collected
(MORDUCH 1999). Nevertheless, in most poor communities, such punishment
fails for one of the two reasons: Either the legal infrastructure does not support
such action, or the borrower has no sizeable assets or wages (KARLAN 2001).55
This is a particular problem of MFIs, which systematically lend to low-income
clients who usually have very few marketable assets. Traditional collateral such
as property, land/land use certificates, or other capital assets is often not available.
Therefore, the absence of physical capital, and thus collateral, has for a long time
been seen as the major access constraint of lower-income households.
55 For instance, in socialist economies like China, land is collectively owned, preventing its
use as collateral (PARK and REN 2001).
Chapter 4
60
Human capital and collateral: At the household level, human capital is a factor
of the amount and quality of available labor, which usually is defined in terms of
the health and education levels of individuals. Lack of human capital is seen as an
access constraint from the household side, as there is often a need to fill in application
forms or draw up small business plans, and this requires a certain amount
of human capital (see for instance PANJAITAN-DRIOADISURYO and CLOUD 1999).
Human capital can compensate for a lack of physical collateral. Then, the decision
of whether or not to grant the credit is based mainly on the profitability of the
investment. However, this practice is still very uncommon among rural lenders.
Usually staff lacks the appropriate skills to assess an investment reliably.
Social capital and collateral: The poor seldom have physical collateral to offer.
The most common way to deal with this problem is by using social collateral.
Here, the borrowers’ reputation, or the social (and political) networks to which
they belong, replace traditional physical collateral (BASTELAER VAN 2003;
PANJAITAN-DRIOADISURYO and CLOUD 1999). Usually this results in applying
credit group schemes with joint liability.56 Under an individual lending contract,
if the borrower defaults, all he has to fear is the penalties the bank can impose,
which, in the absence of collateral, simply means the denial of future loans. In
group lending, he may also be exposed to the wrath of other group members
(BESLEY and COATE 1995).
Infrastructure: The rural infrastructure influences both access to and outreach
of credit equally, and is therefore not depicted in Figure 4-1. Infrastructure not
only refers to the ‘hard’ infrastructure, like roads, but also to ‘soft’ infrastructure,
such as legal frameworks. For instance, FABBRI and PADULA (2003) found
in a recent paper that lax legal enforcement increases the probability of poor
people being access-constrained to formal credit in Italy.
56 Despite its popularity, group lending-oriented microfinance is not a panacea for solving all
problems of access to financial services for the poor in developing countries. Not only specialized
agricultural development banks are prone to the difficulties due to information
asymmetries and moral hazard pointed out by institution economics; these problems can
also occur in group lending. For a critical view of group lending see, e.g., HEIDHUES et al.
(1997) and SCHMIDT and ZEITINGER (1994).
Outreach and access
61
4.2.2 Measuring outreach and access: Econometric models
Principal Component Analysis:57 The breadth of outreach of a microfinance institution
is easy to measure (one simply counts the clients), but other dimensions of
outreach, particularly poverty outreach, are more difficult to measure (WOLLER et al.
1999). A poverty assessment tool was developed in the late 1990s by the International
Food Policy Research Institute (IFPRI), which uses the PCA as econometric
instrument. The objective was to design a tool to assess the poverty level of the
clients of a microfinance institution in relation to their non-clients (who represent
the general population in its area of operation) to give a reliable assessment of the
poverty outreach of the institution (ZELLER et al. 2001; ZELLER et al. 2005).
The PCA is a multivariate technique and its main objective is to reduce the
dimension of the observations (HÄRDLE and SIMAR 2003). Different correlated
variables are aggregated into fewer uncorrelated principal components, which
can be seen as indices. With this technique, most of the information contained in
the data is represented in the new indices. The analysis can be viewed as a ‘data
reduction technique’, since the set of original m variables is reduced to n principal
components (PC), with n«m. This smaller number of components can then be
used for interpretation purposes or for further data analysis. The procedure carried
out by the analysis is to calculate new uncorrelated principal components by
linear combinations of the original, correlated variables. This is done by deriving
(standardized) weights for each indicator. In algebraic terms this means that
PC1 = w11 v1 + w12 v2 + … + w1m vm
PC2 = w21 v1 + w22 v2 + … + w2m vm
PCm = wm1 v1 + wm2 v2 + … + wmm vm
(4.1)
with w is the calculated weight and v is the variable. Applied to poverty assessment,
the PCA determines a subset of indicators that measure the relative poverty
level of a household. In the end, a single indicator for each household is created
that reflects the household’s poverty status in relation to all other households of
the sample (ZELLER et al. 2005). With the weights of the PC1 and the respective
indicators, the poverty index is calculated for each household. Relative comparisons
can then be drawn by ordering the households according to their poverty
57 The next section is based on a review by HÄUSER et al. (2005).
Chapter 4
62
index. In this way, it is possible to identify which households are better or worse
off than others. By creating terciles, quartiles or quintiles using the index, different
wealth groups can be derived. The most important advantage of the PCA is that
on the one hand it creates a single indicator that is easy to use for analysis, while
at the same time this single indicator is not limited to the monetary aspect
addressed by household expenditures as the conventional method of (income)
poverty. The PCA technique allows to take the multiple dimensions of poverty
into account and to integrate qualitative with quantitative variables. Therefore,
indicators capturing different dimensions are included that measure changing
conditions at different levels of welfare. These indicators can be categorized into
three groups (HENRY et al. 2003):
1 Means to achieve welfare
This category includes indicators that reflect the earning capacities of
a household. They are subdivided into human capital, social capital and
ownership of assets.
2 Basic needs
These indicators show the fulfillment of basic needs, such as health
status, food, shelter and clothing, partly obtained by questions asking
the respondent about his or her self-assessment of the situation.
3 Other aspects of welfare
Security, self-assessment of (subjective) poverty, social status and the
environment are captured in this group.58
Binary logit analysis: This research collects household-level credit market information
to determine whether or not households are constrained as regards
access to formal credit. Whether or not a household is access-constrained is depicted
in the decision tree of effective credit demand (see Figure A-1 in the Annex).
Households that had access to formal credit were considered not to be accessconstrained
regardless of whether or not they originally wanted to borrow more
than they were lent. In particular, non-borrowing households were asked their
reasons for not borrowing or for having been rejected. DIAGNE et al. (2000) and
ZELLER and SHARMA (2000) state that households may have chosen not to borrow
even when they had access to credit, while others may have wanted to borrow, but
58 The adapted list of poverty indicators used in this research work can be found in Figure A-2
the Annex.
Outreach and access
63
had no access. For these reasons, one cannot equate observed demand with access.
Finally, the sample households were classified into households with and
without access to formal credit.
Regression models describe the relationship between a dependent variable and
independent explanatory variables (BACKHAUS et al. 1996). Ordinary least
square regression models consider the dependent variable to be continuous in
nature, while the explanatory variables can be either continuous or categorical.
But it is not uncommon that a dependent variable is binary in nature, i.e., that it
can only have two possible values, one for the occurrence of an event, zero other
wise. In this case, the dependent (binary) variable is one for all households with
access to formal credit and zero otherwise. A mixture of continuous and categorical
variables may explain this dependent binary variable. Therefore, the
econometric model used in this research work is a binary logit regression. In the
binary logit regression model, the predicted probabilities for the dependent variable
will never be less than (or equal to) zero, or greater than (or equal to) one, regardless
of the values of the independent variables.
The explanatory variables for the binary logit model and the hypotheses behind
the choice of the explanatory variables are presented in Table 4-1. Descriptive
statistics of these variables can be found in Table A-1 in the Annex. As the
variables have different units of measurement, the independent variables were
standardized using z-transformation to make them comparable. In accordance
with the theoretical framework, the variables have been sorted according to
their character as indicators for the different forms of capital/collateral (see
Section 4.2.1).
Chapter 4
64
Table 4-1: Variables for the binary logistic regression model on credit access
Physical capital-related variables
1. Red or Green
Books*
(yes = 1; no = 0)
Red Books or Green Books, certifying a person’s use rights to farmland
and forests, are often demanded as collateral by rural lenders in
Vietnam (DUFHUES et al. 2004a). It is assumed that possession of a
Red or Green Book will positively influence access to formal credit.
2. Agricultural land
(m2)
Land is an important form of collateral. However, most property
rights relating to land have an informal character in developing countries.
In the absence of Red Books, Vietnamese farmers can apply to
the local People’s Committees for special land use certificates. Only
8% of the loans were used wholly or partly for buying land. Hence,
the endogeneity of the variable is negligible.
3. Value of houses
(VND)
Houses can be used as formal collateral with a max. value of five million
VND. Thus, this variable is assumed to positively influence access
to credit. Only 4% of the loans were invested complete or partly into
constructing houses. Hence, the endogeneity of the variable is negligible.
4. Government
salary
(yes = 1; no = 0)
Government salaries can serve as collateral (DUFHUES et al. 2004a)
and may therefore improve access to credit. Furthermore, government
work in particular allows individuals to enhance their social
network, helping them to stay informed about economic developments
and new laws or policies (ALTHER et al. 2002; DUFHUES et al. 2002),
which may also facilitate access to credit.
5. Cash savings
(VND)
In theory, savings can be used as collateral, too. However, nearly
100% of the savings in this survey are informal and cannot therefore
easily be seized by a formal lender. Nevertheless, savings are the
basis for accumulating physical capital and are therefore a good indicator
of possession of physical capital. In addition, they are a good
indicator for the repayment capacity of households and are therefore
assumed to have a positive influence on access to credit.
Human capital-related variables
6. School years of
HH head
(years)
Better education is assumed to improve access to credit as loan application
procedures demand a certain degree of formal education.
Moreover, it is reasonable to expect that better educated households
perform better in their investment activities. This was also shown in a
recent paper by VAN DE WALLE (2003), who found that marginal
benefit of irrigation increased strongly with the education of the
household. Thus, better educated households are usually perceived as
more creditworthy.
7. Vietnamese
communication
skills of the
married couple
(yes = 1; no = 0)
Vietnamese is the official language and used for the credit application
procedure. The household head and his wife have to sign the
credit contract. Therefore, it is assumed that households in which one
of the two lacks the necessary language skills are more often accessconstrained
(in the event of only husband or wife existing in the
household, his/her language capability is taken into account).
8. Receiving agricultural
extension
service
(yes = 1; no = 0)
Households who receive agricultural extension services are likely to
gain better access to improved knowledge and are thus able to increase
their human capital. But at the same time, households who do not receive
agricultural extension are likely to have fallen through the village
information network and are assumed to have a low social capital base.
Outreach and access
65
9. Active HH
members
(number)
Household members between 15 and 60 years for male members and
15 and 55 for female persons were counted as one workforce. These
age lines represent the official retirement age in Vietnam. This indicator
evaluates the labor capacity of the household and indicates human
capital in the sense of labor force. Each investment activity demands
additional labor. Households with a lower labor supply are assumed
to have less access to formal credit.
10. Share of nonfarm
activities
in total yearly
income
(%)
A high share of non-farm activity in total income may indicate a shift
from traditional farm activities towards more innovative non-farm
investments. It also includes households with employment at government
agencies, which demands a certain level of education and entrepreneurial
skill. Therefore, it is assumed that households with a
higher share of non-farm income also have a higher level of human
capital. Income from day labor is excluded, as this does not require
any human capital except for the labor itself.
11. Lost working
days due to
illness
(days/year/HH)
The number of lost working days per year within a household due to
illness is a good indicator of the quality of human capital. The larger
the number, the lower the human capital. It is assumed that households
with a weaker human capital base are more often accessconstrained.
Social capital-related variables
12. Giving help
(days/year/HH)
13. Receiving help
(days/year/HH)
Receiving and giving help to friends and relatives is seen as an important
indicator of social activity and of being a member of a social
network, and thus of social capital, and this will positively influence
the probability of having access to credit as households are better
protected against income shocks.
14. Interest-free
informal credit
(VND)
Small interest-free informal loans are not a substitute for formal
loans (Section 3.2). Thus, borrowing from informal sources is usually
not the result of an access constraint to the formal financial market.
Possession of an informal interest-free loan is a good indicator of a
functioning informal social network. However, it is also seen as a
mechanism for coping with sudden shocks and it suggests that the
household may have a lower repaying capacity and a low physical
capital endowment. Thus, the a priori sign of the coefficient is ambiguous.
15. Thai/Tay village
(yes = 1; no = 0)
Considered them from a national point of view, the Thai and Tay
ethnic groups are ethnic minorities. However, in the two research
regions, Ba Be and Yen Chau district, the Tay and Thai are the ethnic
majority. For instance, official positions in Ba Be are usually occupied
by members of Tay ethnicity (DUFHUES et al. 2002). Belonging
to one of these ethnicities is seen as important social capital. Besides,
anecdotal evidence suggests that there exists a constraining hierarchy
among the ethnic groups in Northern Vietnam. For example in Ba Be,
Tay and Kinh are the leading ethnic groups, followed by the Nung, Dao,
or Hmong (ALTHER et al. 2002; CASTELLA et al. 2002). Villages in
the research region are of high ethnic homogeneity (see Table 5-3).
Ethnic Thai or Tay usually dwell in valley positions or at mediumhigh
altitudes with very similar agricultural production systems.
Households not belonging to the regional ethnic majority but dwelling
in a village mainly populated by them are also assumed to profit
Chapter 4
66
from this location. Therefore, household observations within a
Tay/Thai village are not necessarily independent. However, significant
differences can be expected between Tay/Thai villages and villages
of other ethnic groups that are not captured by our data (e.g.
different production systems). Therefore, households are grouped by
Tay/Thai village to account for those differences. It is assumed that
households dwelling in a Thai or Tay village are more privileged and
therefore have better access to credit. Village fixed effects which are
related to natural conditions (e.g. climate) are not viewed as relevant
for access to credit in Northern Vietnam. An attempt is made to control
for socio-economic village fixed effects, e.g. market access, by
different variables, such as number of market visits.
16. Only female HH
members going
to the market
(yes = 1; no = 0)
Who goes to the market is regarded as important factor for gaining
access to certain social networks and for gathering essential credit
information. Particularly when only women go to the market, the
household as a whole may be excluded from important information.
But this might also be an indicator of a weak human capital base, as
the male household head could be dead, have left the family, or be
physically or mentally incapable of going to the market. Hence, it is
assumed that when only female household members go to the market,
the chance of being access-constrained is higher.
17. Age of the
household
(years)
This variable can go in two directions. On the one hand, it may be
that the older the household, the wider and stronger the social network
of the household members and thus, the greater its political
influence and its ability to gather credit-relevant information. On the
other hand, however, younger households may be better educated or
more dynamic. Thus, the a priori sign of the coefficient is ambiguous.
Infrastructure-related variables
18. Remoteness
(km)
This variable measures the distance in kilometers of rural households
from the district center where formal bank branches are located.
HUNG and GIAP (1999) state that the distance to the nearest bank
branch strongly influences access.
19. Market visits
per month
(numbers)
The frequency of market visits by household members is assumed to
be an indicator of high social activity. Frequent market visits may
also increase the chance of receiving essential credit information, e.g.
on the availability of loans (DUFHUES et al. 2002). The frequency of
market visits of course also depends on the infrastructure connection
and the remoteness of the household’s dwelling.
20. Different markets
visited
(numbers)
The number of different markets visited is an indicator of the breadth
of a household’s information networks. It is assumed that the broader
the network, the more relevant information is available to the household
for receiving a credit. The number of markets visited also depends
on the infrastructure connection and the remoteness of the
household’s dwelling.
Poverty-related variables**
21. Poverty index Anecdotal evidence suggests that even credits that are targeted at the
poor seem to be bypassing the poorest groups (NEEFJES 2001;
WORLD BANK and DFID 1999). Thus, poorer households are assumed
to have less access to formal loans.
Outreach and access
67
22. Supply of day
labor
(yes = 1; no = 0)
Poor agricultural laborers and other day-laborer (for instance in road
construction) are assumed to be more often access-constrained due to
their low social standing (FALLAVIER 1994). Nevertheless, this may be
also an indicator of lack of agricultural land and thus of physical capital.
23. Receiving aid
from government
(yes = 1; no = 0)
Only very poor or very vulnerable households receive food/equipment
aid from the government. It is thus assumed that these households have
less access to formal credit. However, anecdotal evidence suggests that
it is not always the people with the greatest need who receive the help,
but those who also have good contacts with village authorities, pointing
to a certain degree of social capital, so the a priori sign of the coefficient
is ambiguous.
Source: Own figure.
Notes: * Land is owned by the state in Vietnam. Nevertheless, the government allocates land use
certificates to farm households, the so-called ‘Red Books’ for agricultural land (valid for 20
years) and, ‘Green Books’ (valid for 30-50 years) for forest land. Farmers are allowed to sell
or rent land use certificates, or pass them on to children (LUIBRAND 2002).
** Poverty usually refers to a lack of human, social, and physical capital. Therefore, some variables
that indicate a lack of a mixture of the different forms of capital partly capture the influence
of poverty on access to formal credit.
HH = household.
4.2.3 Regional focus and sampling procedures
Data collection took place from March 2001 to March 2002. Detailed financial
market data were collected at the household level, including information on
household, farm, and business activities, assets, savings and borrowing transactions
with both formal and informal sources, and households’ perceptions of
their formal sector borrowing opportunities. The survey also documented household
consumption and labor market participation.
Two research sites were selected, namely Bac Kan province (Ba Be district) and
Son La province (Yen Chau district). Both provinces are located in the mountainous
regions of Northern Vietnam and are among the poorest provinces of the
country (WORLD BANK 1999). Ba Be district is a very remote area and has only
recently (in 1999/2000) gained access to regional and interregional markets.
Farmers produce mainly for subsistence and a large proportion of them may be
considered poor. Due to the creation of the Ba Be National Park (close to Ba Be
town), huge resettlements took place, aggravating the socio-economic problems
in the region. The Yen Chau district has a much better connection to regional
markets (Son La town) and even to greater Hanoi, and therefore offers a good
contrast to the situation in Ba Be district.
Chapter 4
68
Table 4-2: Research areas and sample composition
Selected
households
Province
and
district
Commune Village Main ethnic
minority
Number of
households
per village
Number %
Dia Linh Pac Nghe 1 Tay (97%) 76 36 47
Nghien Loan Khau Nen Nung (89%) 36 19 53
Province
Ba Kan,
district
Ba Be
Xuan La Thom Meo
Khuoi Khi
Tay (93%)
Dao (100%)
84
40
43
20
51
50
Sap Vat Sai
Na Pa
Dong
Thai (85%)
Thai (100%)
Thai (100%)
80
64
48
42
33
25
53
52
52
Province
Son La,
district
Yen Chau Chieng Hac Bo Kieng Hmong (100%) 20 13 65
Chieng Pan Than
Tat Heo
Kho Mu (100%)
Thai (100%)
38
16
20
9
53
56
Interviewed households in total: 260
Source: Own figure.
The communes and villages were stratified in accordance with pre-defined selection
criteria to ensure a good degree of variance in the sample. These criteria
are:
- Being located at different altitudes (valley, middle slope and top position),
to obtain different stages of market access, ecological zones and ethnic
minorities;
- Being engaged in different phases of the land allocation process (land allocation
completed or not completed, percentage of households with land
use certificates), and
- Having different shares of households with non-farm activities.
An overview of the sample in both regions is given in Table 5-3. Half of all
households in each village were randomly selected after being stratified according
to their living standard into five categories using official poverty data from the
village headmen.59
59 The Vietnamese government classifies every household once a year according to its
living standard into one of five categories: Hungry, poor, medium, better-off, and rich
(DUFHUES et al. 2002; and GEPPERT and DUFHUES 2003).
Outreach and access
69
4.3 Outreach of and access to formal rural lenders in Vietnam
The following section first discusses the general use of collateral by formal
lenders in Northern Vietnam. The situation as regards effective credit demand
is then presented in Section 4.3.2, followed by the outreach of rural lenders
(Section 4.3.3) and econometric analysis of access-constrained households
(Section 4.3.4).
4.3.1 Collateral use
The main collateral demanded by lenders in Vietnam is physical collateral in the
form of land use certificates (Red and Green Books). Social collateral in the
form of references is also a widespread requirement, particularly in Northern
Vietnam, and both forms are often intermingled.60
In Vietnam, lenders face enormous difficulty in enforcing pledges and mortgages
(RIEDEL 2000; UNDP 1999). Banks are not usually allowed to seize land
from defaulting farmers, even if the use rights have been pledged. It is more or
less impossible to evict farmers and auction their land because of the lack of legal
infrastructure and resistance from local authorities (WOLZ 1997). Only a few
cases exist where land has been liquidated in the event of a farmer’s collapse
(DUONG and IZUMIDA 2002). It appears that the underdeveloped legal framework
does not prove effective for the use of physical collateral as a risk management
tool (GOTTWALD and KLUMP 1999). Nevertheless, the VBARD still insists
on land use certificates as collateral and relies mainly on the psychological pressure
related to the possibility of losing land.61 As the liquidation of collateral is
almost impossible in practice (although this may not be known to the farmers),
rescheduling of the loan is often the only possibility for the credit officer to
avoid designating a loan as overdue, not performing, or lost. Therefore, rescheduling
of loans in VBARD/VBP is extremely high (IZUMIDA and DUONG 2001;
VBARD and DANIDA 1999). According to BINSWANGER and SILLERS (1983),
farmers, and particularly poor farmers, in developing countries are almost
60 A detailed description of the use of different kinds of physical collateral in the credit procedures
of rural lenders in Vietnam can be found in DUFHUES et al. (2004a).
61 In contrast to VBARD practice, the government and VBARD headquarters state that
households can take out loans of less than ten million VND without any collateral (see
VBARD 2001b).
Chapter 4
70
universally risk-averse and often reluctant to put their assets at stake as collateral
for a loan. However, as DUFHUES et al. (2004a) found, even poor farmers in
Northern Vietnam are not afraid to pledge their land use certificates as collateral.
They may be convinced that their investment will not fail or, more likely, they
believe that the bank will not seize the land even if they have difficulty repaying
their loan.
Officially, the VBP uses solely group lending schemes with joint liability for
delivering its loans (VBP 1999b).62 Anecdotal evidence, however, revealed that
in some cases of so-called ‘hungry’ households, which are officially excluded
from access to VBP loans as they are considered too poor, the credit officers insist
on collateral in the form of land use certificates, too. In this sample, over
25% of all VBP credits (N=94) were secured by physical collateral in various
forms. The so-called joint liability credit groups are not working effectively, as
some of the regulations enforced by the VBP seem to neutralize the peer pressure
that is important for good credit repayment records. In everyday practice,
the group members are not held liable for each other; the person who fails to
repay the loan is simply expelled from the group and no negative consequences
are imposed on the other group members. As joint liability does not work in the
case of VBP, and physical collateral fails to achieve its intended purpose in the
case of VBARD, both require guarantees from local authorities in the form of
‘certificates of good conduct’ and they rely on an extensive network of non-bank
local officials, who support the banks in screening, monitoring, and enforcing
the loans (DUFHUES et al. 2002). However, supervision of these structures can
be difficult. For instance, TODD (1996) reported that when a loan officer who
was at the same time president of the local political committee resigned, 100
borrowers defaulted as a result. Thus, the delegated task of putting pressure on
borrowers does not always work, as the substitution of physical collateral by social
collateral leads to a delegation problem in which the lender must concern itself
with whether or not the third party charged with imposing social sanctions will
actually carry out this task (BOND and RAI 2002).
62 The VBSP, the successor of the VBP, will continue to provide physical collateral-free
loans to certain target groups (VIETNAM ECONOMY 2003).
Outreach and access
71
In summary, physical collateral works solely through the psychological pressure
it exerts. The social collateral of joint liability groups does not work at all.
The only form of collateral that seems to work is social collateral in the form of
pressure exerted by local authorities and in the form of denial of future credit in
the event of default. However, it remains questionable how long this pressure
can be maintained. Anecdotal evidence from the WORLD BANK (2003, 2004b)
suggests that in poor communes 70% of households default in their installment
payments to VBSP, while around 30% default in their payments to VBARD.
4.3.2 Effective formal credit demand 63
In total, 56% of all households in the sample (total N=251) have had an effective
demand for formal credit (Figure 4-2). This clearly demonstrates the enormous
breadth of outreach of the formal financial sector in Vietnam. Only a mere 16% of
households are involved in the informal market and 23% of the total loans in this
sample are borrowed in the informal sector. DUONG and IZUMIDA (2002) found
that in their survey an even lower number, namely 17% of all loans, were extended
by informal sources, and the WORLD BANK (2003) states that the majority of
households have access to formal credit in one way or another. These figures stand
in contrast to the observations made in many other developing countries, where the
informal sector is still the biggest supplier of financial services. As one important
factor for the reduction of the informal sector from about 80% in the early 1990s
to around 20% today, MCCARTY (2001) mentioned ‘crowding out’ by the VBARD
and VBP, both of which have extended their outreach enormously in recent years.
Nevertheless, the VBP has done this at the cost of financial sustainability. On the
one hand, crowding out moneylenders who charge usury interest rates is a very
welcome effect. However, not all moneylenders charge usury interest rates. So,
crowding out the informal sector can have very negative effects by destroying informal
financial networks without replacing them with a sustainable formal alternative.
In Vietnam these informal structures have been replaced by formal ones, but
the questions remains: Are they sustainable in the long run? The VBP was recently
replaced by the VBSP, which will continue the policy of the VBP. It is just a matter
63 Formal credit is defined as credit from the formal and semi-formal financial institutes,
VBARD, VBP, and the State Treasury. However, as the State Treasury disbursed only few
credits, it is excluded from the later analysis.
Chapter 4
72
of time when the VBSP will become unsustainable, as its interest policy cannot
cover its costs. DIAGNE et al. (2000) state that policies and financial institutions
should be designed to complement the informal market rather then to replace it.
Over 40% of households are not engaged in the formal financial sector. ZELLER
and SHARMA (1998) state that some rural households simply do not apply for a
loan because of the expectation that they will be turned down. This statement
was supported by the work of BUCHENRIEDER and THEESFELD (2000) in a similar
research setting in Northern Vietnam. The empirical results of their research
showed that a lack of bankability from the perspective of the clients exists. This
may foster the assumption that there is still a need to extend the credit outreach
of the formal financial sector still further. Looking at Figure 4-3, it becomes
evident that this figure must be reassessed. A mere 16% of the households in the
sample are access-constrained. When farmers were asked directly why they do
not borrow in the formal sector, three answers were given most frequently:
1. afraid of debt; 2. have no collateral; 3. lack of knowledge. While ‘lack of
knowledge’ clearly refers to a lack of human capital and ‘no collateral’ to a lack
of physical capital, ‘afraid of debts’ refers to a mixture of human, physical, and
social capital, namely a deficit in the risk-bearing capacity of a household.
Figure 4-2: Percentage of households
using different
credit sectors
Figure 4-3: Access-constrained households
(formal financial
sector)
Formal
48%
Informal
8%
No credit
36%
Formal
and
informal
8%
Access
56%
No credit
demand*
28%
Access
constrained
16%
Source: Own figure.
Note: N=251.
Source: Own figure.
Note: N=251.
* Households with no credit demand but
access to the formal financial sector.
The great success of formal credit outreach in Vietnam is clearly due to creditdelivering
technologies that reach out far into the rural area, e.g. offering group
loans at the local level, but also due to the strong promotion of credit by the
government. Finally, the very low interest rates, which are highly subsidized, are
increasing effective demand and thus outreach.
Figure 4-4, Figure 4-5, and Figure 4-6 show that the market segments of the
Outreach and access
73
informal and formal credit markets are clearly separated in terms of interest
rate, loan term, and loan amount. The informal market is separated into two
different segments, first a ‘no interest rate’ segment and second a ‘high interest
rate’ segment.64 Within the first segment, small amounts of money are lent out
to family members or friends at zero interest. These credits are usually used for
consumption and are one method of coping with emergencies or an unexpected
shortage of cash. The second segment is served by traditional moneylenders.
They charge interest rates clearly above the rates of the formal sector (about
three to four times as much). While loans from the moneylender are usually
very short-term, loans from family members or friends are either short-term or
open-ended. Households using moneylenders have either fallen through the
informal safety net or they have to borrow more than their social network can
provide, e.g. in the event of expensive surgery needed by a family member.
The biggest share of formal loans is usually used for investment or productive
purposes.65
Figure 4-4: Interest rates per month of formal and informal credits
Nominal interest rate per month
0% 0.5% 1.0% 1.5% 2.0%
Number of credits
Black bars:
Informal credits (N = 44)
Grey bars:
Formal credits (N = 144)
0
10
20
30
40
C
Source: Own figure.
64 However, some households are paying an interest rate in the informal sector that exactly
matches the formal rate. Usually these households have a close friend or a relative who
borrowed in the formal sector and passed on part of the loan to a friend, who is then
charged the formal interest rate.
65 After financial reforms at the beginning of the nineties, the real interest rate became positive
in 1992 (SENANAYAKE and HO 2001). Since then, however, the real interest rates have
been gradually fallen (DUFHUES et al. 2004a).
Chapter 4
74
Figure 4-5: Formal and informal credit terms
0
20
40
60
80
Open-end
Loan term in years
Number of credits
0.01 1 2 3 4 5
Black bars:
Informal credits (N = 44)
Grey bars:
Formal credits (N = 144)
C
Source: Own figure.
Formal and informal credits are imperfect substitutes for each other. In particular,
formal credit reduces, but does not completely eliminate, informal borrowing
(DIAGNE 1999). Neither of the informal segments, the ‘no interest’ or the ‘high
interest’ segment, can be reached by traditional credit products, and it would be
probably a drain of public recourses if the Vietnamese government tried to do so.
But the two segments could probably be reached by financial products other than
credit. For instance, the lower segment could be served by client-adapted savings
products, and the moneylender segment by a functioning social security system.
Figure 4-6: Loan size of formal and informal credits in VND millions
Informal
(N = 44)
Formal
(N = 144)
Million
VND
0 1 2 3 4 5 6 7 8
Source: Own figure.
Notes: 1 million VND = 70 US$ at time of survey (average conversion rate).
Outreach and access
75
4.3.3 Credit outreach
The poverty outreach of the formal lenders is analyzed using a poverty index
calculated with the PCA. Households that have no effective credit demand in the
formal financial sector are the reference group. These households were first
ranked according to their poverty index and then sorted into five groups of equal
size. The lowest quintile incorporates the poorest households and the upper
quintile embraces the better-off families. When assessing the poverty outreach
of microfinance institutes at the household level, only new clients should be included
in the analysis in order to rule out any impact that could have occurred
due to the financial services obtained from the lender and that could have led to
a change in the poverty status of the client (MATIN et al. 1999).66
When looking at new clients of the formal financial sector, it is clear that betteroff
households are over-represented and that the poor and poorest households
are under-represented (Table 4-3). Nevertheless, clients are considered, only the
poorest group is under-represented. This might indicate a shift by formal lenders
towards the better-off households.67 Of course this is a rather naïve comparison,
as any impact of the program is neglected. But assuming that there is no impact,
the comparison would hold, and if a positive impact is assumed, the gap would
in fact widen. Only in the case of a negative impact would the gap decrease.
Discussions with farmers did not reveal any wide-scale negative credit impacts.
Thus, the last option is unlikely to be realistic. Hence, it is likely that there is a
shift towards the better-off clientele. Whether this shift is actively influenced by
formal lenders, or rather passively, as all creditworthy poor households are already
clients, remains unanswered.
66 HENRY et al. (2003) suggest using only clients who received a loan within the past three
months. However, most of the loans in this sample are invested in livestock production –
mainly pigs, cattle or buffaloes (DUFHUES et al. 2004b). Initial profits are not expected
within the first six to seven months. Thus, it is safe to consider clients who received a loan
within the last seven months as unbiased by poverty impacts of the loan. These households
are grouped in Table 4-3 as ‘new clients’.
67 Almost three-quarters of the new loans were made by the VBP. Thus, a targeting bias
through VBARD, which focuses more on the better-off clientele, can be excluded.
Chapter 4
76
Table 4-3: Depth of outreach of formal lender clients
← increasingly poor/increasingly rich →
Group 1 Group 2 Group 3 Group 4 Group 5
No clients, reference group
(N=111)
20% 20% 20% 20% 21%**
Effective access (new clients)
(N=40)
0% 13% 28% 40% 20%
Effective access (all)* (N=140) 5% 24% 22% 26% 23%
VBP clients (N=94) 4% 30% 23% 27% 16%
VBARD clients (N=35) 6% 9% 26% 23% 37%
Source: Own calculations.
Notes: * Including VBP, VBARD, and State Treasury clients.
** Because of rounding, the figures may not add up to 100%.
When looking at Table 4-4, it is clear that the poorest group (namely group 1) is
clearly less often served by formal credit than the other groups. Only 24% of the
households belonging to the poorest group have a formal loan. Surprisingly, the
other poverty groups are more or less equally served. The VBP uses a targeting
system which focuses on the poor but excludes the poorest, and the VBARD
focuses on non-poor households (DUFHUES et al. 2002). However, if the poorest
have collateral in the form of land use certificates they can usually access credits
from the VBP and VBARD (see Section 4.3.1). Nevertheless, the poorest
households in this survey seem to have difficulty accessing formal bank loans in
comparison to wealthier households, which corresponds to international experience
(see Section 4.1). On the one hand, as stated by MOLISA and UNDP (2004),
bank staff and local authorities fear that the poorest will fail to pay back their
loans. On the other hand, many of the poorest households may simply not demand
the credit products on offer. The most realistic scenario is probably a mixture of
low creditworthiness and low demand, a scenario also supported by research in
other developing countries. For instance, NAVAJAS et al. (2000) state that in Bolivia
the poorest households are less likely to be assessed as creditworthy and/or
to demand loans of the type offered by the industry.
Table 4-4: Outreach of formal credit by poverty group
← increasingly poor/increasingly rich →
Group 1
(N=29)
Group 2
(N=56)
Group 3
(N=53)
Group 4
(N=58)
Group 5
(N=55)
Households with access per group 24% 61% 58% 62% 58%
Source: Own calculations.
The two banks, the VBP and VBARD, serve clearly under-proportionately the
poorest group of the sample population (Table 4-3). This finding is also confirmed
by IZUMIDA (2003), who states that the so-called ‘hungry’ households have
Outreach and access
77
rarely been reached by the VBP. However, in line with their different target
groups, both have a different depth of poverty outreach. If one considers the two
lower quintiles as fairly poor and the three upper quintiles as fairly rich, then
almost 40% of all poor households in the sample were reached with credit by the
VPB. This is also confirmed by the WORLD BANK (2004a), which states that the
VBP has been fairly successful in reaching the poor. Fairly may be interpreted
as a noticeable outreach to the poor, although it is still an under-proportionate
outreach. However, the aim of being a ‘Bank for the Poor’ is obviously not being
achieved, as two-thirds of the VBP’s clients in this sample are not considered
poor.68 Another way of measuring depth of outreach is the loan amount and loan
term. Smaller amounts or shorter loan terms usually mean greater depth
(CHARITONENKO et al. 2004; SCHREINER 2002). Considering only the loan term,
one might think that the VBARD has deeper outreach, with an average loan term
of 2.5 years compared to 2.9 years in the case of the VBP. Nevertheless, there
are two facts that contradict this. First, the interest rate is higher in the case of
VBARD loans. As a result, farmers probably try to keep the loan term as short
as possible, and second, the loan terms are more negotiable in the case of
VBARD loans (Figure 4-7 and Figure 4-8). The VBP loans are ‘one-size-fits-all’
loans. Furthermore, VBP loans are clearly smaller than VBARD loans, sug
gesting that VBP has a higher depth of outreach than VBARD. The average loan
amount of a VBP loan is 2.8 million VND (177 US$) per credit group member
compared to 6.6 million VND (420 US$) from VBARD. Nevertheless,
CHRISTEN et al. (1995) point out that scale determines whether significant outreach
to the poor can be achieved. The VBP not only has greater depth of outreach,
but also greater breadth. Two-thirds of all loans in this sample are disbursed
by the VBP and only one-quarter by the VBARD.69 Thus, considering
only Northern Vietnam, the VBP reaches deeper into the poorer part of the
population than the VBARD.
68 Official data on this issue are much more optimistic. For instance MOLISA and UNDP (2004)
state that three-quarters of the subsidized loans are delivered to poor households.
69 However, when considering national data the VBARD has a much bigger breadth of outreach
then the VBP.
Chapter 4
78
Figure 4-7: Loan term of formal credits in years
VBP
N = 97
VBARD
N = 37
Years
3
2
1
0
Source: Own figure
Figure 4-8: Loan amount of formal credits
VBP
N = 97
VBARD
N = 37
Million VND
10
8
6
4
2
0
Source: Own figure
4.3.4 Credit-constrained households
Of the 260 households interviewed, nine households were dropped prior to the
analysis and eleven were excluded during the analysis due to missing values.
Thus, the parameters are estimated on the basis of 240 households. A binary
logit model is used, where the dependent variable is one for all households with
access to formal credit and zero otherwise. The dependent variable is derived
following the decision tree in the annex Figure A-1. The list of regressors is presented
in the earlier section (see Table 4-1 in section 4.2.2). From 23 potentially
influential parameters, eight variables have a significant influence in the model
presented. These variables are displayed in Table 4-5.
Outreach and access
79
Table 4-5: Parameters influencing households’ access to formal credit –
Binary logit estimation
Standard
error
Significance
Exp(B)
odds ratio
Physical capital-related variables
1. Red or Green Books (yes = 1 no = 0) 0.229 0.049 1.570
2. Agricultural land (m2) 0.506 0.247 1.798
3. Value of houses (VND) 1.119 0.051 8.919
4. Government salary (yes = 1 no = 0) 0.390 0.438 0.739
5. Cash savings (VND) 1.720 0.686 2.005
Human capital-related variables
6. School years of HH head (years) 0.415 0.419 1.399
7. Vietnamese communication
skills of the married
couple
(yes = 1 no = 0) 0.824 0.954 1.048
8. Receiving agricultural
extension
(yes = 1 no = 0) 0.284 0.230 1.407
9. Active household
members
(numbers) 0.310 0.321 0.735
10. Share of non-farm activities
in total yearly income
(yes = 1 no = 0) 0.354 0.623 0.840
11. Lost working days (days/year/HH) 0.482 0.823 0.898
Social capital-related variables
12. Giving help (days/year/HH) 0.348 0.703 1.142
13. Receiving help (days/year/HH) 0.261 0.023 0.553
14. Interest-free informal
credit
(VND) 0.597 0.100 0.374
15. Thai/Tay village (yes = 1 no = 0) 0.770 0.004 9.007
16. Market visits only by
females
(yes = 1 no = 0) 0.304 0.018 0.488
17. Age of the household (years) 0.287 0.622 0.868
Infrastructure-related variables
18. Remoteness (km) 0.393 0.006 2.916
19. Market visits per month (numbers) 0.868 0.041 5.897
20. Different market visited (numbers) 0.416 0.233 1.642
Poverty-related variables
21. Poverty index (index) 0.462 0.553 1.315
22. Day labor (yes = 1 no = 0) 0.331 0.353 0.735
23. Receiving aid form
government
(yes = 1 no = 0) 0.239 0.858 0.958
Chi-square 107.309
Nagelkerke R Square 0.619
Observations in model 240
Source: Own calculations.
Notes: Variables over a significance level of 0.1 are considered to be not significant.
The model predicted 92% of all observations correctly (Table 4-6). In the group
of access-constrained households, the percentage of correctly predicted cases
Chapter 4
80
was good, at 71%; it was very good, at 96%, in the group of households with
access to the formal financial market. The overall fit of the model is satisfactory,
with a Nagelkerke R2 of 0.619. The correlation tables showed no problems due
to multicollinearity between independent variables.
Table 4-6: Classification of correctly predicted access to formal
credit-constrained households
Predicted Percent correct
Observed 0 1
0 27 11 71.1
1 8 194 96.0
Overall percent correctly predicted 92.1
Source: Own calculations.
Notes: The categories of the dependent variable are: 0 = No access to formal credit; 1 = Access to
formal credit.
Physical capital: In contrast to other countries, lack of physical capital in the
form of farmland is not a significant access constraint.70 However, the possession
of land use certificates is significant, but the influence is the lowest of all
significant variables. In the 1990s, one of the most important access constraints
to formal rural credit in Vietnam was the lack of physical collateral in form of
land use certificates (HUNG and GIAP 1999). Today, land use certificates seem to
have lost most of this influence. The ongoing dissemination of ‘Red/Green
Books’ in recent years has brought an increasing number of households into
possession of assets that are useable as collateral, and this has broadened the possible
outreach dramatically (DUFHUES et al. 2004a; MCCARTY 2001).71 Moreover,
the main supplier of formal loans in the research area is the VBP, which tries to
rely totally on social collateral (section 4.3.1 and 4.3.3). This explains the fact that
possession of land use certificates appears to have little influence on access to
formal credit. Nevertheless, households without certificates may still have difficulties
in accessing formal loans, particularly from the VBARD.72
70 For instance, SARAP (1990) found that the smaller size of land holdings in India has an
adverse effect on the access of small farmers to formal credit institutions.
71 In this sample, 89% of the households have a Red Book and/or a Green Book.
72 Lack of collateral was also mentioned by access-constrained households as a reason for
self-exclusion.
Outreach and access
81
A high-value home has much greater influence on the likelihood of obtaining
access to formal credit than land use certificates. Housing is probably used as a
visible indicator of the general wealth of the household and can easily be assessed
by local officials or credit officers. Furthermore, houses can be used as formal
collateral.73 This may explain the considerable influence of the value of housing
on the likelihood of obtaining access to formal credit. The visible wealth of a
household seems to be very important for its access to credit.74
Human capital: None of the human capital-related variables is significant. Particularly
surprising is the fact that the variable ‘school years of HH-head’ is not
significant, as research from other developing countries, e.g. EVANS et al. (1999),
SARAP (1990), and VAESSEN (2001), support the opposite notion. Moreover,
illiteracy levels in Vietnam remain high among the poor ethnic minorities, and
in more remote parts of the country, especially in the northern mountains
(BHUSHAN et al. 2001). This may indicate that the formal credit application
process is not in itself a market entry barrier any more to poorly educated
customers. Group credits from the VBP have probably eradicated this access
constraint. Within the group credit scheme, only the credit group leader who submits
the credit proposal needs a certain degree of literacy (DUFHUES et al. 2002).
Furthermore, investments are not usually very innovative and revolve mostly
around conventional enterprises in animal production. Thus, a high amount of
human capital does not seem necessary to carry out those investments.
Social capital: Dwelling in a Thai/Tay village has the highest influence on the
likelihood of having access to formal credit. As mentioned above, from a national
point of view Tay and Thai are ethnic minorities. However, in the two research
areas these minorities represent the majority and occupy many key positions in
the local administration, including the district bank branches. It is not surprising,
therefore, that inhabitants of non-Thai/Tay villages have a significantly higher
chance of being access-constrained then the ethnic majority in Ba Be and Yen Chau
respectively. However, privileged access to credit may be related not only to
73 However, in this sample only one case was found where the house was used as collateral
for a VBARD loan.
74 The housing status of a household must not be equated with its poverty status. GEPPERT
and DUFHUES (2003) and SIMANOWITZ (2000) stated that the appearance of housing is an
insufficient indicator for poverty, for instance housing could be completely debt-financed.
Chapter 4
82
ethnicity, but also to the fact that the predominant farming system in these villages
is paddy rice production. Paddy rice production was for long time seen by local
officials and bank staff as a farming system with a high developmental priority.
Thus, inhabitants of these villages were likely to obtain easier access to formal
credit. Moreover, these villages usually have a better market connection and are
thus more easily accessed by bank staff, which also promotes access to loans.
The numbers of days of informal help and the amount of interest-free informal
loans that a household receives from its social network may both be indicators of a
functioning social network, but they may also lower the chance of gaining access
to formal credit. A functioning informal social network has apparently no positive
influence on the likelihood of obtaining access to formal credit. But high use of the
social network is obviously a strong indicator of an income shock or a shock that
negatively influences the repaying capacity of the household.75 The incidence of
shocks either raises the chance of being assessed by local authorities as not creditworthy
or adds to the self-exclusion tendency of households.
In households where only female persons go to the market (i.e. the wife of the
household head or a female household head), the chance of being accessconstrained
is higher then in other households. On the one hand, this may be
caused by a lack of human capital, as the male household head is mentally or
physically unable to go to the market. While the number of missed working
days due to illness is not significant, the permanent absence of an important
member of the workforce such as the male household head may be more important.
Moreover, the absence of the male household head can be easily observed
by local authorities or bank staff and may indicate a problem inside the
household, which may lower its creditworthiness. On the other hand, women
probably have less access to official information networks than men.76 Households
of low social standing are often excluded from information networks and
thus lack important information. VAESSEN (2001) draws similar conclusions
from a research study in Nicaragua. One reason mentioned by households that
are access-constrained to formal credit is lack of information.
75 The use of informal borrowing may also be seen as an indicator of poverty. However, other
research, e.g. by RUTHVEN (2001), has shown that reciprocal lending is not necessarily related
to poverty.
76 MCCARTY (2001) points out that the vast majority of rural loans are given to men.
Outreach and access
83
Infrastructure: Surprisingly, the variable ‘distance to the district capital with the
nearest bank branch’ is positive. Former research in Vietnam, e.g. by HUNG and
GIAP (1999), has shown that remoteness is an access constraint to formal credit.
Nevertheless, many programs around the world are set up specifically to serve the
under-served. They locate where financial services have long been weak
(MORDUCH 1999). In the case of Vietnam, those areas are/were certainly the more
remote areas. Through the VBP and special credit programs, the Vietnamese
government has pushed credit outreach for poverty reduction in favor of remote
communes. MINOT (2000) states that poverty is more pronounced in places that
are more remote from markets and cities. Obviously program placement is not
random, and remoter villages or communes with a higher share of poverty are
preferred. However, ALTHER et al. (2002) state that the poorest villages may not
necessarily be located inside the poorest communes. Poor villages in wealthier
areas may have less access to government services.
The variable with the third highest influence is ‘frequency of market visits’. The
high influence in this case may be explained by the lack of information about
credit application procedures and availability in the village, as described in
DUFHUES et al. (2002). Farmers use market days not only for buying and selling,
but also for exchanging information and keeping relations and networks alive.
DUFHUES et al. (2002) also showed that good relations with commune officials
or the credit officer are essential for receiving credit. These contacts are most
likely to be established and maintained on market days. This finding is supported
by VAESSEN (2001), who found that access to certain information and
networks in Nicaragua is essential for receiving credit. The more often a household
goes to the market, the more investment ideas the household may develop.
Consequently, the bigger its demand for credit and the lower its self-exclusion
tendency is likely to be.
Poverty:77 A very surprising result is that the poverty variable does not have a
77 Better access to credit potentially reduces the incidence of rural poverty via a positive impact
of the credit itself and, in the case of not borrowing, via access to consumption loans
at times of income shocks. Therefore, poverty is endogenous to credit access and the inclusion
of a poverty index would result in biased coefficients. However, in this case study,
endogenity of poverty is expected to be negligible for two reasons: First, in the past, formal
Vietnamese lenders did not disburse consumption loans. Second, QUACH et al. (2003, 2004)
Chapter 4
84
significant influence on the chance of a household being access-constrained.
For instance, former research by DUONG and IZUMIDA (2002) concluded that the
poor have great difficulty in accessing formal credits. Furthermore, the descriptive
statistics in section 4.2.2 indicate that particularly the poorest households
tend be excluded. On the one hand, this may be explained by the fact that the
Vietnamese government has made major efforts to increase the supply of formal
credits to the rural poor, and clearly these efforts have been successful. On the
other hand, the poorest households may potentially be able to obtain access to
formal credit, but they do not demand it as their investment possibilities are very
limited, their debt-taking capacity is low and their risk-aversion is high.78
4.4 Conclusions and policy recommendations
This research paper has contributed to enhance our understanding of the broad
outreach of formal rural lenders in Northern Vietnam. Compared to a decade
ago, the informal market now plays only a minor role. The poverty outreach of
the formal rural lenders is satisfactory (e.g. about 50% of the households belonging
to the two lower poverty quintiles have effective credit demand). However,
the outreach analysis has shown that the poorest households are seldom
clients of formal lenders. Only slightly more than 20% of the poorest households
are served by formal lenders, while among the wealthiest quintile this proportion
found in their studies only a very small impact of formal credit on household welfare in
Vietnam. As mentioned in Section 3.3, households who received a formal loan within the
last seven months are likely to be unbiased by poverty impacts of the loan. Therefore, a
new model was calculated with a sub-sample of households. In this sub-sample, households
that had a formal loan with duration of more than seven months were excluded. The
results from this model also indicate that poverty status had no influence on the likelihood
of a household being access-constrained to formal credit.
78 First the poverty index variable was squared and then multiplied by its original value,
which can also be negative, to give a higher weight to the extreme cases. Second a dummy
variable for the lowest quintile of the poverty index was incorporated in the regression instead
of the poverty index itself to give a higher weight to the poorest households. But, in
both cases the poverty variable turned out to be not significant. However, when dropping
all variables except the poverty index as explanatory variable, this variable becomes significant.
This indicates that poverty does play a role but its single dimensions or characteristics
which are already reflected by other variables in the model turn out to be more important.
As mentioned above the correlation table did not reveal any problematic correlations between
the independent variables.
Outreach and access
85
is around 60%. When considering access to formal credit, this figure must be
re-evaluated. As the logit analysis revealed, general poverty (as captured with
the poverty index) does not significantly influence access to formal credit. The
results indicate that only certain aspects of poverty, e.g. low quality of housing,
have an important influence on access to formal credit in Vietnam. Thus, the
poorest households use formal credit less often, but are not significantly more
often access-constrained. This means that the poorest households simply have
much less demand for the types of formal credit products on offer. Improving
credit products or offering new credit lines would only slightly improve the
credit coverage of poorer households. A more promising approach would be to
introduce a specialized pro-poor extension service to widen the scope of their
investment ideas and opportunities, combined with a general improvement in the
infrastructure. One factor that very positively influenced access to formal credit
was the connection to the market. A good market connection serves credit outreach
in a twofold manner: First, households have better access to credit-relevant
information; and second, through better market access they may find new investment
opportunities. The most appropriate tool to incorporate poorer households
into the formal financial system would the mobilization of savings. As stated by
several scholars, all people can save, even the poorest of the poor, and therefore
deposit services have deeper outreach than credit (CHARITONENKO et al. 2004;
SCHREINER 2002; ZELLER and SHARMA 2000). Moreover, DUFHUES et al. (2004a)
revealed an enormous and unmet demand for formal savings among the rural
population in Northern Vietnam.
Nevertheless, the number of access-constrained households is surprisingly low.
One reason for the low number is the weakening or eradication of former access
constraints. A major access constraint in the last decade was lack of collateral.
Land use certificates are nowadays widespread and they have only limited influence
on access to formal credit. If the process of issuing land use certificates is
finally completed and/or if the VBARD and the VBSP enforce their national
policy of granting collateral-free loans, the number of households without access
to the formal financial market due to lack of land use certificates will dwindle,
and the importance of land use certificates as physical collateral will become
even less important. Other access constraints, namely remoteness (distance to
the nearest bank branch) and literacy requirements of the household head, have
been eradicated through locally disbursed group credits. However, considering
the anecdotal reports of very low repayment rates, the price of eradicating these
access constraints has likely been a decrease in sustainability of the formal lenders.
As VON PISCHKE (1991:83) points out: "Binding financial constraints to the formal
sector are not necessarily anti-developmental. If they stand in the way of
Chapter 4
86
bad investment, they are socially beneficial. For the financial system to make
good loans and allocate funds for high-return investments, it must reject poor
proposals, unfit applicants, and low-return investments. To generate good loans
requires a highly selective screening of applicants."
Nevertheless, some barriers to access continue to exist. To reduce these access
barriers, locally-oriented rather than general actions should be taken, catering to the
needs and the circumstances of those households which lack access, particularly
ethnic minorities and female-led households. Such measures will be particularly
useful within the organizational structure of the lenders themselves.
DUFHUES et al. (2002), for example, suggested employing ethnic minority credit
officers, which would create more awareness of those groups inside the institution.
Furthermore, special female credit officers for female-headed households
could work in the same direction. However, the recent efforts of the Vietnamese
government, for instance the establishment of the VBSP, represent an attempt to
broaden access in general. But this increase in outreach will go hand in hand
with an increase in access to credit for non-creditworthy households, thus resulting
in decreased repayment rates. Moreover, it is questionable whether households
that do not have access today, or do not demand the existing products, will demand
credits from the VBSP. If rural lenders in Vietnam were one day forced to
work in a competitive environment according to market principles, there would
be a great chance that large parts of the population would be access-constrained
as a result of previous loan defaults.
Whether the breadth and depth of formal credit outreach that was financed by
enormous government subsidies contributes to poverty reduction, however, remains
a subject for future research. Finally, it should be borne in mind that the
conclusions drawn in this paper are based on a rather small and regionally limited
sample. Therefore, further confirmation of the results by a larger and more representative
study would be desirable.
5 Participatory product design by using Conjoint Analysis
in the rural financial market of Northern Vietnam79
The following sections discussed the development of client adapted financial
products by measuring customer preferences in a participatory way. In this sense
credit and saving products are discussed separately.
5.1 Introduction
Vietnam has been engaged in a process of enhancing the role of market forces in
the economy since 1986-87. This transition process, known in Vietnamese as
‘Doi Moi’ (‘renewal’), is aimed at restructuring Vietnam’s legal, regulatory, administrative,
investment and foreign trade apparatus and policies to transform its
centrally planned economic system into a market economy with ‘socialist characteristics’
(BRYANT 1998). In some sectors the reforms have achieved impressive
results, bringing economic growth rates up to an annual average of 7.5% for the
years 1995-99. Nevertheless, Vietnam is still an extremely poor country
(WORLD BANK 1999). Three-quarters of Vietnam’s population are engaged in
subsistence farming and the majority of them lives in rural communities (BA 1997).
In the transition process, the reform of the national financial system plays a key
role (SCHRIEDER and HEIDHUES 2000). The tasks ahead include: Strengthening
popular faith in the financial system, moving the banking sector to financial
self-sustainability, and expanding its outreach to include the newly emerging
private sector not only in urban but also in rural areas. A financial system that
enjoys the confidence of the population will be able to boost the domestic financial
79 This section is based on the following article: "Participatory product design by using conjoint
analysis in the rural financial market of Northern Vietnam", written by Thomas Dufhues,
Franz Heidhues, and Gertrud Buchenrieder and published 2004 in the Asian Economic
Journal 18 (1): 81-114. I would like to thank Prof. Dr. Manfred Zeller for his helpful
comments to an earlier Version of this artikel.
Chapter 5
88
savings rate, now hovering at around 20%, up to rates of 30-40%, as attained in
other East Asian countries (WORLD BANK 1995; WORLD BANK 1998). Vietnam
is still largely a cash economy, with cash accounting for about 50% of the M3
money supply (WOLFF 1999). There is consensus that in transition countries
such as Vietnam, for the financial system to develop a comprehensive service portfolio
to all market segments, the regulatory and supervisory framework for
banking needs strengthening. In this context, the paper refers to the development
of demand-driven financial services in a participatory way.
Broad access to appropriate and sustainable financial services has been pointed out
repeatedly as being important for poverty reduction. It contributes to higher incomes
and better food security (ADB 2000; BUCHENRIEDER and THEESFELD 2000;
HEIDHUES 1995; SCHRIEDER 1996; ZELLER et al. 1997). In Vietnam many poor
households are confronted with transitory food insecurity, even though their incomes
seem to provide an adequate livelihood base over several years. Thus,
there is a potential demand for savings, credit, and insurance services to more
effectively stabilize consumption and to increase the ability to escape chronic
poverty (KANBUR and SQUIRE 2001; SHARMA 2001; ZELLER 1999). Hunger
eradication and poverty alleviation, particular in the northern Uplands, is of
enormous concern to the Vietnamese government (VBP 1999b). MARD proposes
financial services as a powerful tool for poverty reduction (SIDA-MARD
1998). From the early 1990s onwards, the Vietnamese government has begun to
establish and promote FFIs such as the VBARD, the VBP and the PCFs to provide
the rural poor with cheap loans. This policy is based on the assumption that
(1) the rural population is too poor to repay credits at market interest rates and
that (2) capital is the factor constraining production and thus limiting food
security.
The Vietnamese FFIs have been very successful in increasing their level of outreach.
The state-owned VBARD and VBP have reached almost 4.5 million and
2.5 million rural households with credit, respectively (BAC 2001; HANH 2001).
These are more than 58% of all rural households in Vietnam. Despite this success
in terms of outreach, many rural households demanding credit still lack access.
Even those credits that are targeted towards the poor seem to be bypassing the
poorest groups, which often are identical to ethnic minorities (NEEFJES 2001;
WORLD BANK and DFID 1999). The supply of formal savings schemes is insufficient
in Vietnam. Formal savings schemes have not been in the focus of
development efforts. Rural households are widely assumed to be too poor to
save, and therefore the VBP, for example, does not offer any saving schemes.
However, theoretical and empirical evidence suggests that even poor people
Participatory product design
89
want to, need to, and indeed do save for various purposes (KANBUR and SQUIRE
2001; RUTHERFORD 2000; SHARMA 2000; WRIGHT and MUTESASIRA 2001;
ZELLER 2001).
The development of adapted financial products has been the orphan of research
on financial markets in developing countries and is only now coming into its
own (RUTHERFORD 2000). Thus, there is a pressing need to examine ways of
designing and introducing new financial service products into MFIs to improve
their outreach, depth, and access to them (WRIGHT 1999).
This research work aims at developing innovative client-oriented financial services
in a participatory way to improve access to financial services. The objectives
can be formulated as: 1. analysis of the supply of formal financial services,
and 2. design of client-oriented financial services by means of participatory
research.
5.2 Material and methods
The complexity of the rural financial market system requires a whole set of different
research methods and tools. Data from three different levels, the household
level (demand side), the level of financial institutes (supply side), and the
community level need to be collected and analyzed.80 However, this research
sets a special focus on the demand side, i.e. the household level. Primary data
collection took place in the period from Mach 2001 to March 2002.
Secondary data were collected at all administrative levels. Semi-structured and
unstructured interviews with key persons such as officials of mass organizations
or political cadres at the commune and district level provided general information
on the research region and on the financial landscape in general.81 Furthermore,
primary data were collected at the commune level. Secondary data from the FFIs
80 Community level is defined as all institutions, policies, and infrastructure above village
level.
81 Due to Vietnam’s socialist history, the so-called ‘mass organizations’ are found everywhere
in Vietnam. They are hierarchically structured according to the administrative levels,
e.g. Farmers’ Union, Women’s Union, Youth Union, and Veterans’ Union. Some of these
organizations offer small loans or organize training courses or saving groups at village
level. The Fatherland Front is the holding organization of all other mass organizations and
also has political influence on the local people’s committees (GEPPERT et al. 2002).
Chapter 5
90
and non-formal financial institutes, like non-governmental organizations and
special credit programs carried out by the state, were collected. Semi-structured
interviews were conducted with bank staff from all hierarchical levels (credit
officer to bank manager and district branch to head-quarters). These data were
analyzed from the point of view of performance indicators and information
economics.
The main part of this analysis is based on the household/village database, comprising
households of different ethnicity. Cross-sectional household-level data
from the Yen Chau and Ba Be districts in Northern Vietnam were collected,
whereby the sample contains agricultural households with and without access to
services from the formal financial sector (see Section 5.2.2).
The perception of the target group, namely the rural population, is the most
important part of this research work, particularly while designing the conjoint
survey. Their view of reality permeates the whole research study to supplement
and to validate all other data. Therefore, different participatory rural appraisal
(PRA) tools were applied and had their influence on all parts of the research
and its analysis. Depending on the purpose, different stakeholders were contacted
with the PRA-tools: Women – men, poor – rich, old – young, individuals
– groups. PRA-tools applied included: Cash-flow diagrams, mobility maps,
wealth rankings, unstructured interviews (partly supported by photographs),
different rankings, visualization workshops and role-plays with an external
moderator, social mappings, and Venn diagrams.
5.2.1 Conjoint survey82
Conjoint Analysis (CA) is commonly used in commercial marketing studies
and analysis of consumers’ preferences and has its origin in psychological research
(WITTINK and CATTIN 1989).83 Assuming that a product can be defined
as a vector in a multidimensional attribute space, and that the evaluation of the
82 A comprehensive description of the methodology of CA can be found in GREEN and
SRINIVASAN (1978, 1990).
83 Preference is a one-dimensional psychological variable. It reflects the perceived advantage
between two or more alternatives. The given preference to an alternative does not mean
that this alternative is assessed as positive or good. A preference shows merely a relative
assessment of alternatives (HAMMANN and ERICHSON 1994).
Participatory product design
91
product is based on its attribute levels, it becomes theoretically possible to
relate preference to attributes (JANSSEN et al. 1991).84
Table 5-1: Credit attributes and their levels
Attributes Levels
1. Interest rate
(percent/month)
1) High (1.2%/month)
2) Low (0.5%/month)
2. Insurance of
investment in
livestock
1) Insurance of livestock investment
(livestock that died due to accident or disease (buffaloes and pigs) will
be replaced by their value at time of death; premium 5,000/month per
animal)
2) No livestock insurance
3. Disbursal time
of the loan
Disbursal time of loan in days from the first day of action to receiving
the loan (e.g. in the case of VBARD getting the application form; in
the case of VBP creating the group)
1) Quick (7 days)
2) Slow (60 days)
4. Lending scheme 1) Group lending scheme
During a group lending scheme, all negotiations with the bank and
commune authorities will be conducted by the CGL. The application
form will be filled in by the group leader, the interest will be collected
by him, etc.
2) Individual lending scheme
5. Collateral 1) Land use certificates (Green and Red Books)
2) Durable consumer goods
3) No collateral required
6. Place of credit
negotiations and
information
All necessary negotiations, credit disbursal, collecting of interest,
collecting of principal will take place at one of these levels.
1) District
2) Commune
3) Village
Source: Own figure.
Each product has an almost infinite number of attributes. Many of these attributes
do not have a measurable influence on the purchasing decision of a potential
consumer or are considered as important only by a very small market segment.
Therefore, it is neither possible nor useful to grasp all existing attributes and
their levels in market research. It is necessary to reduce the attributes and their
levels to a manageable size and to those, which are most relevant to consumers
84 A possible attribute of a credit product is the interest rate, with the possible level of 20%
interest rate per year.
Chapter 5
92
in forming their preferences. Such a reduction requires interaction with the potential
consumers to determine the most relevant attributes and their levels.
From the perspective of the target population, the attributes and their levels have
to be determined in a participatory process because this is pre-eminent for obtaining
true-to-life results in the statistical analysis. Engineers and/or economists
assigned with developing new products or services may have other priorities
than the potential customer.
The rural population, and particularly the farmers, were encouraged during open
discussions to describe their financial background and economic conditions with
the help of several participatory research tools. The qualitative data gathered
allowed to specify possible attributes of financial services that are microcredit
and microsavings services. Then relevant microfinance attributes and corresponding
levels were pre-selected. These pre-selected attributes were again presented
to the rural target group during group discussions and rankings. Here the
importance of each attribute was verified or the attribute was dropped (Table 5-1
and Table 5-2).
Table 5-2: Deposit attributes and their levels
Attributes Levels
1. Savings term 1) Interest-bearing (0.5% per month), three-month time deposit
If money is not withdrawn after this time, automatic extension for
another 3 months.
2) Interest-bearing (0.3% per month), one-month time deposit
If money is not withdrawn after this time, automatic extension for
another month.
3) No interest-bearing checking account
Withdrawal and deposit at any time.
2. Incentive 1) With a lottery scheme
Clients receive a free ticket for the monthly lottery for each 10,000
VND deposit. After withdrawals, clients have to skip three months of
lottery unless they deposit at least 10,000 VND more than they have
withdrawn. For every 50,000 VND on the account they receive one
ticket.
2) No lottery scheme
3. Place of
transaction
The savings transaction will be done at one of the following locations.
1) District
2) Commune
3) Village
4. Minimum
deposit at opening
1) 20,000 VND
2) No minimum deposit necessary
Source: Own figure.
Participatory product design
93
Orthogonal design: Even if the number of attributes and levels is reduced to the
most relevant and important ones, the number of possible concepts which have to
be assessed is usually too large to be managed effectively.85 For instance, in this
research, in the case of the credit product, four attributes with two levels and two
attributes with three levels have been identified as most relevant. This would result
in 144 possible concepts. According to BACKHAUS et al. (1996), the CA design
should not exceed 20 concepts. Therefore, a reduced design was applied. The basic
idea behind a reduced design is to create a manageable number of concepts that
represents the full design as closely as possible. The number of concepts is selected
in such a way as to permit the statistical decomposition and quantification of
each attribute level’s contribution to the consumers’ choice (RANDOLPH and
NDUNG'U 2000). The Orthogonal Main Effect Design for asymmetrical factorial
experiments has been applied to reduce the number of concepts (ADDELMAN 1962).86
An Orthogonal Main Effect Design was created with SPSS 9.0. Hence, the full
design for the credit CA was reduced from 144 concepts to 16, and in the case of
the savings CA, from 36 to nine without losing important information.
Stimuli: Typically, a CA is carried out using hypothetical descriptions of the
service, or so-called stimuli. In this context, a stimulus is defined as the presentation
of the attributes’ levels to the respondent. Data for CA experiments may
be collected by three types of stimulus presentation: (1) verbal, (2) paragraph
(descriptive cards), and (3) pictorial or in-kind presentation (GREEN and
SRINIVASAN 1978). These stimuli describe distinct concepts and will be assed by
the respondents (BACKHAUS et al. 1996). This research study follows the recommendation
of SCHRIEDER and HEIDHUES (1991) for presentation of financial
services, which constitutes a mixture between verbal, paragraph, and pictorial
design for the creation of stimuli in developing countries.
Ideas for the visualization of the levels were gained through the whole research
process and later on discussed during village workshops, which were moderated
by a PRA specialist. The visualization ideas were finally discussed with a
painter of the Tay ethnic minority. He converted the ideas, images, and figures
into pictograms (BARISCH and DUFHUES 2001; GEPPERT and DUFHUES 2003).
85 Different compositions of attributes’ levels create hypothetical product concepts.
86 An asymmetrical design is defined as a set of attributes with different numbers of levels
(BACKHAUS et al. 1996).
Chapter 5
94
The pictograms were arranged on DIN-A4 cards according to the orthogonal design
and titled with an explanatory headline in Vietnamese. A detailed description of
the participatory creation of stimuli can be found in (DUFHUES et al. 2003).
The traditional CA involves asking consumers to rank or rate in order of preference
different product alternatives. However, this research uses the so-called ‘Choice
Based Conjoint Analysis’ (CBC) approach. CBC does not involve any ranking
or rating, but simply asks customers which option they would choose or purchase.
This approach is more realistic. Another advantage of the CBC method is
the ‘none’ option. As in the real world, respondents can decline to purchase in a
CBC interview by choosing the ‘none’ option (ORME 1996).
One representative of each household was invited to participate in the CA survey.
The respondent was asked to choose the three best alternatives represented by
the stimuli-cards, or none. Furthermore, a short interview was conducted to collect
data for market segmentation (e.g. sex, age, etc.) and complementary questions
on the ideal savings/credit product (e.g. amount of savings).
5.2.2 Regional focus and sampling procedures
Two research sites were selected, namely Bac Kan (Ba Be district) and Son La
province (Yen Chau and Mai Son district). Both provinces are located in the
mountainous regions of Northern Vietnam and belong to the poorest provinces
of the country (WORLD BANK 1999).87 Ba Be district is a very remote area and
has only recently (in 1999/2000) gained access to regional and interregional
markets. Farmers produce mainly at subsistence level and a large proportion are
to be considered poor. Therefore, the area is suitable for developing clientadapted
financial products that can be used as a mechanism to prevent, mitigate,
and cope with poverty. Due to the creation of the Ba Be National Park, huge resettlements
took place and these have aggravated the problems in the region.
Hence, the main focus of this research was set on the Ba Be district.
The communes and villages were selected in accordance with pre-defined selection
criteria. These criteria are:
87 The average GDP per capita for Son La province is 206 US$ and for Bac Kan province
149 US$. These figures lie clearly under the national average of 370 US$ per capital (1998)
(STATISTICAL PUBLISHING HOUSE 1999).
Participatory product design
95
- Along the slope, to get different stages of market access, ecological zones
and ethnic minorities;
- Different phases of the land allocation process, and
- One village with a high proportion of non-farm activities.
Research in rural regions involving foreigners is sometimes perceived by the
local authorities as being troublesome for the village society. To keep the intrusion
to a minimum, the research was arranged together with the local authorities
and local research partners. For example, three communes were selected together
with the Thai Nguyen University of Agriculture and Forestry and the local authorities
in Ba Be town according to the criteria presented above. Then the leaders
of the communes identified three villages in each commune in line with the
selection criteria. From these nine villages, four were chosen for the sample. An
overview of the research sample in both regions is given in Table 5-3. The
households in Ba Be and Yen Chau district were selected using a stratified sample
according to the living standard of the population. Different wealth groups are of
particular interest for this research in order to take into account the diverse
preferences of people.
Table 5-3: Research areas and sample composition
Selected
households
Province
and
district
Commune Village Main ethnic
minority
Number of
households
per village
Number %
Dia Linh Pac Nghe 1 Tay (97%) 76 36 47
Nghien Loan Khau Nen Nung (89%) 36 19 53
Province
Ba Kan,
district
Ba Be
Xuan La Thom Meo
Khuoi Khi
Tay (93%)
Dao (100%)
84
40
43
20
51
50
Sap Vat Sai
Na Pa
Dong
Thai (85%)
Thai (100%)
Thai (100%)
80
64
48
42
33
25
53
52
52
Province
Son La,
district
Yen Chau Chieng Hac Bo Kieng Hmong (100%) 20 13 65
Chieng Pan Than
Tat Heo
Kho Mu (100%)
Thai (100%)
38
16
20
9
53
56
Interviewed households in total: 260
Source: Own figure.
5.2.3 Econometric analysis
Consumers make their consumption decisions based on a joint comparison of
different attributes. The CA assumes that a consumer assigns a utility value to
each level of each attribute and makes his or her final decision based on the total
utility values across attributes for a given product (RANDOLPH and NDUNG'U 2000).
Applied consumer research focuses on determining the contributed portion
Chapter 5
96
(part-worth utility) of each attribute level to the dependent variable (MOORE 1980).
The part-worth utility is defined as the contributed portion of various attribute
levels to the overall acceptance perceived (GREEN and SRINIVASAN 1978). The
respondent of a CA interview shows his preferences for different concepts. By
using an estimation procedure, the value of each attribute level can be calculated
from the overall preference (ALBRECHT 1997). An advantage of the technique is
that it can be used to assess hypothetical as well as existing products, and so it is
often used to evaluate new commercial products before they are put on the market,
or even before they are developed (RANDOLPH and NDUNG'U 2000). Thus,
one of the main objectives of the CA is to develop a new product/service according
to the true multi-attribute preferences for the product/service of a certain
target group (market segment). In addition, it quantifies the impact on consumer
acceptance if the demanded attribute concepts are not met (KÖCHER 1990).
The analysis of main effects regarding financial service attributes can lead to
valuable insights for designing financial products (SCHRIEDER 1996). The main
effects are investigated by using part-worth and total utilities, and the relative
importance of single attributes. A utility is a measure of relative desirability or
worth. When computing utilities using logit, every attribute level is assigned a
utility (also referred to as part-worth). The higher the utility, the more desirable
is the attribute level. Attribute levels with high utilities have a large positive impact
on influencing respondents to choose products. Just because a level receives
a negative utility value does not mean that this level is unattractive. But relatively
speaking, other levels are better. Utilities are zero-centered within the attribute
and therefore sum to zero in each attribute. The relative attractiveness of
a concept can be assessed by adding up the utilities for its component attribute
levels (total utility). Utilities cannot be compared between different attributes.
Therefore, the relative importance of each attribute is calculated to enable comparisons
between attributes.
The CBC 2.6 software was used for data analysis. The CBC 2.6 software applies
a multinomial logit analysis. Logit analysis is an iterative procedure to find the
maximum solution for fitting a multinomial logit model to the data. Within logit,
a multinomial framework is advocated over a binary one. In a binary model of
choice, the dependent variable has just two states, to choose one product or not.
A multinomial model estimates the probabilities of choosing a product from a
number of competing alternatives (HUBER 2000).
The logit analysis is evaluated by the chi-square statistics. Both the credit and
savings models have seven degrees of freedom. With seven degrees of freedom, a
chi-square of about 25 would be significant at the 0.001% level. The chi-squares
Participatory product design
97
obtained from the logit analysis, which are recorded in the Tables 5-7, A-2, and
A-3, are safely larger than this. Therefore, it is safe to say that respondent choices
are significantly affected by the attribute composition of the concepts.
5.3 Results and discussion
The following section is based on the analysis of primary data, secondary data,
and an extensive literature review. In the first part, the supply side of the rural
financial market will be examined and in the second the demand side will be the
in focus of the analysis.
5.3.1 Institutional assessment88
Vietnam’s financial system is still in a formative stage with some legislative issues
still in process. Nevertheless, after a decade of financial sector reforms,
Vietnam expects the emergence of an effective and efficient banking system
(NGUYEN-KHAC and VON GURETZKY-CORNITZ 1996; TEUFEL 1997). By 1997,
the former mono-bank system had been changed into a two-tier banking system
consisting of the SBVN as central bank and supervisory institution (tier 1), and
an operating system (tier 2) consisting of four SOCBs (one of them is the
VBARD), one non-profit, state-owned bank (the VBP)89, about 150 commercial
banks in the form of private banks, joint-stock banks, joint venture banks and
foreign banks (foreign bank branch or representative office), 947 PCFs (commune
level savings and credit cooperatives, which are supervised by the State
Bank), and one Vietnam Postal Savings Service Company (VPSC) (QUE 1997;
WORLD BANK 1999).90
88 The institutional assessment is based on the data collected in Ba Be district.
89 As mentioned before, the VBP was replaced by the VBSP. The reasons for the establishment
of the bank are the separating of policy and commercial credits. This aims at enhancing
financial transparency in the country’s banking system (SGT DAILY 2003).
90 The VPSC was established in 1999 and operates under the authority of Vietnam Post and
Telecom. Its main functions are to provide a savings product for the underserved (rural,
women, and poor) populations of Vietnam and to mobilize savings for government development
investments. VPSC offers a time savings and a collection savings (demand deposits)
and a limited money transfer system for clients. The operating cost of VSPC is de facto
subsidized by the Vietnam Post and Telecom through the use of its staff, as the VSPC itself
has only 100 staff on board. The postal savings system offers the use of a structure that is
Chapter 5
98
The formal financial institutes of the rural financial system are the VBARD,
VBP, and PCFs.91 The semi-formal sector consists of special credit programs
administered by the State Treasury and by mass organizations. The informal sector
(non-governmental organizations, moneylenders, and family or friends) covers
the demand not served by the formal and semi-formal sector. Nevertheless, the
share of the informal sector has been heavily reduced during the last ten years in
favor of the formal sector from 78% (1992/93) of all outstanding loans to 54%
(1997/98) (GSO 1995, GSO 2000).92 It is quite safe to say that this development
will continue as the interest rates of the formal sector are constantly lowered
(see below Table 5-5). The main players in the rural financial system are the
VBARD and VBP. Therefore, in the following sections the structure and performance
of these two will be discussed.
5.3.1.1 Structure of VBARD/VBP
Before the financial reform (1988-1990), the VBARD was a department of the
SBVN. Its former objective was to provide credit to state farms and cooperatives.
The share of credits to SOEs has decreased drastically in the last years and
the new objective of the VBARD is to offer financial services, and in particular
credit, to all rural households and SMEs (BAC 2001). However, the VBARD
concentrates on the better-off market segment. The branch system is split into
four levels: 1. Headquarters (Hanoi) and two representative offices (Da Nang
and Ho Chi Minh City), 2. provincial, 3. district, and 4. sub-district banks.93
The VBP was established in 1995 as the poor people’s lending outlet of the VBARD.
The VBP is a so-called policy bank, specialized in lending to poor households.
obligated to have equal geographic distribution, not access driven by profitability. Up to
now, the VPSC is offering its services only in Hanoi, Ho Chi Minh City, and Quang Ninh
Province with 710 postal savings services (WORLD BANK 2002a).
91 PCFs are not the object of this research report, as they are of minor importance due to their
little outreach (only 5% of the rural population and the fact, that they are not active in either
of the two research provinces (Bac Kan and Son La) (WOLZ 1999).
92 DUONG and IZUMIDA (2002) found that, in their survey, only 17% of the loans were borrowed
from informal sources. MCCARTY (2001) mentioned as one important factor for the
reduction of the informal sector the ‘crowding out’ by the VBARD and VBP, by which
both extended their outreach in the last years.
93 The sub-district level includes the so-called ‘mobile car branches’. At the moment, 240 mobile
branches are active (BAC 2001).
Participatory product design
99
The VBP uses what is called ex-ante targeting. Only a certain part of the population
is eligible for a loan, namely the rural poor.94 The Vietnamese government
classifies once a year every household according to its living standard into one
of five classes: Hungry, poor, medium, better-off, and rich (for details see
DUFHUES et al. (2002) and GEPPERT and DUFHUES (2003). The target group of
the VBP is the rural poor. Households from other wealth classes are officially
excluded. The purpose of the VBP is not to maximize profit but to reduce poverty
(HANH 1999; VBP 1999b).95
The VBP basically consists of a head office and it has no own structures below
this level. For example, the vice-head of the VBARD branch is at the same time
the head of the VBP branch. The VBP uses the operational facilities and staff of
the VBARD and of mass organizations at commune and village level in extending
its services to the target group. From the monthly interest rate charged to
the clients, 0.1% is paid to the local mass organizations and 0.25% to the
VBARD for its services.96 Currently, the average operational costs of the VBP,
expressed as interest rate spread, is 0.45% per month (HANH 2001). In 2000, the
monthly interest rate charged by the VBP was 0.6%. This means that 75% of the
interest revenues are used to cover operational costs.97
The poverty focus of the VBP and its high operational costs do not mean that the
bank consequently operates at a loss. Due to the highly subsidized interest rates,
however, many international agencies consider VBP not to be financially sustainable.
98 Vietnamese policy-makers have recognized this statement and declare
the subsidized interest rate policy as a temporary strategy. In the long term,
market rates should be implemented (VBP 1999b). However, this objective has
94 The successor of the VBP, the VBSP, has broadened its targeting criteria. Now, not only
‘the poor’ are eligible for a loan, but also other disadvantaged groups, e.g. pupils, students,
unemployed persons seeking jobs (even abroad), individuals and organizations in remote
areas (VIETNAM ECONOMY 2003).
95 The VBSP continues the policy of non-profit making. It is explicitly responsible for granting
the preferential loans to underprivileged groups (VIETNAM ECONOMY 2003).
96 The head of the district branch stated that the payment from the VBP for the VBARD services
does not cover the costs (CHAN 2001).
97 The VBSP will establish its own branch network independently from the VBARD with 80
district branches across the country (SGT DAILY 2003).
98 The VBP now is recognized as loss-making (VBARD and DANIDA 1999).
Chapter 5
100
not yet been achieved (see below Table 5-5). It looks rather like a fictitious
statement to satisfy the international donor community with their requirements
for financial sustainability and thus attracting more funds.99
Outreach: The VBARD is the biggest supplier of rural credit. In 2001, 4.5 million
households were given credit. This figure represents 38% of all rural households.
However, the share of VBARD’s outstanding credit to mountainous provinces is
significantly lower than to rural areas (CECARDE 1999).100 As the VBARD
targets the better-off clientele in the rural population, the bank has been able to
meet only a small proportion of the funds demanded by very poor households in
rural areas, particular in Northern Vietnam (DUONG and IZUMIDA 2002;
VIETNAM-CANADA RURAL FINANCE OUTREACH PROJECT 1999a). In South Vietnam,
the share of poor people with loans from the VBARD is higher, as the
VBP is less active there (IZUMIDA and DUONG 2001). In Northern Vietnam the
VBP is covering this market segment. In total, the VBP reaches 21% of all rural
households (HANH 2001). Both VBP and VBARD have enormous outreach.
This enormous outreach is reflected by the research sample in Ba Be district. In
total, 57% of all households have had an effective demand for formal credit.
About 30% of the households were not interested in taking a formal credit. Only
15% of the households in the sample were access-constrained. Keeping in mind
that this figure also includes non-creditworthy households the number of constrained
but creditworthy households would be even smaller.
Staff: The efficiency of the VBARD has been increased over the last decade. In
1990, 30,000 people worked for the VBARD. From then, the number of staff
was steadily reduced to 24,000 people in 2000 (VBARD 2000a). At the same
time, the branch system was extended from fewer that 1,300 branches at the end
of 1993 to 1,568 branches in 2000 (BAC 2001; SCHENK 1998; VBARD 2000a).
Along with the increase in efficiency, however, an overload of work due to
99 The law on credit institutions approved by the National Assembly in 1997 stipulates that
the state should establish banks that operate on a non-profit basis (Article 10). This implies
that the state will continue to provide cheap credit to rural areas and the poor (MINOT and
GOLETTI 2001). Besides, the newly established VBSP continues to provide soft-loans to
special target groups (VIET NAM NEWS 2003).
100 For instance, in 1998 the mountainous provinces received on average 150 billion VND
compared to Hanoi, which received 1,200 billion VND.
Participatory product design
101
numerous/transactions was observed in most of the district branches (VIETNAMCANADA
RURAL FINANCE OUTREACH PROJECT 1999a). This was confirmed by
the research undertaken in Ba Be district. For instance, it was found that some
credit officers had to administer more than 1,000 households (CAT 2001). It has
been realized that one credit officer, if he exercises all his duties including receiving
clients, checking project proposals, disbursing money, checking loan
usage, collecting principal and interest, can take care of 500 families as a maximum
(CECARDE 1999). Certainly, if the credit officer serves more than this, he
cannot fulfill all his duties. Nevertheless, it is vital that the credit officers visit
all clients in their home to check their creditworthiness and to prevent heavy
debts. This is not possible in case of VBARD/VBP. The VBARD/VBP has
therefore delegated many responsibilities to local authorities (heads of the communes).
In fact, the credit officer relies totally on the commune and village officials
to assess the creditworthiness of the potential borrower.
Both banks apply a group lending scheme to shift responsibilities of screening
and monitoring to the clients themselves. However, the principle of peer pressure
does not work in the applied group lending scheme (see Credits below).
After the disbursement of the credit to the group members, the credit officer may
verify with the CGL whether the group members use the credit as stated in the
loan application. Normally he trusts the report of the CGL (giving him room for
opportunistic behavior) and does not monitor the credit directly. Basically, no
bank staff, but persons outside the bank are involved in the monitoring: The
head of the commune and CGL. During discussions with the commune heads
and the CGL, however, they admitted that this is not a regular activity. Officially,
the loan ought to be repaid immediately if it is not used as stated in the
credit contract. Nevertheless, farmers pointed out that as long as they pay the
interest and the principal in time, the credit officer and the other local authorities
do not bother about the actual use of the credit (DUFHUES et al. 2002).
Local information sources in the form of key informants can be an important
low-cost mechanism to reduce screening and enforcement costs. Too much dependence
by a bank on local networks as sources of information and recommendation,
however, can have negative effects in terms of social exclusion of
those households not included in the social and political network of the territory
(VAESSEN 2001). In general, important information tends to be segmented
and to circulate within specific groups or networks to the exclusion of others
(ROBINSON 2001). Particularly the very poor households often find themselves
in this position, as they are socially excluded and lack access to fruitful relationships
with powerful allies (HICKSON 2001).
Chapter 5
102
Mobilization of funds: The financial sources of the VBARD consist mainly of hot
money.101 For instance, in the year 2000 over 95% of the sources were mobilized
through savings. The insufficiency of deposit mobilization by formal institutions at
the grass-roots level stands in contrast to the successful development of deposit
mobilization at national level. The majority funds are mobilized in urban areas. In
Vietnam, money is flowing from urban to rural areas (HUNG and GIAP 1999;
IZUMIDA and DUONG 2001). The VBARD offers the only formal savings scheme
available to households in rural areas in Northern Vietnam. However, even
VBARD deposits are overwhelmingly urban, with only a small portion coming
from rural households (MCCARTY 2001). Therefore, most branches in rural and
mountainous areas have to borrow from the head office. Regulations from the
head office do not motivate district branches to mobilize funds (for details see
HUNG and GIAP 1999; and VIETNAM-CANADA RURAL FINANCE OUTREACH PROJECT
1999a). Moreover, these regulations make branches become used to depending
on head office funds rather than competing for funds and improving efficiency
in fund mobilization. However, the VBARD branch in Ba Be district seems to
rely totally on hot money (BAO 2001). But only one-quarter of this money is
based on savings from the population. The rest of the money originates from
savings from other institutes, mainly the State Treasury (60%).
The sources of the VBP consist purely of cold money. One of the major reasons
for establishing the VBP was to attract international donor funds (VBP 1999b).
However, the VBP failed to attain this objective, and international donors provide
only small amounts. The government is the main supplier of funds to the
VBP. If the government does not have sufficient funds to finance the soft loans
of the VBP, the SBVN must refinance the VBP without charging interest on a
long-term basis and without a pre-fixed repayment date (NHGIA 2001). The procedure
for distributing the funds from the national to the lower levels is rather
cumbersome. For its credit allocation process, the VBP relies on information
from several public organizations and subordinated branches at the district and
provincial levels, and then distributes the funds on that basis (for a detailed description
see DUFHUES et al. (2002) and Theesfeld (THEESFELD 2000).
101 Hot money is defined as savings. Cold money is defined as grants and subsidies.
Participatory product design
103
Credits: Lending scheme: Most of the VBARD loans follow an individual lending
scheme and only on a very limited scale group lending schemes. In contrast,
the VBP uses only group lending schemes (so-called ‘joint liability groups’)
(VBP 2001b).102 However, in day-to-day practice, the group members are not
held liable for each other. If one group member defaults, the only consequence
for the other members is that this particular group does not receive credit any
more. But individual members who repaid on time can join a new group. This
means that it is enough to expel only the defaulter from the group in order to
obtain a new credit, as this is officially a new group. Therefore, joint liability
does not exist and the concept of peer pressure does not work. Each group has a
group leader. He has an obligation to support his group members in all matters
related to the loan (a detailed description about the obligations of a CGL can be
found in DUFHUES et al. 2002). The VBARD/VBP regulations points out that the
CGL is also liable (VBP 2001b). He has to repay the loan in the event of default
by a single borrower. However, this regulation is also not practiced in day-today
business.
Loan Terms: Most loans from the VBARD were short-term production loans
until around 1996. Recently, however, the situation, has changed quite significantly,
with a larger proportion of long-term loans (IZUMIDA and DUONG 2001).
Table 5-4 shows the development of the loan structure in the Ba Be district. The
demand for short-term loans is decreasing, while demand for long-term loans is
increasing year by year. This might be by explained by the fact that the VBARD
has eased the criteria for long-term credits and that very popular investments are
e.g. husbandry of large ruminants or tree cultivation. For an investment such as
these, one year is too short to gain any profit. The government of Vietnam recognized
this development and recently gave permission to the VBARD to use 30%
of its short-term deposits as medium and long-term loans for farmers to provide
more funds for long-term investments.
The VBP uses predefined loan products for all households. In March 1999, the
maximum loan size for a loan from the VBP had been set at three million VND
and a maximum term of three years. In 2001, the loan size was raised to five
102 The group lending scheme of VBARD/VBP was supposed to be a copy of the Grameen
Bank group lending scheme.
Chapter 5
104
million VND and the term to five years.103 In comparison to the VBP, the
VBARD determines the loan amount by the value of the collateral. Short-term
loans are usually not requested (see Table 5-4).
Table 5-4: Loan structure of VBARD/VBP in Ba Be district
1998 1999 2000 2001
Amount % Amount % Amount % Amount %
VBARD
short-term
826 19 722 16 779 10 698 6
VBARD
long-term
3,562 81 3,801 84 7,240 90 10,316 94
VBP short-term 121 8 50 2 0 0 0 0
Long-term 1,313 92 2,291 98 3,471 100 3,326 100
Source: VBARD (1998, 1999, 2000b, 2001a) and VBP (1998a, 1999a, 2000, 2001a).
Notes: Figures are in VND millions. Classification of loan term: Short-term: ≤ one year; long-term:
> one year (BAC 2001).
Interest rates: The Vietnamese Government sets an interest rate ceiling through
the SBVN to provide the population with soft loans. After financial reforms at
the beginning of the nineties, the real interest rate became positive in 1992
(SENANAYAKE and HO 2001). Since then, however, the nominal interest rates, particularly
those of VBARD and VBP, have been gradually reduced (see Table 5-5).
Interestingly, apart from the nominal price, the credit conditions have remained
more or less the same over the years. Over this period, the effective demand increased
continuously. The demand was satisfied through the Vietnamese government
by supplying additional funds to VBARD and VBP for lending on. Apart
from two years, 2000 and 2001, when Vietnam experienced deflation, the real
credit interest rate also declined (see Figure 5-1). The increasing credit demand
is certainly positively correlated with the declining price, but also with the improvement
of the transport infrastructure, which allowed the credit officers to
reach more remote villages that were formerly not serviced.
Nevertheless, commercial banks such as VBRAD are allowed to adjust their
credit interest rates within a range of 0.25% for short-term and 0.3% for medium-
and long-term loans (NGUYEN 2001). However, the VBARD has to give a
103 The VBSP will provide short- and medium-term loans only. Long-term loans will be con
sidered later (VIETNAM ECONOMIC TIMES 2002). A single borrower will be entitled to get a
maximum loan of ten to fifteen million (SGT DAILY 2003).
Participatory product design
105
discount of 15% on the interest rate for farmers who are living in a poor commune
and 30% for those living in a very poor commune. The interest rate payments are
collected on a monthly basis, or every three or six months (depending on negotiations
with the credit officer) and the principal is usually paid at the end of the term.
Figure 5-1: Real credit interest rate of VBP in Ba Be district, 1998-2003
-4
-2
0
2
4
6
8
10
12
14
1998 1999 2000 2001 2002 2003
Inflation rate*
Nominal interest rate p.a.
Real interest rate p.a.
Source: *ADB (2001), CIA (2002), and EIU (2002).
Notes: The monthly interest rate was lowered in 2001 from 0.6% to 0.5%. Therefore, the average
interest rate of 0.55% was used for the calculation of the real interest rate. It was assumed
that the monthly interest rate of 2001 (0.5%) will stay the same in 2002 and 2003. The estimated
inflation rate for the years 2002 and 2003 is based on EIU (2002).
Table 5-5: Nominal interest rates per month of VBARD/VBP in Ba Be district
June 1997 1998 1999 2000 2001
VBARD Short-term 1.25% 1.2% 1.2% 1.1% 1.0%
Long-term 1.05% 1.0% 1.0% 0.9% 0.8%
VBP 1.20% 1.0% 0.8% 0.6% 0.6/0.5%
Source: VBARD (1997, 1998, 1999, 2000b, 2001a) and VBP (1997, 1998a, 1999a, 2000, 2001a).
Notes: Classification of VBARD, short-term: ≤ one year; long-term: > one year (BAC 2001).
Loan use: Households desire to use funds for various purposes while formal financial
institutions only finance a number of specific purposes. In former times the
VBARD provided only financing for agricultural production, but little or none to
other production activities or services (VIETNAM-CANADA RURAL FINANCE
OUTREACH PROJECT 1999a). The fungibility of money has been noticed by the
VBARD’s headquarters and the formerly strict rule has been attenuated recently so
that nowadays trading and small-scale processing will also be financed (VBARD
2001b). Rural households are now allowed to spend their loan for any purpose,
even for consumption (IZUMIDA and DUONG 2001). Due to the close connections
between VBARD and VBP, it is very likely that the headquarters of the VBP will
Chapter 5
106
soon adapt its policies, too. However, this new development has not yet reached the
Ba Be district. The VBP still finances only a limited range of ‘obvious’ or ‘fashionable’
purposes related to agricultural production (e.g. rice cultivation, pig raising).
There is not much room for innovative or idiosyncratic proposals (UN 1998).
Collateral: The VBARD requires collateral for its loan. Usually land use certificates,
‘Green Books’ or ‘Red Books’ are taken as collateral.104 The collateral often
does not reflect the real value as no differentiation is made in the assessment of the
land. Different qualities of land are given the same value (e.g. paddy rice or forest).
During the loan term, the land use certificates are kept in the bank, until the principle
has been repaid (BAC 2001, CAT 2001). The land use certificate can bear only
one loan, even if the loan amount is less than the value of the certificate.
Other accepted forms of collateral are government wages and houses.105 The collateral,
except land use certificates, must be notarized at the certification office in
Ba Be town if the value exceeds ten million. Below ten million an approval from the
commune is sufficient (CAT 2001). If the loan exceeds 50 million, the collateral must
be notarized in the province capital (BAO 2001). Only part of the value of the collateral
will be granted as a loan. For instance, in the Dia Linh commune 70% of the value of
the collateral will be granted, while in the other communes it is only 60% (DOAN 2001).
Although, the government states that farm households can take out loans of less
than ten million without any collateral, the VBARD still requires certificates of land
use certificates and guarantees from local authorities as loan security. Therefore,
these households without certificates have difficulties in accessing formal loans from
the VBARD (MCCARTY 2001). However, the ongoing dissemination of ‘Red Books’
in recent years brought an increasing number of households into possession of assets
that are useable as collateral, which broadened the possible outreach dramatically.
Nevertheless, the land market is still underdeveloped. Only few households in fact
sell or buy land and usually it is traded within the village. Therefore it is difficult
104 and is owned by the state in Vietnam. Nevertheless, in 1993 the renewal of the law was
completed and since then, the government allocates land use certificates to farm
households, the so-called ‘Red Books’ for agricultural land (valid for 20 years) and, since
1999, ‘Green Books’ (valid for 30-50 years) for forest land. Farmers are allowed to sell,
rent, or pass land on to children (LUIBRAND 2002).
105 The VBARD recently announced that it will also accept assets other than land use certificates,
houses, and wages to improve their outreach to SMEs.
Participatory product design
107
for the banks to liquidate land. Often, the only possibility would be the expulsion
of the farmers from the land before liquidation (BAC 2001). At the moment the
VBARD uses only psychological pressure to the farmers relating to the possibility
of losing their land. None of the villagers or key persons interviewed knows of
any case of land liquidation in this area. Therefore, there might be the danger of a
moral hazard. If farmers find out that the VBARD is not going to liquidate their
land in the event of default, the bank might end up in a landslide of bad debts.
The VBP does not require any physical collateral for its loans. However, the approval
of the head of the commune can be considered as social collateral according
to DUFHUES et al. (2002). Besides, anecdotal evidence from the village survey
showed that in some cases of ‘hungry’ households, which are officially excluded
as they are assessed as too poor, the credit officers insist on collateral in
the form of a ‘Red Book’ for a VBP loan too. However, this is in contrast to the
national policy of the VBP, namely to provide collateral-free loans.106
Repayment rate: The official repayment rate is quite impressive. In the year 2000,
only 1.1% of the loans were not performing (VBARD 2001b). Nevertheless, in
this case the repayment rate is an inadequate indicator of good performance, as
the VBARD has so-called ‘frozen debts’, which are not counted as ‘not performing’
or ‘overdue’ as the principal is guaranteed by the SBVN. These debts
will not bear any further interest.107 The loan recovery is generally overestimated.
Besides, the arrears of the VBARD do not contain rescheduled loans.
The details of rescheduling loans are not clear in annual reports or in interviews
with VBARD staff. A loan can be rescheduled several times but not beyond one
production cycle for short-term loans or 36 months for medium/long-term loans
(IZUMIDA and DUONG 2001). As the liquidation of collateral is not possible,
rescheduling of the loan is often the only possibility for the credit officer to
avoid designating a loan as overdue or not performing.108
106 The VBSP will continue to provide collateral free loans to certain target groups (VIETNAM
ECONOMY 2003).
107 In the event of a natural calamity or something unavoidable, the SBVN admit some loans
from VBARD as frozen debts.
108 The new VBSP will take over most of the non-performing loans made by the VBARD to
poor households. This of course will serve the VBARD very well but the VBSP has to
carry a heavy burden right from the start (VIR 2003).
Chapter 5
108
The VBP too has a very high repayment performance of 98% (HANH 2001). But
again this is only a poor measurement of good performance, as rescheduling of loans
in VBP too is extremely high; it is reported to be as high as 70% in some provinces
(VBARD and DANIDA 1999). As the staff of VBARD and VBP is identical, it is not
surprising that the instrument of rescheduling is also heavily used within the VBP.
Procedure: The procedure for obtaining an individual loan from the VBARD is
described in Figure 5-2. The credit officer supplies the farmer with the application
form. But it is not uncommon that the credit officer does not have the form
with him. This means that the farmer has to go to the district or wait until the
next visit of the credit officer to get the form.
In general, TCs such as credit fees, cost for forms, opportunity cost of time, etc.,
were not seen as an access barrier by the farmers themselves. Nevertheless, the
striking disadvantage of this credit product is the high transportation costs due to
the frequent journeys required (at least three) to the district capital. For the more
remote households, this is a very strong market entry barrier. Particularly, the
first two visits were mentioned by farmers as representing access barriers, as the
farmers do not know at that time whether the loan will be granted or not. This
problem could be solved by implementing so-called profit centers and through a
real decentralization of the decision-making structures.
The directors authorize all small loans as recommended by the credit officers. In
essence, credit officers, though not authorized to impose any limit, are authorizing
small loans. This impression was confirmed by research work done within the
VIETNAM-CANADA RURAL FINANCE OUTREACH PROJECT (1999a). However, this
practice is still far from being a decentralized profit center as implemented for
example by the Bank Rakyat Indonesia (BRI) (SEIBEL and SCHMIDT 2000).109 This
form of decentralization is called ‘deconcentration’, and is predominant in Vietnam
(GEPPERT et al. 2002; HIEN 1999). Within the deconcentration process, the
authority remains with the central agency.
The procedure of the VBP is very similar to the procedure for VBARD loans, but
most of the steps are carried out by the CGL. Therefore, the TCs for the single borrower
are much lower. Nevertheless, as it is a group lending scheme, the whole procedure is
109 Profit center is defined as a semi-autonomous, independently accounting corporate unit
responsible for its own operations, profits, and losses (TOFFLER 1990).
Participatory product design
109
much more time-consuming (for details see DUFHUES et al. 2002). Recent credit policy
developments, however, show that the credit products of the VBARD and VBP are
becoming more and more similar, and so their market segments are converging. For
instance, the VBARD now tries to offer a credit product that does not require collateral
for loans of less than ten million VND (VBARD 2001b). At the same time, the VBP
has broadened its credit term to five years and raised the maximum amount to five
million VND (VBP 2001b).110 It is not clear to potential customers whether this credit
is a VBP product or whether it is a VBARD product, offered by the VBP. Some
farmers even switched from the VBARD to the VBP because of its lower interest rate,
despite having collateral and not belonging to the eligible target group of the VBP.
This marketing strategy can affect the sustainability of both institutions.
Figure 5-2: The six-step procedure to obtain an individual loan from the
VBARD in Ba Be district
1. The farmer contacts the credit officer in the district or in the commune to discuss with
the credit officer the purpose of the loan, the amount, the profitability of the investment,
and further steps (if farmers can not calculate the profitability of the investment, the
credit officer helps to do so); then the farmer receives the application form.
2. The farmer fills in the form at home, because the wife and any children over 18 in
the household have to sign as well.
3. The farmer has to take the form to the commune to get a signature and a stamp from
the head of the commune.
4. The farmer brings the form and the land use certificate to the bank. (The credit officer
visits the household to check if everything is correct and arrange an appointment at the
bank (usually Tuesdays and Wednesdays). He mainly checks the collateral. Many farmers
stated that the credit officer never visited their households, as they have to give him
the land use certificate to keep it in the bank, he has no need to visit the household.)
Internal steps of application: a) The credit officer signs the application.
b) The head of credit office signs the application.
c) The director signs the application.
5. The farmer has to come to the bank to ask whether his application was approved or
not; if yes, he agrees a date with the credit officer for the disbursal of the money.
6. The farmer comes on the arranged date and receives the money disbursed.
Source: Own figure.
Notes: 1 The signature also confirms that the applicant lives in the area. The head of the commune
may refuse the signature if he knows of a disagreement between the married couple as it relates
to the credit application, in cases of drug abuse of the applicant, or so-called social evil in the
households.
110 The new VBSP adjusted its credit scheme even further to the credit scheme of the VBARD
(see Footnote 103).
Chapter 5
110
5.3.1.2 Supply of savings products
The VBARD’s head office offers different savings schemes: A demand deposit
scheme, which is mainly used by companies and term deposit accounts with
three, nine, twelve, and more than twelve-month terms. Bonds are available for
two, three or five years. However, long-term deposits are not popular and are
rarely demanded (BAC 2001).
Savings accounts at VBARD’s branches are traditional, simple and not attractive
enough to customers. Branches offer only a very limited choice of savings
products, for instance the branch in Ba Be district offers only demand deposits
(0.15% per month) and three month (0.25% per month), six month (0.4% per
month), and twelve/thirteen month (0.45% per month) deposits, while customers’
need a more diverse supply. There is no linkage between credit and savings in
branches’ and credit officers’ operations. For example, loans are disbursed directly
in the communes (as in the case of VBP) and credit officers also visit the villages,
but to save money, customers have to go personally to the branch offices.
For remote households this is an invincible market entry barrier, as transport
costs to the branch office are often much higher than the amount saved
(VIETNAM-CANADA RURAL FINANCE OUTREACH PROJECT 1999a). Another barrier,
particularly for poorer households, is the minimum deposit of 50,000 VND.
Many poor households are not able to save this amount at one time. Despite the
fact that the interest rate on deposits has been sinking year on year, the numbers
of savers and the amount saved have increased in the last four years (Table 5-6).
Nevertheless, as discussed above, almost 100% of the savings are of urban origin.
Savings mobilization plays no role in the range of services of the VBP.
As a reason for the lack of formal savings schemes in rural areas, the high cost
of deposit collection is often mentioned. Rural households are scattered geographically
and their potential deposits are very small. In developing countries
the bad infrastructure aggravates the situation. Nevertheless, it is now widely
agreed that microsavings can be collected in a profitable and sustainable way
(for discussion see ROBINSON 2001). However, in Vietnam the cost argument
did, if at all, influence only slightly the decision of not implementing a rural saving
scheme. It is the national consensus, adopted by the staff of the VBARD/VBP,
Participatory product design
111
that poor households are not capable of saving (CAT 2001; CHAN 2001; HANH
2001; HUNG and GIAP 1999).111 Many non-governmental organizations and research
projects in Vietnam found evidence to the contrary. The negative attitude
towards savings is one of the main obstacles for improving savings products in
order to achieve wider outreach to the rural population.
Table 5-6: Savings products at VBARD in Ba Be district
1998 1999 2000 2001
Amount Saver Amount Saver Amount Saver Amount Saver
Demand
deposit
/ / 336 171 206 146 278 152
3-month
deposit
148 23 94 19 189 21 57 16
6-month
deposits
1,374 116 2,732 215 2,962 247 4,599 379
13-month
deposit
1,978 133 / / / / / /
Source: VBARD (1998, 1999, 2000b, 2001a).
Notes: Amount in VND millions. This table includes only savings of the population and SMEs,
excluding public agencies.
5.3.2 Potential demand
This section empirically applies CA to identify the specific demand for financial
services and to formulate market and client-oriented financial policies. First the
potential demand for credit services then saving services will be discussed. In
the following the households are segmented by wealth (three classes) and by
gender.
5.3.2.1 Potential demand for credit
Without any market segmentation, the credit attributes are assessed with similar
relative importance. However, within the wealth segmentation the relative importance
of attributes differs greatly. The indigent and medium households valued
the attribute ‘disbursement time’ as most important (see Table A-2 in the Annex).
Both groups have a high preference for quick disbursal of loans. This finding is
supported by KANBUR and SQUIRE (2001), who state that poor households often
111 In all interviews with VBARD/VBP staff, the costs of saving collection were never mentioned
as an obstacle for the implementation of a rural saving scheme.
Chapter 5
112
use quickly accessible credits as a tool for income and consumption-smoothing
and to cope with external shocks. This explains why poorer households have a
stronger preference for quick loans than rich households.
All groups value the livestock insurance positively. However, it plays a minor
role in the decision process relating to taking a credit, except for indigent households.
The existence of livestock insurance is the second important parameter in
their decision process relating to taking a credit. Yet, the differences between the
attributes in this class are rather small. Nevertheless, poor households are well
known for their risk-averse behavior.112
Surprisingly, medium households prefer the district to the commune (the district
is more remote from the rural household’s perspective), and indigent households
value commune and district almost equally as the place for the credit transactions.
This might be explained by the fact that the regional market also takes
place in the district, which gives the district an additional attraction, especially
for the poor who seldom leave the village. However, for indigent and particularly
for medium households it seems quite important (second highest relative
importance) to conduct all credit transactions in the village. In comparison, rich
households prefer the commune as the transaction place. This result is not significant,
but it is supported by qualitative research. Members of rich households
are often working at the commune or have family members or close friends who
do so, and therefore have close contacts with commune officials and especially
to the head of the commune. The head of the commune has a prominent position
in the local rural credit market of Northern Vietnam (DUFHUES et al. 2002).
Therefore, it may be that those households expect to take advantage of this relationship.
Rich households assess the attribute ‘collateral’ as most important, and within
this attribute, the level ‘no collateral required’. This is surprising, as rich households
are considered to be in possession of sufficient collateral. Although this is not
significant, one explanation for this outcome might be that rich households are
more afraid of losing their collateral in the event of default than other households.
112 Over 80% of all households would take out livestock insurance even without a credit.
Particularly the poorer households demanded this service. Therefore, the implementing of
such a livestock insurance scheme would probably directly benefit the poor.
Participatory product design
113
This seems to be particularly true for ‘durable consumer goods’. These goods
might be assessed by local authorities and bank staff as luxury goods and are
probably easier to seize than land use certificates.113 Therefore, the rich value
‘durable consumer goods’ very negatively. In comparison, poorer households do
not seem to care much about collateral. The relative importance of the attribute
is only 17%. It is striking here that they place land use certificates on the same
level as ‘no collateral’.114 Indigent farmers are not afraid to give their land use
certificates as collateral. They may be convinced that their investment will not
fail or, more likely, they believe that the bank will not seize the land. In the
event of loan default, they expect to be helped by the bank or the government.
Usually this means that the credit officer will extend the loan, or in the event of
a natural disaster, the government will step in and freeze any debt obligations.
The finding is confirmed by DUONG and IZUMIDA (2002), who state that in rural
Vietnam only few cases are known where land has been sold because of a farmer’s
default.
Men and women did not show great differences in their preference for credit
products. Therefore, we refrained from showing the detailed table here with all
utility figures and t-values. Only a slight difference was found within the attribute
‘lending scheme’. Both sexes favored the individual lending scheme. However,
male respondents preferred the individual lending scheme to a greater extent
than women (15% to 11% relative importance). This might be explained by the
fact that women are usually less educated and more reluctant to deal directly
with local officials or bank staff. Therefore on average they are more loath to the
individual lending scheme than men.
The share of households who chose the ‘none’ option (indicating no credit demand)
was highest in the case of the rich and second in the case of the indigent. Rich
households do possess a certain self-financing capacity and indigent households
often lack investment ideas due to a low educational level. Yet, when the indigent
113 Land markets are still in a rudimentary stage in Vietnam.
114 88% of the surveyed households in Ba Be district have Green Books or Red books. The
average in the three research communes is slightly lower, at 70%, but still high. However,
not every household possesses a land use certificate. The share of households with certificates
varies greatly between the villages. In some villages, not a single household has a
certificate and in others all of them do. Villages with better access to the infrastructure
Chapter 5
114
group is separated into poor and hungry households, it shows that only 5% of
the hungry households chose the ‘none option’. The poorer the households, the
greater the share of households who want to use a credit. It is likely that some of
these households are so pressed for funds that they would take any credit, no
matter what the credit terms or features are. The numbers of male respondents
who picked the ‘none option’ was almost double compared to the number of
females. Therefore, it seems that women do have a much greater demand for
credit than men. MCCARTY (2001) points out that the vast majority of rural loans
are given to men. Therefore women seem to have an unsatisfied demand for
credit products. Nevertheless, in general the demand for credit is still enormous
due to the low interest rates set by the government.
The assumption that certain attributes of a credit product could compensate for a
higher interest rate is not confirmed in this research. Almost 100% of the respondents
choose the cheaper credit. This strong focus on interest rates might be
explained by the fact that rural Vietnam has gone through a decade of continuous
reduction of interest rates and a history of public pronouncements on how important
low interests are for rural development and the improvement of rural living.
Besides, farmers often have the impression that the VBP particularly is not a
bank but an institution to help the poor. Nevertheless, farmers are willing to pay
extra fees for special services, as the insurance attribute has proven.
5.3.2.2 Potential demand for savings
With regard to developing savings products for the rural poor, emphasis ought
to be placed on liquidity and low TCs (ZELLER 1999). Nevertheless, a monetary
incentive to save is also important. The results from the logit modeling
show that households in all wealth groups give special emphasis to a high interest
rate for savings (Table A-3 in the Annex). Corresponding to economic
theory (time preference rate), this tendency is more distinct in poor and hungry
households. Men clearly put a high interest rate above everything else. Women
too are in favor of a high interest rate, but they rate a sight deposit account
without any interest more highly than a one-month time deposit with a 0.3%
interest rate. A possible interpretation for this outcome is that women are more
often confronted directly with family emergencies and therefore prefer an
have a greater probability of possessing land use certificates.
Participatory product design
115
always-accessible sight deposit account providing access to the deposit immediately
in case of sudden need.
As saving is a much more regular activity than obtaining a credit, it is not surprising
that the level ‘saving in the village’ is valued highly by the indigent
households. Poor households seldom leave the village and the small amounts they
intend to save are easily eaten up by travel costs. OWENS and WISNIWSKI (1999)
pointed out that close physical contact is essential to reach the poor with savings
products. Rich and medium households favor the commune as the place to save,
but this result is not significant. Nevertheless, some farmers mentioned that they
prefer saving in the commune or in the district because they would not trust
people collecting savings in the village, even if this were a bank employee.
Farmers mentioned a case where the credit officer collected the interest rate
payments in the village and gambled it away on the way back to the bank. It was
not clear whether farmers had to bear the loss or the bank. Although cases like
this are very rare, they create an environment of distrust. Trust in the bank is essential
for attracting savings. LEDGERWOOD (1999) states that MFIs providing
credit services must select borrowers whom they trust to repay the loan. When
collecting savings, however, it is the customer who must trust the MFI.
Lottery linked deposit accounts have proved to be successful to attract savings in
many countries (GUILLEN and TSCHOEGL 2002). The lottery scheme was also
analyzed for different wealth groups. In comparison to indigent households, medium
and rich households accept the lottery as an incentive to save. However,
these results are not significant. Indigent households are very indifferent about
the lottery incentive. The lottery itself is assessed as positive but the attribute
plays almost no role in the decision process whether to save or not. All farmers
understood the lottery, but some had difficulties to comprehend that this is a
kind of incentive. This idea was very unfamiliar to them. However, the riskaverse
behavior of poor households might deliver another explanation, namely
that winning a prize in a lottery is not secure, but interest rates are. Some poor
farmers mentioned that they would never win the prize. It is surprising that
women favor the existence of a lottery more than men. Usually women are more
risk-averse than men. However, when considering only indigent women, the
utility of the lottery became negative. But this result is not significant.
From rich to indigent households, an increase in choosing the ‘none’ option
(indicating no interest in savings schemes) was observed. Rich households have
more money to save in the bank. Besides, members of rich households are used
to dealing with local officials and therefore can better assess their trustworthiness.
Chapter 5
116
Women strongly prefer to save in the village as compared to men (Table 5-7).
RANDOLPH and NDUNG'U (2000) state that TCs may vary by gender, e.g., a
woman farmer with reproductive responsibilities may face higher opportunity
costs of time when leaving the village to seek any kind of services than a male
farmer. Women are responsible for many tasks in the household and on the
farm. It is much more difficult for them than for men to reallocate time towards
other activities. Therefore, WORLD BANK and DFID (1999) concluded in a study
about poverty among ethnic minorities in Northern Vietnam, that any kind of
policy intervention must consider women’s high opportunity costs of time and
their tight time schedules.
Table 5-7: Logit estimation of average utility values for saving attributes,
by sex
Male
(189 respondents)
Female
(68 respondents)
Utilities t-values Utilities t-values
Interest rate and term
No interest rate / demand deposit -2.247 -3.130*** -0.217 -0.360
0.3% Per month for a one-month deposit 0.110 0.225 -1.167 -1.745*
0.5% Per month for a three-month
deposit
2.137 5.791*** 1.384 4.502***
Relative importance in % 50% 24%
Incentive scheme
No lottery -0.446 -1.110 -0.944 -2.119**
Lottery 0.446 1.110 0.944 2.119**
Relative importance in % 10% 18%
Location of depositing and withdrawing
Village 1.447 6.026*** 1.947 3.966***
Commune -0.517 -1.270 0.430 0.817
District -0.929 -3.265*** -2.377 -2.963***
Relative importance in % 27% 41%
Minimum requirement for account
opening
No minimum requirement 0.518 2.114** 0.933 2.248**
20,000 VND -0.518 -2.114** -0.933 -2.248**
Relative importance in % 12% 18%
None option 2.138 5.086*** 1.937 3.088***
Percentage of households choosing none
option
13% 11%
Chi-square 350.894 117.704
Source: Own calculations.
Notes: * Significant at 10% level. ** Significant at 5% level. *** Significant at 1% level.
Participatory product design
117
5.4 Conclusions and policy recommendations
The CA provided some valuable insights into how to improve the financial services
of FFIs to the population by adapting the existing services. The results of
this research show that the VBP particularly can improve its credit services by
offering individual loans, as this kind of lending scheme was strongly preferred
by the sample population. This approach would also reduce the administration
time of the credit procedure, i.e. TCs.115 It could also go hand in hand with the
increasing introduction of physical collateral instead of group liability, as farmers
showed that they are able and willing to use their land use certificates (Red Books)
as collateral. However, without an effective land market for trading Red Books,
it will probably be risky for the credit institutes to rely more on land use certificates
as collateral. Nevertheless, VBP and VBARD together already have enormous
outreach. At this stage, implementing a consolidation policy and establishing
financially sustainable structures would deserve priority over boosting credit
outreach further by implementing new structures. An important element for inducing
innovation in the microfinance industry is to nurture conditions for
greater competition between different suppliers. As long as no effective competition
exists in the rural financial market in Northern Vietnam, there is little incentive
for the institutes to improve their products. The only competition exists
between the VBP and the VBARD.
This research has shown that poor households are able and willing to save. Over
80% of the households demand a formal savings scheme. The supply of savings
services offers firstly the possibility to create financially sustainable structures
within the existing institutes, and secondly to boost the outreach of the formal
financial sector. When offering savings services to the rural population, especially
to the poor, close physical proximity to customers is seen as a key factor.
This proximity could be achieved by creating decentralized profit centers. Credit
officers would collect and pay out savings as well as performing all credit activities.
Thus, the deposit collection could be done within the village. The local savings
collection by the credit officer would also influence very positively the credit
business. The credit officer has access to a much broader range of information to
115 The recently established VBSP, successor of the VBP, will also offer collateral free soft
loans to individuals.
Chapter 5
118
assess the client’s creditworthiness. These profit centers also guarantee a good
internal monitoring of most operational costs involved in financial intermediation.
A good example is offered by the BRI. Besides, savings instruments need
to be promoted much more than credit. Therefore, a marketing and advertisement
strategy is essential to absorb substantial amount of rural savings.
Nevertheless, the main challenge will be to implement a safe, attractive and
cost-covering deposit collection system at the village level. Simple savings
products can coexist with more complex market-segment-oriented saving products.
Therefore, a range of products should be implemented and promoted. The implementation
of the Vietnam Postal Saving Company (VPSC) is a right step in
this direction. Although the VPSC potentially offers a very deep outreach down
to the commune and village level, the 710 postal savings services are far away
from having a significant outreach to the rural population and particularly to the
rural poor. As WORLD BANK (2002a:46) states: "In a country where 80% of the
population lives in rural areas, even the 2000 branches of VBARD cannot reach
wide, nor deep enough to the population at large." Nevertheless, the politics of
VBARD and VBP respectively VBSP have not changed so far regarding the
implementation of a savings scheme to reach the rural population. Before this
can happen a paradigm change is called for. The rural financial intermediaries in
Vietnam need to recognize the ability and the demand of the rural population to
save.
6 Final conclusion
The rural financial market in Vietnam is, and in the near future will continue to
be, strongly dominated by state-owned financial intermediaries. The liberalization
of interest rates is an important step towards full transformation of the financial
system and offers the potential for financial intermediaries to supply costcovering
services. However, as long as the Government continues to supply subsidized
credits to major parts of the population, it is unlikely that any viable services
(private or state-owned) can be provided in the rural financial market. The
VBSP and the VBARD continue to crowd out competition, inhibiting the deepening
of the financial system and impeding innovations.
A promising development is the attempt to integrate the savings of the rural
population into the financial system. This attempt has resulted in the creation of
the VPSC and might indicate a paradigm change, namely recognition of the rural
population’s ability to save financially, and consequently recognition of their
demand for deposit instruments. Finally, this paradigm change may bring to an end
the tradition of considering credits as the only financial tool for development.
The VBARD, however, despite its immense network, has never reached deep
enough into the country to attract rural savings. Furthermore, the VBSP does not
show any intention to offer savings instruments to its customers. Therefore, discontinuing
the Vietnamese soft loan policy will likely take more time.
Though the creation of the VBSP, the Vietnamese government has at least formally
separated political lending from commercial lending. Earlier evidence from agricultural
development banks in other countries, e.g. by ADAMS and VON PISCHKE (1992)
and HEIDHUES and SCHRIEDER (1999), to name but a few, suggests that VBARD,
now freed from political lending, is likely to dismiss its peasant clientele and concentrate
on more lucrative rural business with bigger and wealthier farmers. The
VBSP will offer subsidized loans on a broad scale and will obviously be a drain on
public resources. The question is, how long can the Vietnamese Government
Chapter 6
120
finance the VBSP, and who will serve the rural poor after the collapse of the
VBSP, should the government discontinue subsidizing the VBSP?116
Despite all the progress achieved in the transformation of the financial system,
its sustainability is also still threatened by an accumulation of non-performing
loans amassed by SOEs over the years. In addition, the problem of nonperforming
loans is spreading to the private sector and particularly to rural
households. Apart from macro-economic threats to the financial system, this
moral hazard behavior is hindering the establishment of any viable rural financial
intermediation.
Compared to a decade ago, the informal market now plays only a minor role.
The breadth of outreach of the major formal rural lenders (VBARD and VBP,
now VBSP) is immense and their poverty outreach is satisfactory since about
50% of all predominantly poor rural households have access to formal credit.
However, the outreach analysis has shown that the poorest households are
seldom clients of formal lenders. Only slightly more than 20% of the poorest
households are served by formal lenders. When considering access to formal
credit, this figure must be re-evaluated. As the logit analysis (see Section 4)
revealed, general poverty (as captured with the poverty index) does not significantly
influence access to formal credit. The results indicate that only certain
aspects of poverty, e.g. low quality of housing, have an important influence on
access to formal credit in Vietnam. Thus, the poorest households use formal
credit less often, but are not significantly more often access-constrained. This
means that the poorest households simply have much less demand for the types
of formal credit products on offer. Improving credit products or offering new
credit lines would only slightly improve the credit coverage of poorer households.
A more promising approach would be to introduce a specialized pro-poor
extension service to widen the scope of their investment ideas and opportunities,
combined with a general improvement in the infrastructure. One factor that very
positively influenced access to and demand for formal credit was the connection
to the market. A good market connection serves credit outreach in a twofold
manner: First, households have better access to credit-relevant information; and
second, through better market access they may find new investment opportunities.
116 The VBSP is assumed to be financially not sustainable (IZUMIDA 2003).
Final conclusion
121
Nevertheless, the share of access-constrained households is surprisingly low
(only 16%). One reason for the low figure is the weakening or eradication of
former access constraints (e.g. lack of physical collateral, literacy requirements,
remoteness) through locally disbursed group credits. Considering the anecdotal
reports of very low repayment rates, the price of eradicating these access constraints
has likely been a decrease in financial sustainability of the formal lenders.
However, some barriers to access continue to exist, particularly for ethnic
minorities or female-headed households. To reduce these access barriers, locallyoriented
rather than general actions should be taken, catering to the needs and
the circumstances of those households which lack access.
Nevertheless, the recent efforts of the Vietnamese government, for instance the
establishment of the VBSP, represent an attempt to broaden access in general.
But this increase in outreach will go hand in hand with an increase in access to
credit for non-creditworthy households, thus resulting in decreased repayment
rates. The situation will possibly be aggravated by the fact that the VBSP will
supply subsidized loans for higher education. Without the creation of appropriate
jobs, however, newly graduated young debtors will face difficulties in paying
back the loan, as has recently been the case in China (THE ECONOMIST 2004).
Furthermore, many people consider political lending to be a gift from the
government, one that they are not required to repay. There is a very real possibility
that expanding accessible policy credit could create a mountain of bad debt
(IZUMIDA 2003). If rural lenders in Vietnam were one day forced to work in a
competitive environment according to market principles, there would be a great
chance that large parts of the population would be access-constrained as a result
of previous loan defaults. Moreover, it is questionable whether households that
do not demand the existing products today will demand credits from the VBSP
in the future.
A more sustainable way to promote outreach would be to improve the rural
population’s knowledge in general, and in particular that of fringe groups such
as ethnic minorities or female-headed households, of credit application procedures,
as access (especially by fringe groups) is often hampered simply by a lack
of information. The information supply to rural households as regards essential
credit information is impeded by the supply-oriented flow of information and by
relying on local authorities for the dissemination of credit relevant information.
Ethnic diversification of bank employees could broaden the information networks
available and could create more awareness of those groups inside the institution.
Furthermore, special female credit officers for female-headed households could
work in the same direction. Local information centers, in combination with the
Chapter 6
122
introduction of regular information days (preferably market days), when bank
staff is locally available for questions and discussion, would further improve the
availability of information. Designing employees’ contracts to include incentives
to pass on information to different social networks would round off a transparent
credit policy.
As mentioned above, however, state-owned rural lenders have already an enormous
outreach. At this stage, implementing a consolidation policy and establishing
financially sustainable structures would deserve to take priority over boosting
credit outreach further by introducing new structures, which are unlikely to improve
credit outreach significantly – particularly as the poorest households
would profit the least from broadening credit outreach on a large scale. The most
appropriate tool to incorporate poorer households into the formal financial system
would the mobilization of savings. As stated by several scholars, all people
can save, even the poorest of the poor, and therefore deposit services have
deeper outreach than credit (CHARITONENKO et al. 2004; SCHREINER 2002;
ZELLER and SHARMA 2000). In contrast to the enormous credit outreach in
Northern Vietnam, formal savings are used by rural households only on an insignificant
scale. However, this low effective demand for savings instruments is
caused by inappropriate savings products and not by the inability or unwillingness
of the rural population to save. This research has proved again that even the
poorest households in Northern Vietnam are able and willing to save. Over 80%
of all households in this sample demand a formal savings scheme. The supply of
client-adapted savings services not only boosts the outreach of the formal financial
system, but also offers the possibility to create financially sustainable structures
within the existing institutes. However, savings products need to be tailored
closely to the needs of the rural population. Otherwise they will not be adopted
by the local people. When offering savings services to the rural population, especially
to the poor, close physical proximity to customers is seen as a key factor
for success. This proximity could be achieved by creating decentralized profit
centers. Credit officers would collect and pay out savings locally as well as performing
all credit activities. Thus, deposit collection could be done within the
village. Local savings collection by the credit officer would also influence the
credit business very positively. The credit officer has access to a much broader
range of information to assess the client’s creditworthiness. These profit centers
also guarantee good internal monitoring of most operational costs involved in
financial intermediation.
Whether the enormous government subsidies channeled into rural credit contribute
to poverty reduction, however, remains a subject for future research. But it is clear
Final conclusion
123
that these subsides bypass households that have no demand for credit, and
these households are often identical to the poorest. Finally, the low credit interest
rates imply low saving rates, which again affect the poor most, as stated
by ADAMS (1984: 72): "The low rates on deposits hurt poor households the most
because they cannot assemble enough savings to buy lumpy, non-financial assets
such as land and cattle. The poor are forced to accept a tax on their savings if
they bother to open accounts or to consume their surplus. The backlash of cheap
credit is that the poor take a beating on their financial savings." This statement
of Adams reflects the situation in Vietnam. The transformation polices of the
Vietnamese government concerning the rural financial system are biased towards
the supply of preferential credits and this policy discriminates the poorest
households against their access to save formal savings products.
Reference list
Adams, D. W. 1984. Are the arguments for cheap agricultural credit sound? In Undermining
rural development with cheap credit. Adams, D. W. et al. (ed.): 92-104. Boulder, USA:
Westview Press.
Adams, D. W., D. Graham, and J. von Pischke. 1984. Undermining rural development with
cheap credit. Boulder, USA: Westview Press.
Adams, D. W., and J. von Pischke. 1992. Microenterprise credit programs: Déja vu. World
Development 20 (10): 1463-1470.
ADB. 2000. Finance for the poor: Microfinance development strategy. Manila, Philippines:
Asian Development Bank (ADB).
ADB. 2001. Asian Development Outlook 2001. Manila, Philippines: Asian Development Bank.
Addelman, S. 1962. Orthogonal main-effect plans for asymmetrical factorial experiments.
Technometrics 4 (1): 21-46.
Administration Book Dia Linh. 2001. Administration book Dia Linh Commune. District Ba Be,
Vietnam. July 2001.
Ahmed, R., and F. Goletti. 1997. Food policy and market reform in Vietnam and Bangladesh;
Annual report essay 1997. Washington DC, USA: International Food and Policy Research
Institute (IFPRI).
Albrecht, J. 1997. Validitätsarten und ihre Überprüfung bei der Conjoint-Analyse. In Märkte
der Agrar- und Ernährungswirtschaft, Schriften der Gesellschaft für Wirtschaft- und Sozialwissenschaften
des Landbaus Band 33. 299-311. Münster-Hiltrup, Germany: Landwirtschaftsverlag
GmbH.
Alther, C., J. C. Castella, P. Novosad, E. Rousseau, and T. T. Hieu. 2002. Impact of accessibility
on the range of livelihood options available to farm households in mountainous areas
of Northern Vietnam. In Doi Moi in the mountains – Land use changes and farmers' livelihood
strategies in Bac Kan Province, Vietnam. Castella, J. C. et al. (ed.): 121-146. Hanoi,
Vietnam: The Agricultural Publishing House.
Amin, S., A. Rai, and G. Topa. 2003. Does microcredit reach the poor and vulnerable? Evidence
from Northern Bangladesh. Journal of Development Economics 70 (1): 59-82.
Anh, V. T. 1997. Criteria for rural poverty monitoring. Vietnam's Socio-Economic Development
9 (1): 3-16.
References
125
Ardrey, W., I. Watson, and N. T. Ha. 1999. Regulation of prudential ratios for banks and
credit organizations in Vietnam. Hanoi, Vietnam: The World Bank and State Bank of
Vietnam (SBVN).
Ba, L. X. 1997. Informal credit and its impact on the poor. Vietnam's Socio-Economic Development
9 (1): 21-26.
Bac, T. H. 2001. Expert interview with the Chief of Statistics and Translation Department and
International Relations Department of the Vietnam Bank of Agriculture and Rural Development.
Hanoi, Vietnam: 06.02.2001/2001.
Backhaus, K., B. Erichson, W. Plinke, and R. Weiber. 1996. Multivariate Analysemethoden –
Eine anwendungsorientierte Einführung. Berlin, Germany: Springer-Verlag.
Bao, N. V. 2001. Expert interview with the head of the VBP and Vice-director of VBARD at
Ba Be district. Ba Be, Vietnam: 24.04., 21.05., and 18.09. 2001/2001.
Barham, B. L., S. Boucher, and M. R. Carter. 1996. Credit constraints, credit unions, and
small-scale producers in Guatemala. World Development 24 (5): 793-806.
Barisch, T., and T. Dufhues. 2001. Facilitating farmers workshops through PRA techniques to
support the visualization of financial services in rural mountainous areas in Northern Vietnam.
Internal report for the subproject F2 'Rural finance and sustainable development in
Northern Vietnam'. Hanoi, Vietnam: The Uplands Program.
Barth, J. R., G. Jr. Caprio, and R. Levine. 2002. Bank regulation and supervision: What works
best? NBER Working Paper 9323. Cambridge, USA: National Bureau of Economic
Research (NBER).
Bastelaer van, T. 2003. Does social capital facilitate the poor's access to credit – A review of
the microeconomic literature. Social Capital Initiative Working Paper No 8. Washington DC,
USA: The World Bank.
Besley, T., and S. Coate. 1995. Group lending, repayment incentives and social collateral.
Journal of Development Economics 46 (1): 1-18.
Bester, H. 1987. The role of collateral in credit markets with imperfect information. European
Economic Review 31 (4): 887-899.
Bhushan, I., E. Bloom, N. H. Huu, and N. M. Thang. 2001. Human capital of the poor in
Vietnam. Manila, Philippines: Asian Development Bank.
Bijl, d. P. W. 1995. Entry deterrence and signaling in markets for search goods. Tilburg,
Netherlands: Tilburg University, Centre for Economic Research.
Binswanger, H. P., and D. A. Sillers. 1983. Risk aversion and credit constraints in farmers'
decision-making: A reinterpretation. Journal of Development Studies 20 (1): 5-21.
Bond, P., and A. Rai. 2002. Collateral substitutes in microfinance. Mimeo. Accessed October
2004 at <http://www.cid.harvard.edu/cidpublications/limits-july2.pdf>.
Bowles, P., and G. White. 1994. Central bank independence: A political economy approach.
Journal of Development Studies 31 (2): 235-264.
Chapter 7
126
Bryant, J. 1998. Communism, poverty, and demographic change in North Vietnam. Population
and Development Review 24 (2): 235-269.
Buchenrieder, G. 2002. Sequencing of institution building in Central and Eastern Europe's
financial system transformation. Frankfurt am Main, Germany: Peter Lang Verlag.
Buchenrieder, G., T. Dufhues, F. Heidhues, and P. T. M. Dung. 2003. Rural finance and sustainable
rural development in Northern Vietnam – Final report of the subproject F2. Stuttgart,
Germany: University of Hohenheim, The Uplands Program.
Buchenrieder, G., and F. Heidhues. 2005. Rural financial markets for food security. In Food
and nutrition security in the process of globalization. Schulz, M. et al. (ed.): 289-304.
Münster, Germany: Lit-Verlag.
Buchenrieder, G., and I. Theesfeld. 2000. Improving bankability of small farmers in Northern
Vietnam. Savings and Development 24 (4): 385-403.
Castella, J. C., T. Q. Hoa, O. Husson, V. H. Nam, and D. D. Quang. 2002. The declining role of
ethnicity in farm household differentiation: A case study from Ngoc Phai Commune, Cho
Don district, Bac Kan Province, Vietnam. In Doi Moi in the mountains – Land use changes
and farmers' livelihood strategies in Bac Kan Province, Vietnam. Castella, J. C. et al. (ed.):
47-71. Hanoi, Vietnam: The Agricultural Publishing House.
Cat, M. V. 2001. Expert interview with a credit officer from the VBARD/VBP at Ba Be district.
Ba Be, Vietnam: 09.10. and 08.11. 2001/2001.
CECARDE. 1999. A rural credit in Vietnam: A case study in Tuyen Quang and Than Hoa.
Hanoi, Vietnam: Center for Consultation on Agriculture and Rural Development
(CECARDE).
Chan, M. N. 2001. Expert interview with the head of the VBARD at Ba Be district. Ba Be,
Vietnam: 09.10. 2001/2001.
Charitonenko, S., A. Campion, and N. A. Fernando. 2004. Commercialization of microfinance:
Perspectives from South and Southeast Asia. Manila, Philippines: Asian Development
Bank (ADB).
Chin, A. T. H., and Ng. H. Guan. 1996. Towards a framework for the transforming economies
of East Asia. In Economic management and transition towards a market economy – An
Asian perspective. Chin, A. T. H. et al. (ed.): 1-11. Singapore: World Scientific Publishing
Co. Pte. Ltd.
Christen, R. P., E. Rhyne, R. Vogel, and C. McKean. 1995. Maximizing the outreach of microenterprise
finance: An analysis of successful microfinance programs. Washington DC,
USA: United States Agency for International Development (USAID).
Christen, R. P., and R. Rosenberg. 2000. Regulating microfinance – The options. Small Enterprise
Development 11 (4): 4-23.
CIA. 2002. The world factbook 2002 – Vietnam. Washington DC, USA: Central Intelligence
Agency (CIA).
Coleman, J. S. 1999. Social capital in the creation of human capital. In Social capital: A multifaceted
perspective. Dasgupta, P. et al. (ed.): 13-39. Washington DC, USA: The World Bank.
References
127
Collier, P., and C. Mayer. 1989. The assessment: Financial liberalization, financial systems,
and economic growth. Oxford Review of Economic Policy 5 (4): 1-12.
Conning, J. 1999. Outreach, sustainability, and leverage in monitored and peer-monitored
lending. Journal of Development Economics 60 (1): 51-77.
Credit book Dia Linh. 2001. Credit book Dia Linh Commune. District Ba Be, Vietnam. Juli 2001.
Demirgüc-Kunt, A., and E. J. Kane. 2001. Deposit insurance around the globe: Where does it
work? Washington DC, USA: The World Bank and Boston College.
Diagne, A. 1999. Determinants of households' access to and participation in formal and informal
credit markets in Malawi. FCND Discussion Paper No 67/1999. Washington DC,
USA: International Food and Policy Research Institute (IFPRI) – Food Consumption and
Nutrition Division (FCND).
Diagne, A., M. Zeller, and M. Sharma. 2000. Empirical measurements of household's access
to credit and credit constraints in developing countries: Methodological issues and evidence.
FCND Discussion Paper No 90/2000. Washington DC, USA: International Food
and Policy Research Institute (IFPRI) – Food Consumption and Nutrition Division (FCND).
Diehl, M. 1998. The role of institutions in the economic transition process: The Reform of
state-owned enterprises in Vietnam. Journal for Institutional Innovation, Development,
and Transition 2: 49-59.
Doan, D. Q. 2001. Expert interview with a credit officer from the VBARD/VBP at Ba Be
district. Ba Be, Vietnam: 21.06., 17.10., and 09.11. 2001/2001.
Du, T. T. N. 2001. Governance and participation in Vietnam. Project discussion Paper
No 12/2001. Duisburg, Germany: University of Duisburg, Institute of East Asian Studies/
East Asian Politics.
Duckett, J. 2001. Bureaucrats in business, Chinese-style: The lessons of market reform and
state entrepreneurialism in the People's Republic of China. World Development 29 (1): 23-37.
Dufhues, T., P. T. M. Dung, H. T. Hanh, and G. Buchenrieder. 2002. Information and targeting
policies and their principal-agent relationships – The case of the Vietnam Bank for the Poor.
Quarterly Journal of International Agriculture 41 (4): 335-362.
Dufhues, T., M. Geppert, and G. Buchenrieder. 2003. Enriching quantitative analysis through
participatory research – Developing demand-driven microsavings services with Conjoint
Analysis in Vietnam. Poster accepted at the 25. International Conference of Agricultural
Economists (IAAE) in Durban, South Africa, August 16-22, 2003 on 'Reshaping Agriculture's
Contributions to Society'.
Dufhues, T., F. Heidhues, and G. Buchenrieder. 2004a. Participatory product design by using
Conjoint Analysis in the rural financial market of Northern Vietnam. Asian Economic
Journal 18 (1): 81-114.
Dufhues, T., U. Lemke, and I. Fischer. 2004b. Constraints and potential of livestock insurance
schemes – A case study from Vietnam. Research in Development Economics and Policy.
Discussion Paper No 5/2000. Stuttgart, Germany: Grauer Verlag.
Chapter 7
128
Dung, P. T. 2002. Overview on insurance and agricultural insurance of Vietnam. Hanoi, Vietnam:
The Uplands Program and Hanoi Agriculture University No. 1 (HAU).
Dung, P. T. M. 2001. Expert interview with rural finance specialist at Faculty of Economics and
Rural Development at Hanoi Agricultural University No 1. Hanoi, Vietnam: 6.10.2001/2001.
Duong, P. B., and Y. Izumida. 2002. Rural development finance in Vietnam: A microeconomic
analysis of household surveys. World Development 30 (2): 319-335.
Eggertsson, T. 1990. Economic behavior and institutions. Cambridge, USA: Cambridge University
Press.
EIU. 2002. Vietnam at a glance. London, UK: The Economist Intelligence Unit Limited (EIU).
Erhardt, W. 2000. Small enterprise finance in urban and rural Northern Thailand. Frankfurt
am Main, Germany: Peter Lang Verlag.
Evans, T., A. M. Adams, R. Mohammed, and A. H. Norris. 1999. Demystifying nonparticipation
in microcredit: A population-based analysis. World Development 27 (2): 419-430.
Fabbri, D., and M. Padula. 2003. Does poor legal enforcement make households creditconstrained?
FAME Research Paper No 81. Geneva, Switzerland: International Center for
Financial Asset Management and Engineering (FAME). Accessed October 2003 at
http//:<www.nccr-finrisk.unizh.ch/media/pdf/fabbri_creditconstrained.pdf>.
Fallavier, P. 1994. Developing microfinance institutions in Vietnam – Policy implications to
set up an enabling environment. M.A. Thesis. Vancouver, Canada: School of Community
and Regional Planning, University of British Columbia. Accessed June 2004 at
<http://www.arts.uwaterloo.ca/~vecon/download/fallavier/mfi_vnm.pdf>.
Fforde, A. 2004. Vietnam in 2003 – The road to ungovernability? Asian Survey 44 (1): 121-129.
Fiebig, M., A. Hannig, and S. Wisniwski. 1999. Saving in the context of microfinance – State
of knowledge. In Challenges of microsavings mobilization: Concepts and views from the
field. Hannig, A. et al. (ed.): 3-26. Eschborn, Germany: Gesellschaft für Technische Zusammenarbeit
(GTZ).
Geppert, M., G. Buchenrieder, and N. T. Dang. 2002. Participatory agricultural research and
decentralization in Vietnam. The Journal of Agricultural Extension and Education 8 (4):
171-180.
Geppert, M., and T. Dufhues. 2003. Visualizing rural financial market research in Northern
Vietnam through pictures. Research in Development Economics and Policy Discussion
Paper No 2/2003. Stuttgart, Germany: Grauer Verlag.
Gibbons, D., B. Jr. Quiñones, J. Remenyi, and H. D. Seibel. 2000. Microfinance for and by
the poor: Lessons from experience. In Microfinance and poverty alleviation – Case studies
from Asia and the Pacific. Remenyi, J. et al. (ed.): 253-269. London, UK: Pinter.
Glewwe, P., M. Gragnolati, and H. Zaman. 2002. Who gained from Vietnam's boom in the
1990's? Economic Development and Cultural Change 50 (4): 773-792.
Gottwald, K., and R. Klump. 1999. Financial reform in Vietnam: Lessons from the Asian crisis.
Passau, Germany: Universität Passau.
References
129
Green, P. E., and V. Srinivasan. 1978. Conjoint Analysis in consumer research: Issues and
outlook. Journal of Consumer Research 5 (2): 103-123.
Green, P. E., and V. Srinivasan. 1990. Conjoint analysis in marketing: New developments
with implications for research and practice. Journal of Marketing 54 (4): 3-19.
Grindle, S. G. 2001. In quest of the political: The political economy of development policymaking.
In Frontiers of development economics – The future perspective. Meier, G. M. et al.
(ed.): 345-380. Washington DC, USA: Oxford University Press.
GSO. 1995. Vietnamese Living Standard Survey 1992/93. Hanoi, Vietnam: General Statistical
Office (GSO), Statistical Publishing House.
GSO. 2000. Vietnamese Living Standard Survey 1997/98. Hanoi, Vietnam: General Statistical
Office (GSO), Statistical Publishing House.
GTZ. 2002. Support for the Establishment of a People's Credit Fund (PCF) System. Hanoi,
Vietnam: Gesellschaft für Technische Zusammenarbeit (GTZ). <http://www.gtz.de/vietnam/
projects/projects_refpriv_pcf_eng.htm> (accessed: March 2002).
Guillen, M. F., and A. E. Tschoegl. 2002. Banking on gambling: Banks and lottery-linked
deposit accounts. Financial Services Research 21 (3): 219-231.
Hammann, P., and B. Erichson. 1994. Marktforschung. Stuttgart, Germany: Gustav Fischer Verlag.
Hanh, H. T. 1999. Presentation of the General Director of VBP in Hanoi at the Agricultural
University No 1. 13.03.1999. Hanoi Vietnam: Agricultural University No 1.
Hanh, H. T. 2001. Expert interview with the General Director of the VBP in Hanoi. Hanoi,
Vietnam: 25.05.2001/2001.
Härdle, W., and L. Simar. 2003. Applied multivariate statistical analysis. Berlin, Germany:
Springer Verlag.
Häuser, I., T. Dufhues, and G. Buchenrieder. 2005. Reliability and applicability of different
poverty measurements – A case study from Northern Vietnam. Research in Development
Economics and Policy. Discussion Paper No 01/2005 (forthcoming). Stuttgart, Germany:
Grauer Verlag.
Heidhues, F. 1995. Rural financial markets – An important tool to fight poverty. Quarterly
Journal of International Agriculture 34 (2): 105-108.
Heidhues, F., and G. Schrieder. 1998. Rural development and financial markets in Romania.
Paper presented at the 46th International Atlantic Economic Conference (IAES) in Boston,
October 8-11, 1998. Boston, USA: IAES.
Heidhues, F., and G. Schrieder. 1999. Rural financial market development. Research in
Development Economics and Policy. Discussion Paper No 1/1999. Stuttgart, Germany:
Grauer Verlag.
Heidhues, F., B. Sossoh, and G. Schrieder. 1997. Transaction costs of group and individual
lending as an influencing factor of access: The case of poverty oriented microfinance in
Cameroon. Sacramento, USA: Paper presented at the 22nd International Conference of
Agricultural Economists (IAAE) on food security, diversification, and resource management:
"Refocusing the role of agriculture". August 10-16, 1997.
Chapter 7
130
Henry, C., M. Sharma, C. Lapenu, and M. Zeller. 2003. Microfinance poverty assessment
tool. Washington DC, USA: The World Bank.
Herring, R. J., and A. M. Santomero. 2000. What is optimal financial regulation? In The new
financial architecture – Banking regulation in the 21st century. Gup, B. E. (ed.): 51-84.
Westport, USA: Quorum books.
Hickson, R. 2001. Financial services for the very poor – Thinking outside the box. Small
Enterprise Development 12 (2): 55-67.
Hien, D. P. 1999. Development under decentralized administration: A case study of Soc son
district, Vietnam. Master Thesis. Bangkok, Thailand: Asian Institute of Technology,
School of Environment and Development.
Hoff, K., and J. E. Stiglitz. 2001. Modern economic theory and development. In Frontiers of
development economics – The future in perspective. Meier, G. M. et al. (ed.): 389-459.
Washington DC, USA: Oxford University Press.
Holtmann, M. 2001. Designing financial incentives to increase loan officer productivity:
Handle with care! The Microbanking Bulletin 6 (April): 5-9.
Huber, J. 2000. The importance of multinominal logit analysis of individual consumer
choices. In CBC system manual 2.6. F1-F18. Washington DC, USA: Sawtooth Software.
Hulme, D., and P. Mosley. 1996. Finance against poverty – Volume 1. London, UK: Routledge.
Hung, B. D. V., and B. M. Giap. 1999. Country Report: Microfinance sector in Vietnam. Hanoi,
Vietnam: Asian Development Bank (ADB).
Izumida, Y. 2003. Programmed lending for social policies – Challenges for the Vietnam Bank
for Social Policies. Report No. 03-F-02. Tokyo, Japan: The University of Tokyo.
Izumida, Y., and P. B. Duong. 2001. Measuring the progress of rural finance in Vietnam. Savings
and Development 15 (2): 139-166.
Janssen, W., J. Ashby, M. Calier, and J. Castano. 1991. Targeting new technology at consumer
food preferences in developing countries. Food and Quality Preference 3 (2): 175-182.
Jensen, M. C., and W. H. Meckling. 1976. Theory of the firm: Managerial behavior, agency
costs and ownership structure. Journal of Financial Economics 3 (4): 305-360.
Kanbur, R., and L. Squire. 2001. The evolution of thinking about poverty: Exploring the interactions.
In Frontiers of development economics – The future perspective. Meier, G. M. et al.
(ed.): 183-226. Washington DC, USA: Oxford University Press.
Karlan, D. S. 2001. Social capital and group banking. Cambridge, USA: Massachusetts Institute
of Technology (MIT), Department of Economics.
Klump, R., and A. Spitzenpfeil. 1998. Vietnam: The financial system of an economy in transition.
In Asian financial markets. Menkhoff, L. et al. (ed.): 335-352. Baden-Baden, Germany:
Nomos Verlag.
Köcher, W. 1990. Die Steuerung von Forschung und Entwicklung mit Hilfe der Maieutischen
Systemanalyse. Miami, USA: Conference on Management of Technology.
References
131
Kovsted, J., J. Rand, F. Tarp, L. V. Thai, V. N. Huong, and N. M. Thao. 2003. Financial sector reforms
in Vietnam: Selected issues and problems. Discussion paper 0301. Hanoi, Vietnam: Central
Institute of Economic Management (CIEM) and Nordic Institute of Asian Studies (NIAS).
Ledgerwood, J. 1999. Sustainable banking with the poor – Microfinance handbook – An institutional
and financial perspective. Washington DC, USA: The World Bank.
Leung, S., and J. Riedel. 2000. The role of the state in Vietnam's economic transition. Canberra,
Australia: Australian National University.
Livingstone, I. 2000. Agriculture, small enterprise development and poverty eradication in
Vietnam. Journal of the Asia Pacific Economy 5 (3): 173-189.
Llanto, G. 2000. Vietnam. In The role of central banks in microfinance in Asia and the Pacific:
Vol. 2, Country studies. 331-356. Manila, Philippines: Asian Development Bank (ADB).
Luibrand, A. 2002. Transition in Vietnam – Impact of the rural reform process on an ethnic
minority. Frankfurt am Main, Germany: Peter Lang Verlag.
Mas, I. 1995. Central bank independence: A critical view from a developing country perspective.
World Development 23 (10): 1639-1652.
Matin, I., D. Hulme, and S. Rutherford. 1999. Financial services for the poor and poorest:
Deepening understanding to improve provision. Finance and Development Research Programme.
Working Paper No 9. Manchester, UK: University of Manchester, Institute for
Development and Policy Management.
McCarty, A. 1999. The employment and social consequences of restructuring in Vietnam.
Hanoi, Vietnam: Paper presented at UNIDO/UNDP/CIEM 'How to make integration work'
conference, 31 March 1999.
McCarty, A. 2001. Microfinance in Vietnam: A survey of schemes and issues. Hanoi, Vietnam:
Department for International Development (DFID) and the State Bank of Vietnam (SBVN).
McGuire, P. B., J. D. Conroy, and G. B. Thapa. 1998. Getting the framework right – Policy
and regulation for microfinance in Asia. Brisbane: Australia: The Foundation for Development
Cooperation.
Mehran, H., and B. Laurens. 1997. Interest rate: An approach to liberalization. Finance &
Development 34 (2): 33-35.
Minot, N. 2000. Generating disaggregated poverty maps: An application to Vietnam. World
Development 28 (2): 319-331.
Minot, N., and F. Goletti. 2001. Rice market liberalization and poverty in Vietnam. Washington
DC, USA: International Food Policy Research Institute (IFPRI).
MOLISA, and UNDP. 2004. Taking stock, planning ahead: Evaluation of the national targeted
program on hunger eradication and poverty reduction (135). Hanoi, Vietnam: United
Nations Development Program (UNDP).
Montes, M. F. 2001. Vietnam: Transition as a socialist project in East Asia. In Transition and
Institutions – The Experience of gradual and late reformers. Cornia, G. A. et al. (ed.): 94-113.
New York, USA: Oxford University Press.
Chapter 7
132
Montgomery, R. 1996. Disciplining or protecting the poor? Avoiding the social costs of peer
pressure in micro-credit schemes. Journal of International Development 8 (2): 289-305.
Moore, W. L. 1980. Levels of aggregation in Conjoint Analysis: An empirical comparison.
Journal of Marketing Research 17 (November): 516-523.
Morduch, J. 1999. The microfinance promise. Journal of Economic Literature 37 (December):
1569-1614.
Navajas, S., M. Schreiner, R. L. Meyer, and C. Gonzalez-Vega. 2000. Microcredit and the
poorest of the poor: Theory and evidence from Bolivia. World Development 28 (2): 333-346.
Neefjes, K. 2001. Poverty reduction amongst ethnic minorities in the mountainous areas of
Vietnam – Reaching targets. Paper presented at the Poverty Task Force Workshop
17.-19.09.2001. Hanoi, Vietnam: United Nations Development Programme (UNDP).
Nelson, P. 1970. Information and consumer behavior. Journal of Political Economy 78 (2):
311-329.
Nguyen, N. X. 1998. Review of rural finance environment in Vietnam – In the context of
outreach (VCRFOP) DID-7001. Hanoi, Vietnam: Vietnam-Canada Rural Finance Outreach
Project.
Nguyen, N. X. 2001. Rural credit schemes, poverty alleviation and nature conservation in
mountainous areas (results of a field study). Vietnam's Socio-Economic Development 26 (2):
38-60.
Nguyen, T. 2002. Vietnams Börse und die Entwicklung eines Kapitalmarktes. Südostasien
aktuell 2002 (Januar): 84-88.
Nguyen-Khac, T. Q., and G. von Guretzky-Cornitz. 1996. Das vietnamesische Bankwesen.
Die Bank 1996 (5): 274-279.
Nhgia, L. X. 2001. Expert interview with the General Director of the department of Economic
Strategy of the State Bank of Vietnam. Hanoi, Vietnam: 16.10. 2001/2001.
North, D. C. 1989. Institutional change and economic history. Journal of Institutional and
Theoretical Economics 145 (1): 238-245.
Oh, S.-N. 1999. Financial deepening in the banking sector – Vietnam. In Rising to the challenge
in Asia: A study of financial markets, Vol. 12: Vietnam. 19-52. Manila, Philippines:
Asian Development Bank (ADB).
Orme, B. 1996. ACA, CBC, or both? Effective strategies for conjoint research. Washington
DC, USA: Sawtooth Software.
Owens, J. V., and S. Wisniwski. 1999. Microsavings: What can we learn from informal savings
schemes. Frankfurt am Main, Germany: Third Annual Conference on Microfinance.
Panjaitan-Drioadisuryo, R. D. M., and K. Cloud. 1999. Gender, self-employment and microcredit
programs – An Indonesian case study. The Quarterly Review of Economics and
Finance 39 (5): 769-779.
Park, A., and C. Ren. 2001. Microfinance with Chinese characteristics. World Development
29 (1): 39-62.
References
133
Pill, H., and M. Pradhan. 1997. Financial liberalization in Africa and Asia. Finance & Development
34 (2): 7-10.
Posen, A. S. 1993. Why central bank independence does not cause low inflation: There is no
institutional fix for policies. In Finance and the international economy: The AMEX-Bank
review prize essays – In memory of Robert Marjolin. O'Brien, R. (ed.): 40-54. Oxford, UK:
Oxford University Press.
Quach, M. H., A. W. Mullineux, and V. Murinde. 2003. Microcredit and household poverty
reduction in rural Vietnam: A cross-sectional study. Paper presented at the Development
Studies Association (DSA) International Conference in Glasgow, 9-11 September 2003.
Devon, UK: Development Studies Association.
Quach, M. H., A. W. Mullineux, and V. Murinde. 2004. Rural credit and household poverty
reduction in Vietnam: Evidence using panel data from household surveys. Paper presented
at the International Conference on Access to Finance, 28-29 October 2004. Brussels, Belgium:
World Savings Bank Institutes.
Que, L. X. 1997. Rural credit: Reality and problems. Vietnam's Socio-Economic Development 9 (1):
27-32.
Randolph, T., and L. Ndung'u. 2000. Gender and transaction costs: A conjoint analysis of
choice livestock health services among smallholder dairy farmers in Kenya. Berlin, Germany:
Background paper for a poster presented at the 24th Congress of the International Association
of Agricultural Economists (IAAE), Berlin, 18-24 August 2000.
Richter, R. 1990. Sichtweise und Fragestellungen der Neuen Institutionenökonomik. Zeitschrift
für Wirtschafts- und Sozialwissenschaften 110 (4): 571-591.
Richter, R., and E. Furubotn. 1996. Neue Institutionenökonomik – Eine Einführung und kritische
Würdigung. Tübingen, Germany: Mohr Siebeck.
Riedel, J. 1999. Needed: A strategic vision for setting reform priorities in Vietnam. In Vietnam
and the East Asian Crisis. Leung, S. (ed.): 9-30. London, UK: Edward Elgar Publications
Ltd.
Riedel, J. 2000. The role of the state and the market in the economy of Vietnam. Hanoi, Vietnam:
Ministry of Planning and Investment (MIP) and United Nations Development Program
(UNDP).
Riedel, J., and W. S. Turley. 1999. The politics and economics of transition to an open market
economy in Vietnam. Paris, France: Organization for Economic Co-operation and Development
(OECD) – OECD Development Centre.
Robinson, M. S. 2001. The microfinance revolution – Sustainable finance for the poor.
Washington DC, USA: The World Bank.
Rondinelli, D. A., and J. Litvack. 1999. Economic reform, social progress, and institutional
development: A framework for assessing Vietnam's transition. In Market reform in Vietnam –
Building institutions for development. Litvack, J. et al. (ed.): 1-30. Westport, USA: Quorum
Books.
Rutherford, S. 2000. The poor and their money. New Delhi, India: Oxford University Press.
Chapter 7
134
Ruthven, O. 2001. Money mosaics: Financial choice and strategy in West Delphi squatter settlement.
Finance and Development Research Programme Working Paper No. 32. Manchester,
UK: Institute for Development Policy and Management, University of Manchester.
Sarap, K. 1990. Factors affecting small farmers' access to institutional credit in rural Orissa,
India. Development and Change 21 (2): 281-307.
SBVN. 2001. Annual report 2000. Hanoi, Vietnam: State Bank of Vietnam (SBVN).
Schenk, R. 1998. Einflußfaktoren auf den Kreditzugang bei Kleinbauern in Nordvietnam.
Institute of Agricultural Economics and Social Sciences in the Tropics and Subtropics (490a),
Chair of Agricultural Development Theory and Policy, University of Hohenheim.
Schlüter, A. 2001. Institutioneller Wandel in der Transformation – Restitution, Transformation
und Privatisierung in der tschechischen Landwirtschaft. Aachen, Germany: Shaker Verlag.
Schmidt, R. H. 1999. Development of finance, microfinance, and banking regulation. Savings
and Development 24 (1): 111-121.
Schmidt, R. H., and E. Terberger. 1996. Grundzüge der Investitions- und Finanzierungstheorie.
Wiesbaden, Germany: Gabler Verlag.
Schmidt, R. H., and C.-P. Zeitinger. 1994. Critical issues in small and microbusiness finance.
Paper presented at the seminar on financial services and the poor: U.S. and developing
country experience. September 28-30. Washington DC, USA: The Brookings Institution.
Schreiner, M. 2002. Aspects of outreach: A framework for discussion on the social benefits of
microfinance. Journal of International Development 14 (5): 591-603.
Schrieder, G. 1996. The role of rural finance for food security of the poor in Cameroon.
Frankfurt am Main, Germany: Peter Lang Verlag.
Schrieder, G., and F. Heidhues. 1991. Conjoint Analysis and its predictive power for financial
market development: A methodological framework. In Finance and Rural Development in
West Africa. Proceedings of the 12th Seminar on Rural Economics. Cuevas, C. et al. (ed.):
83-91. Montpellier, France and Columbus Ohio, USA: Centre de Coopération Internationale
en Recherche Agronomique pour le Dévelopment (CIRAD) and Ohio State University.
Schrieder, G., and F. Heidhues. 1998. Transition and rural finance. Quarterly Journal of International
Agriculture 37 (4): 265-282.
Schrieder, G., and F. Heidhues. 2000. Transition and rural finance. In Romania – Rural finance
in transition economies. Heidhues, F. et al. (ed.): 3-18. Frankfurt am Main, Germany: Peter
Lang Verlag.
Schrieder, G., and M. Sharma. 1999. Impact of finance on poverty reduction and social capital
formation – A review and synthesis of empirical evidence. Savings and Development 13 (1):
67-94.
Seibel, H. D. 1992. The making of a market economy: Monetary reform, economic transformation
and rural finance in Vietnam. Saarbrücken, Germany: Breitenbach Verlag.
Seibel, H. D. 1996. Microfinance strategies: Strategies for developing viable microfinance
institutions and sustainable micro-financial services in Asia. Working Paper No 2 1996.
Cologne, Germany: University of Cologne, Development Research Center.
References
135
Seibel, H. D. 1997. Aufbau eines Verbundsystems für ländliche Spar- und Darlehenskassen in
Vietnam. Gesamtbericht No 95.3521.2-006.00. Eschborn, Germany: Deutsche Gesellschaft
für technische Zusammenarbeit (GTZ).
Seibel, H. D., and P. Schmidt. 2000. How an agricultural development bank revolutionized
rural finance: The case of Bank Rakyat Indonesia. Rome, Italy: International Fund for
Agricultural Development (IFAD).
Senanayake, S. M. P., and D. P. Ho. 2001. What makes formal rural financial institutions successful
in Vietnam? Savings and Development 15 (4): 475-489.
SGT Daily. 2003. First policy bank set to make debut this month: Governor. March 6, 2003.
(accessed: April 2003, <www.vneconomy.com.vn>). Saigon Times (SGT) Daily/2003.
Ho Chi Minh City, Vietnam: Saigon Times Group.
Sharma, M. 2000. The scope for policy reforms in rural microfinance. In microFINANCE:
A pathway from poverty. Sharma, M. (ed.): Washington DC, USA: International Food and
Policy Research Institute (IFPRI).
Sharma, M. 2001. Microfinance. In Empowering women to achieve food security.
Quisumbing, A. R. et al. (ed.): 19-20. Washington DC, USA: International Food Policy
Research Institute (IFPRI).
Shimomoto, Y. 2003. Developing the capital market – Vietnam. In Rising to the challenge in
Asia: A study of financial markets, Vol. 12: Vietnam. 53-72. Manila, Philippines: Asian
Development Bank (ADB).
Sida-MARD. 1998. Strategy for rural financial services. Hanoi, Vietnam: Sweden International
Development Agency (SIDA) and Ministry of Agriculture and Rural Development
(MARD).
Simanowitz, A. 2000. Targeting the poor – Comparing visual and participatory methods.
Small Enterprise Development 11 (1): 29-39.
Statistical Publishing House. 1999. Socio-economic statistical data of 61 provinces and cities.
Hanoi, Vietnam: Statistical Publishing House.
Statistics book Nghien Loan. 2001. Statistics book Nghien Loan Commune. District Ba Be,
Vietnam. Juli 2001.
Statistics book Xuan La. 2001. Statistics book Xuan La commune. District Ba Be, Vietnam.
Juli 2001.
Stiglitz, J. E. 1986. The new development economics. World Development 14 (2): 257-265.
Stiglitz, J. E. 2002. Globaslization and its discontents. New York, USA: W.W. Norton &
Company.
Stiglitz, J. E., and A. Weiss. 1981. Credit rationing in markets with imperfect information.
American Economic Review 71 (3): 393-410.
Tam, L. M. 2000. Reforming Vietnam's banking system: Learning from Singapore's model.
Singapore: Institute of Southeast Asian Studies.
Teufel, H. 1997. Banken und Finanzmärkte in Vietnam. Kreditwesen 1997 (2): 77-81.
Chapter 7
136
Thai, T. D. 1998. Financial sector reform in Vietnam. In Financial deregulation, capital flows
and macroeconomic management in the Asia Pacific. Hansen, T. et al. (ed.): 213-220.
Seoul, Korea: The Korea Development Institute.
Thayer, C. A. 2000. Vietnam, the Asian financial crisis, and Doi Moi 2. Harvard Asia Quarterly 4 (1):
The Economist. 2003. A survey of global finance – The trouble with banks, nobody loves
them, everybody needs them. The Economist 367/2003. London, UK: The Economist
Newspaper Limited.
The Economist. 2004. Poor payers to a degree. The Economist 8379/2004.
Theesfeld, I. 2000. Vietnam's rural credit market – Determinants of farmers' bankability.
Master Thesis. Stuttgart, Germany: Institute of Agricultural Economics and Social Sciences
in the Tropics and Subtropics (490a), Chair of Agricultural Development Theory and Policy,
University of Hohenheim.
Thu, H. 3-19-2001. Bank for the Poor 'under-capitalised'. Vietnam Investment Review (VIR)/
3-19-2001. Hanoi, Vietnam: Ministry of Planing and Investment (MIP).
Todd, H. 1996. Cloning Grammen Bank: Replicating a poverty reduction model in India, Nepal
and Vietnam. London, UK: Intermediate Technology Publications.
Toffler, A. 1990. Power shift. New York, USA: Bantam Books.
Tompson, W. 1998. The politics of central bank independence in Russia. Europe-Asia Studies
50 (7): 1157-1182.
Tran, H. A., and K. Dang. 1996. Vietnam – Banking regulatory regime in Vietnam. World
Reports 7 (3): (accessed: May 2000,< www.hg.org/supp48.html>).
UN. 1998. Expanding choices for the rural poor – Human development in Vietnam. Hanoi,
Vietnam: United Nations (UN).
UNDP. 1999. Completion of Vietnam's legal framework for economic development. Hanoi,
Vietnam: United Nations Development Program (UNDP).
UNIDO, DSI, and MPI. 2003. Vietnam industrial competitiveness review. Hanoi, Vietnam:
United Nations Industrial Development Organization (UNIDO), Development Strategy
Institute (DSI), and Ministry of Planning and Investment (MPI).
Vaessen, J. 2001. Accessibility of rural credit in Northern Nicaragua: The importance of networks
of information and recommendation. Savings and Development 25 (1): 5-32.
van de Walle, D. 2003. Are returns to investment lower for the poor? Human and physical
capital interactions in rural Vietnam. Review of Development Economics 7 (4): 636-653.
VBARD. 1997. Annual report VBARD, District Ba Be. Ba Be, Vietnam: Vietnam Bank for
Agriculture and Rural Development (VBARD).
VBARD. 1998. Annual report VBARD, District Ba Be. Ba Be, Vietnam: Vietnam Bank for
Agriculture and Rural Development (VBARD).
VBARD. 1999. Annual report VBARD, District Ba Be. Ba Be, Vietnam: Vietnam Bank for
Agriculture and Rural Development (VBARD).
References
137
VBARD. 2000a. Annual report – Agribank – Steady development and integration into the
international Community. Hanoi, Vietnam: Vietnam Bank for Agriculture and Rural Development
(VBARD).
VBARD. 2000b. Annual report VBARD, District Ba Be. Ba Be, Vietnam: Vietnam Bank for
Agriculture and Rural Development (VBARD).
VBARD. 2001a. Annual report VBARD, District Ba Be. Ba Be, Vietnam: Vietnam Bank for
Agriculture and Rural Development (VBARD).
VBARD. 2001b. Leaflet on guidelines, pre-conditions, and procedures for loans at the
VBARD (available only in Vietnamese). Hanoi, Vietnam: Vietnam Bank for Agriculture
and Rural Development (VBARD).
VBARD and Danida. 1999. Agriculture sector program support – Vietnam – Component description
document. Hanoi, Vietnam: Government of Vietnam and Vietnam Bank for Agriculture
and Rural Development (VBARD) and Danish Ministry of Foreign Affairs (Danida).
VBP. 1997. Annual report VBP, District Ba Be. Ba Be, Vietnam: Vietnam Bank for the Poor (VBP).
VBP. 1998a. Annual report VBP, District Ba Be. Ba Be, Vietnam: Vietnam Bank for the Poor
(VBP).
VBP. 1998b. Policy guidelines of the Ba Be district. Ba Be, Vietnam: Vietnam Bank for the
Poor (VBP).
VBP. 1999a. Annual report VBP, District Ba Be. Ba Be, Vietnam: Vietnam Bank for the Poor
(VBP).
VBP. 1999b. Vietnam Bank for the Poor – Annual report. Hanoi, Vietnam: Vietnam Bank for
the Poor (VBP).
VBP. 2000. Annual report VBP, District Ba Be. Ba Be, Vietnam: Vietnam Bank for the Poor
(VBP).
VBP. 2001a. Annual report VBP, District Ba Be. Ba Be, Vietnam: Vietnam Bank for the Poor
(VBP).
VBP. 2001b. Policy document on training of credit groups. Hanoi, Vietnam: Vietnam Bank
for the Poor (VBP).
Viet Nam News. 2002a. Joint stock banks held back by low funds. October 31, 2002. (accessed:
December 2003, <http://vietnamnews.vnagency.com.vn/202-10/30/Stories/18.htm>). Viet
Nam News/2002a. Hanoi, Vietnam: Vietnam News Agency.
Viet Nam News. 2002b. Voice of Vietnam (VOV) reaches out to ethnic groups. Viet Nam
News 3815/2002b. Hanoi, Vietnam: Vietnam News Agency.
Viet Nam News. 2003. Government enlist new bank in poverty fight. March 12, 2003. (accessed:
April 2003, <www.vneconomy.com.vn>). Viet Nam News/2003. Vietnam, Hanoi: Vietnam
News Agency.
Vietnam Economic Times. 2002. Chance for the poor people to have easy access to bank
loans. October 16, 2002. (accessed: April 2003, <www.vneconomy.com.vn>). Vietnam
Economic Times/2002. Ho Chi Minh City, Vietnam: Ringier Pacific Ltd.
Chapter 7
138
Vietnam Economy. 2003. Vietnam's social policy bank established. March 12, 2003. Vietnam
Economy/2003. (accessed: April 2003, <www.vneconomy.com.vn>).
Vietnam-Canada Rural Finance Outreach Project. 1999a. Diagnostic report on capacity of
Vietnam Bank for Agriculture and Rural Development to outreach to low-income households.
Hanoi, Vietnam: Vietnam-Canada Rural Finance Outreach Project.
Vietnam-Canada Rural Finance Outreach Project. 1999b. Proposal on testing units at
VBARD. Hanoi, Vietnam: Vietnam-Canada Rural Finance Outreach Project.
VIR. 2001. Poverty fight given $ 1.4 bn price tag by officials. March 12, 2001. Vietnam Investment
Review (VIR)/2001. Hanoi, Vietnam: Ministry of Planing and Investment (MPI).
VIR. 2003. Policy Bank opens to boost anti-poverty campaign. March 18, 2003. (accessed:
April 2003, <www.vneconomy.com.vn>). Vietnam Investment Review (VIR)/2003.
Hanoi, Vietnam: Ministry of Planing and Investment (MPI).
von Pischke, J. 1991. Finance at the frontier – Debt capacity and the role of credit in the private
economy. Washington DC, USA: The World Bank.
Warner, B. 2001. Vietnam – Economic and financial management. Canberra, Australia:
Australian Agency for international Development (AusAID).
Williamson, O. E. 1985. The economic institutions of capitalism – Firms, markets, relational
contracting. New York, USA: Free Press.
Wittink, D. R., and P. Cattin. 1989. Commercial use of Conjoint Analysis: An update. Journal
of Marketing 53 (3): 91-96.
Wolff, P. 1999. Vietnam – The incomplete transformation. London, UK: Frank Cass Publishers.
Woller, G., C. Dunford, and W. Woodworth. 1999. Where to microfinance? International
Journal of Economic Development 1 (1): 29-64.
Wolz, A. 1997. The transformation of rural finance systems in Vietnam. Discussion Paper No 60.
Heidelberg, Germany: Universität Heidelberg, Forschungsstelle für Internationale Wirtschafts-
und Agrarentwicklung e.V. (FIA).
Wolz, A. 1999. Achievements and problems of people's credit funds in Vietnam. Heidelberg,
Germany: Lehrstuhl für internationale Wirtschafts- und Entwicklungspolitik, Universität
Heidelberg.
World Bank. 1995. Viet Nam – Environmental program and policy priorities for a socialist
economy in transition. Hanoi, Vietnam: The World Bank.
World Bank. 1998. Vietnam – Deepening reform for growth. Hanoi, Vietnam: The World Bank.
World Bank. 1999. Vietnam attacking poverty – Vietnam development report 2000. Hanoi,
Vietnam: The World Bank.
World Bank. 2000a. Selected world development indicators. Washington DC, USA: The
World Bank.
World Bank. 2000b. Vietnam 2010: Entering the 21st century – Vietnam development report
2001 – Partnership for development. Hanoi, Vietnam: The World Bank.
References
139
World Bank. 2000c. Vietnam 2010: Entering the 21st century – Vietnam development report
2001 – Pillars of development. Hanoi: Vietnam: The World Bank.
World Bank. 2001. Vietnam development report 2002 – Implementing reforms for faster
growth and poverty reduction. Hanoi, Vietnam: The World Bank.
World Bank. 2002a. Banking sector review. Hanoi, Vietnam: The World Bank.
World Bank. 2002b. Vietnam – Delivering on its promise – Development report 2003. Hanoi,
Vietnam: The World Bank.
World Bank. 2003. Poverty – Vietnam development report 2004. Hanoi, Vietnam: The World
Bank.
World Bank. 2004a. Financial sector policy issue note: Vietnam Bank for Social Policies.
Hanoi, Vietnam: The World Bank.
World Bank. 2004b. Governance – Vietnam development report 2005. Hanoi, Vietnam: The
World Bank.
World Bank, and DFID. 1999. Vietnam, voices of the poor; synthesis of participatory poverty
assessments. Hanoi, Vietnam: The World Bank and Department for International Development
(DFID).
Wright, G. 1999. Beyond basic credit and savings: Developing new financial service products
for the poor. In Challenges of microsavings mobilization: Concepts and views from the
field. Hannig, A. et al. (ed.): 45-54. Eschborn, Germany: Gesellschaft für Technische
Zusammenarbeit (GTZ).
Wright, G., and L. Mutesasira. 2001. The relative risks to the savings of poor people. Small
Enterprise Development 12 (3): 33-45.
Zeller, M. 1999. The role of rural financial services for alleviation of food insecurity and poverty.
Agriculture and Rural Development 6 (2): 28-32.
Zeller, M. 2001. Promoting institutional innovation in microfinance – Replicating best practices
is not enough. Development and Cooperation (D+C) 2001 (1): 8-11.
Zeller, M., G. Schrieder, J. Braun von, and F. Heidhues. 1997. Rural finance for food security
of the poor: Concept, review, and implications for research and policy. Washington DC,
USA: International Food Policy Research Institute (IFPRI).
Zeller, M., and M. Sharma. 1998. Rural finance and poverty alleviation. Food Policy Report.
Washington DC, USA: International Food Policy Research Institute (IFPRI).
Zeller, M., and M. Sharma. 2000. Many borrow, more save, and all insure: Implications for
food and microfinance policy. Food Policy 25 (2): 143-167.
Zeller, M., M. Sharma, C. Henry, and C. Lapenu. 2001. An operational tool for evaluating
poverty outreach of development policies and projects. FCND Discussion Paper No 110/2001.
Washington DC, USA: International Food Policy Research Institute (IFPRI) – Food Consumption
and Nutrition Division (FCND).
Chapter 7
140
Zeller, M., M. Sharma, C. Henry, and C. Lapenu. 2005. An operational method for assessing
the poverty outreach performance of development projects: Results of case studies in
Africa, Asia and Latin America. World Development 34 (3): 446-464.
Zeller, M., M. Wollni, and A. A. Shaban. 2003. Evaluating the poverty outreach of development
programs: Results from case studies in Indonesia and Mexico. Quarterly Journal of
International Agriculture 42 (3): 371-383.
Annex
Table A-1: Descriptive statistics of the independent variables for the binary
logistic regression model on credit access
Variable name Unit Minimum Maximum
Mean Std. deviation
Red or Green Books (yes/no) 0.00 1.00 0.89 0.32
Agricultural land (1000 m2) 0.00 65.00 9.51 7.82
Value of houses (Mill.
VND)
0.00 130.00 10.11 12.92
Government salary (yes/no) 0.0.0 1.00 0.13 0.33
Cash savings (Mill.
VND)
0.0.0 55.00 1.48 5.57
School years of HH head (years) 0.0.0 14.00 5.20 3.02
Vietnamese communication
skills of the married couple
(yes/no) 0.0.0 1.00 0.83 0.38
Receiving agricultural extension
service
(yes/no) 0.00 1.00 0.82 0.39
Active HH members (numbers) 0.50* 7.00 2.83 1.26
Share of non-farm activities
in total yearly income
(%) 0.00 100.00 15.17 21.51
Lost working days due to
illness per year/HH
(days) 0.00 509.00 42.46 58.29
Giving help per year/HH (days) 0.00 200.00 36.26 30.75
Receiving help per year/HH (days) 0.00 300.00 25.76 36.96
Interest-free informal credit (Mill.
VND)
0.00 18.00 0.18 1.36
Thai/Tay village (yes/no) 0.00 1.00 0.72 0.45
Market visits only by female
HH members
(yes/no) 0.00 1.00 0.36 0.48
Age of the household (years) 1.00 81.00 19.66 14.14
Remoteness (km) 1.50 24.00 10.48 7.47
Market visits per month (numbers) 0.00 30.00 5.63 6.46
Different markets visited (numbers) 0.00 5.00 1.13 0.47
Poverty index -1.82 3.09 0.00 1.00
Supply of day labor (yes/no) 0.00 1.00 0.35 0.48
Receiving aid from government
(yes/no) 0.00 1.00 8.76 0.28
Source: Own calculations.
Notes: * Households containing only persons over the retirement age were counted as having 0.5
work-forces.
HH = household.
Annex
142
Figure A-1: Decision tree of the effective credit demand in the formal sector
Credit application:
yes/no?
(N = 251)
Received full credit:
yes/no
(N = 151; 60 %)
No demand
(N = 70; 28%)
Access problems
(N = 28; 11%)
No credit demand
(N = 70; 28 %)
No access to credit
(N = 41; 16%)
No Yes
No Yes
Credit refused
(N =13; 5 %)
Credit rationed1
restricted access
Access to credit
(N = 140; 56%)
•No investment needs
•Sufficient capital
•Too complicated
•Lack of knowledge
•No collateral
•Credit conditions
•Bribes
•Afraid of debts
•Do not know why
•Loan funds exhausted
•No/not enough collateral
Source: Adapted from BARHAM et al. (1996) and HEIDHUES and SCHRIEDER (1998).
Notes: 1 Credit-rationed households do have access to the formal financial system and are therefore
not separated in the analysis. Besides, in one research area (Ba Be) not a single household
was rationed. Nine households were excluded from the sample because of missing values.
This decision tree includes the formal and semi-formal financial institutes, VBARD, VBP,
and the State Treasury.
Figure A-2: Principal Component Indicators
Ten indicators were chosen for the PCA, including the benchmark indicator, three assetrelated
variables, one food-related variable, two dwelling-related variables and three indicators
related to human resources. The poverty component is given by:
PC1 = 0.587 * per person expenditure on clothes and footwear
+ 0.674 * total value of assets per person
+ 0.662 * value of electronics and appliances per household (HH)
+ 0.596 * value of transportation-related assets per HH
− 0.497 * months without enough to eat per year
+ 0.531 * type of roof
+ 0.566 * electricity supply
+ 0.675 * percentage of adults with only primary education
+ 0.592 * percentage of adults with college education
+ 0.542 * percentage of literate adults
This equation accounts for 35.4% of the total variance of the original data.
Source: HÄUSER et al. (2005).
Table A-2: Logit estimation of average utility values for credit attributes, by wealth classes
Indigent (N = 134) Medium (N = 82)1 Better-off/rich (N = 42)1 Total (N = 258)
Utilities t-values Utilities t-values Utilities t-values Utilities t-values
Livestock
insurance
Yes (5,000 VND per month
and animal)
1.541 6.800*** 1.239 4.794*** 8.949 0.170 1.562 9.353***
No -1.541 -6.800*** -1.239 -4.794*** -8.949 -0.170 -1.562 -9.353***
Relative importance in % 21% 15% 21% 20%
Disbursal time Seven days 2.078 4.801*** 2.654 3.510*** 6.019 0.130 1.947 7.032***
60 days -2.078 -4.801*** -2.654 -3.510*** -6.019 -0.130 -1.947 -7.032***
Relative importance in % 28% 32% 14% 24%
Lending
scheme
Group lending -1.109 -3.643*** -1.275 -2.942*** -3.903 -0.074 -1.184 -5.414***
Individual lending 1.109 3.643*** 1.275 2.942*** 3.903 0.074 1.184 5.414***
Relative importance in % 15% 15% 9% 15%
Collateral Land use certificates 0.857 2.616** 0.880 1.916* 5.106 0.098 1.261 5.744***
requirement Durable consumer goods -1.645 -2.822*** -0.873 -1.011 -14.739 -0.196 -2.298 -5.989***
No collateral 0.788 1.940** -0.007 -0.013 9.632 0.128 1.037 3.658***
Relative importance in % 17% 11% 29% 22%
District -0.947 -2.648*** -0.756 -1.563 -11.481 -0.148 -1.388 -5.763***
Commune -0.931 -1.622* -1.894 -2.214** 10.405 0.146 -0.281 -0.823
Village 1.878 4.857*** 2.650 4.264*** 1.075 0.023 1.669 6.519***
Location of
credit negotiation,
disbursal,
etc. Relative importance in % 19% 27% 26% 19%
None option 3.566 6.466*** 3.895 4.338*** 16.330 0.185 3.521 9.100***
Percentage of households choosing none
option
10% 7% 14% 10%
Chi-square 385.170 277.465 113.648 734.149
Source: Own calculations.
Notes: * Significant at 10% level. **Significant at 5% level. ***Significant at 1% level. 1 Due to the small sample size or little variance, the default settings
did not yield interpretable results for all attributes or the regression model did not converge. Therefore, the settings were changed to 30 iterations in
stead of 20, with a smaller step size between the iterations 0.5 instead of one, and the change of log-likelihood from one iteration to the next was
changed from 1e-005 to 1e-004. The attribute ‘interest rate’ was excluded from the analysis, as the level ‘0.5% interest rate/month’ was chosen in
100% of the cases.
Table A-3: Logit estimation of average utility values for saving attributes, by wealth classes
Indigent (N = 134) Medium (N = 82)1 Better-off /rich (N = 42)1 Total (N = 258)
Utilities t-values Utilities t-values Utilities t-values Utilities t-values
Interest rate and term
No interest rate / demand deposit -1.750 -2.355** -2.120 -0.271 -2.026 -0.243 -1.401 -3.702***
0.3% Per month for a one-month deposit -0.063 -0.105 -1.063 -0.136 -1.789 -0.214 -0.347 -1.075
0.5% Per month for a three-month deposit 1.813 5.614*** 3.184 0.204 3.815 0.229 1.747 8.243***
Relative importance in % 45% 26% 30% 38%
Incentive scheme
No lottery -0.069 -0.154 -3.802 -0.163 -3.297 -0.132 -0.612 -2.070**
Lottery 0.069 0.154 3.802 0.163 3.297 0.132 0.612 2.070**
Relative importance in % 2% 37% 33% 15%
Location of depositing and withdrawing
Village 1.935 5.820*** 0.645 0.083 -0.124 -0.015 1.448 8.077***
Commune -0.571 -1.083 1.537 0.099 1.827 0.110 -0.251 -0.839
District -1.364 -3.018*** -2.182 -0.279 -1.704 -0.205 -1.196 -4.755***
Relative importance in % 42% 18% 18% 32%
Minimum requirement for
account opening
No minimum requirement 0.458 1.483 1.828 0.156 1.884 0.151 0.589 3.045***
20,000 VND -0.458 -1.483 -1.828 -0.156 -1.884 -0.151 -0.589 -3.045***
Relative importance in % 12% 18% 19% 14%
None option 2.283 6.012*** 4.251 0.218 2.726 0.131 1.864 6.483***
Percentage of households
choosing none option
17% 11% 5% 13%
Chi-square 249.034 168.742 73.906 459.531
Source: Own calculations.
Notes: * Significant at 10% level. **Significant at 5% level. ***Significant at 1% level. 1 Due to the small sample size or little variance the default settings
did not yield in interpretable results for all attributes or the regression model did not converge. There-fore, the settings were changed to 30 iterations
instead of 20, with a smaller step size between the iterations 0.5 instead of one, and the change of log-likelihood from one iteration to the next was
changed from 1e-005 to 1e-004.
Studies on the Agricultural and Food Sector in Central and Eastern Europe
edited by Leibniz Institute of Agricultural Development in Central and Eastern Europe
(IAMO)
ISSN 1436-221X
Vol. 1 The Importance of Institutions for the Transition in Central and
Eastern Europe with Emphasis on Agricultural and Food Industry
ed. by Klaus Frohberg and Witold-Roger Poganietz
1998, 137 pages, ISBN 3-8175-0258-3
Vol. 2 The Significance of Politics and Institutions for the Design and
Formation of Agricultural Policies
ed. by Klaus Frohberg and Peter Weingarten
1999, 254 pages, ISBN 3-8175-0289-3
Vol. 3 Food Processing and Distribution in Transition Countries. Problems
and Perspectives
ed. by Monika Hartmann and Jürgen Wandel
1999, 349 pages, ISBN 3-8175-0293-1
Vol. 4 Die private Nachfrage nach Nahrungsmitteln im Transformationsprozeß
Tschechiens und Polens
Stephan Brosig (PhD)
2000, 171 Seiten, ISBN 3-8175-0319-9
Vol. 5 Integrating Estonia into the EU: Quantitative Analysis of the
Agricultural and Food Sector
Achim Fock (PhD)
2000, 286 pages, ISBN 3-8175-0320-2
Vol. 6 Competitiveness of Agricultural Enterprises and Farm Activities in
Transition Countries
ed. by Peter Tillack and Frauke Pirscher
2000, 216 pages, ISBN 3-8175-0322-9
Vol. 7 Конкурентоспособность сельскохозяйственных предприятий и
фермерской деятельности в странах переходного периода
под редакцией Петера Тиллака и Фрауке Пиршер
2000, 253 страницы, ISBN 3-8175-0324-5
Vol. 8 Perspectives on Agriculture in Transition: Analytical Issues,
Modelling Approaches, and Case Study Results
ed. by Witold-Roger Poganietz, Alberto Zezza, Klaus Frohberg and
Kostas G. Stamoulis
2000, 433 pages, ISBN 3-8175-0323-7
Vol. 9 Land Ownership, Land Markets and their Influence on the Efficiency
of Agricultural Production in Central and Eastern Europe
ed. by Peter Tillack and Eberhard Schulze
2000, 485 pages, ISBN 3-8175-0325-3
Vol. 10 Landwirtschaft und Industrie in Russland – der Transformationsprozeß
in der Ernährungsindustrie
Jürgen Wandel (PhD)
2000, 361 Seiten, ISBN 3-8175-0334-2
Vol. 11 Food Consumption in Russia. An Econometric Analysis Based on
Household Data
Karin Elsner (PhD)
2001, 256 pages, ISBN 3-8175-0335-0
Vol. 12 Alexander Wasiljewitsch Tschajanow – die Tragödie eines großen
Agrarökonomen
hrsg. u. übers. von Eberhard Schulze
2001, 192 Seiten, ISBN 3-8175-0342-3
Vol. 13 Analysis of Food Consumption in Central and Eastern Europe:
Relevance and Empirical Methods
ed. by Stephan Brosig and Monika Hartmann
2001, 253 pages, ISBN 3-8175-0349-0
Vol. 14 Wettbewerbsprozesse und Firmenwachstum in der Transformation
am Beispiel der polnischen Fleischindustrie
Agata Pieniadz (PhD)
2002, 291 Seiten, ISBN 3-8175-0360-1
Vol. 15 Agricultural Enterprises in Transition: Parallels and Divergences in
Eastern Germany, Poland and Hungary
ed. by Ludger Hinners-Tobrägel and Jürgen Heinrich
2002, 455 pages, ISBN 3-8175-0366-0
Vol. 16 Agricultural Technology and Economic Development of Central and
Eastern Europe. Results of the Workshop in Halle, 2nd – 3rd July 2001
ed. by Peter Tillack and Ulrich Fiege
2002, 160 pages, ISBN 3-86037-199-1
Vol. 17 Региональные аспекты аграрных преобразований: политика,
реструктуризация, рыночная адаптация
под редакцией Петера Тиллака и Виталия Зиновчука
2003, 236 страницы, ISBN 3-928466-55-0
Vol. 18 Alexander Vasilievich Chayanov – the Tragedy of an Outstanding
Agricultural Economist
ed. by Eberhard Schulze
2003, 188 pages, ISBN 3-86037-201-7
Vol. 19 Development of Agricultural Market and Trade Policies in the CEE
Candidate Countries
by the Network of Independent Agricultural Experts in the CEE Candidate
Countries
2003, 72 pages, ISBN 3-86037-212-2
Vol. 20 Large Farm Management
ed. by Alfons Balmann and Alexej Lissitsa
2003, 396 pages, ISBN 3-86037-213-0
Vol. 21 Success and Failures of Transition – the Russian Agriculture between
Fall and Resurrection
ed. by Eberhard Schulze, Elke Knappe, Eugenia Serova, Peter Wehrheim
2003, 521 pages, ISBN 3-9809270-1-6
Vol. 22 Subsistence Agriculture in Central and Eastern Europe: How to
Break the Vicious Circle?
ed. by Steffen Abele and Klaus Frohberg
2003, 233 pages, ISBN 3-9809270-2-4
Vol. 23 Pfadabhängigkeiten und Effizienz der Betriebsstrukturen in der
ukrainischen Landwirtschaft – Eine theoretische und empirische
Analyse
Andriy Nedoborovskyy (PhD)
2004, 197 Seiten, ISBN 3-86037-216-5
Vol. 24 Nichtmonetäre Transaktionen in der ukrainischen Landwirtschaft:
Determinanten, Spezifika und Folgen
Olena Dolud (PhD)
2004, 190 Seiten, ISBN 3-9809270-3-2
Vol. 25 The Role of Agriculture in Central and Eastern European Rural
Development: Engine of Change or Social Buffer?
ed. by Martin Petrick and Peter Weingarten
2004, 426 pages, ISBN 3-9809270-4-0
Vol. 26 Credit rationing of Polish farm households – A theoretical and
empirical analysis
Martin Petrick (PhD)
2004, 254 pages, ISBN 3-9809270-6-7
Vol. 27 Drei Jahrhunderte Agrarwissenschaft in Russland: Von 1700 bis zur
Gegenwart
Alexander Alexandrowitsch Nikonow und Eberhard Schulze
2004, 232 Seiten, ISBN 3-9809270-8-3
Vol. 28 Russlands Weg vom Plan zum Markt: Sektorale Trends und
regionale Spezifika
Peter Voigt (PhD)
2004, 270 Seiten, ISBN 3-9809270-9-1
Vol. 29 Auswirkungen des Transformationsprozesses auf die sozioökonomischen
Funktionen ukrainischer Landwirtschaftsunternehmen
Helga Biesold (PhD)
2004 182 Seiten, ISBN 3-938584-00-9
Vol. 30 Agricultural policies and farm structures – agent-based modelling
and application to EU-policy reform
Kathrin Happe (PhD)
2004, 291 pages, ISBN 3-938584-01-7
Vol. 31 How effective is the invisible hand? Agricultural and Food Markets in
Central and Eastern Europe
ed. by Stephan Brosig and Heinrich Hockmann
2005, 361 pages, ISBN 3-938584-03-3
Vol. 32 Erfolgsfaktoren von landwirtschaftlichen Unternehmen mit
Marktfruchtanbau in Sachsen-Anhalt
Kirsti Dautzenberg (PhD)
2005, 161 Seiten, ISBN 3-938584-06-8
Vol. 33 Agriculture in the Face of Changing Markets, Institutions and
Policies: Challenges and Strategies
ed. by Jarmila Curtiss, Alfons Balmann, Kirsti Dautzenberg,
Kathrin Happe
2006, 544 pages, ISBN 3-938584-10-6
Vol. 34 Making rural households’ livelihoods more resilient – The importance
of social capital and the underlying social networks
ed. by Gertrud Buchenrieder and Thomas Dufhues
2006, 106 pages, ISBN 3-938584-13-0
Vol. 35 Außerlandwirtschaftliche Diversifikation im Transformationsprozess.
Diversifikationsentscheidungen und -strategien ländlicher Haushalte
in Slowenien und Mazedonien
Judith Möllers (PhD)
2006, 323 Seiten, ISBN 3-938584-14-9
Vol. 36 Accessing rural finance – The rural financial market in Northern
Vietnam
Thomas Dufhues (PhD)
2007, 166 Seiten, ISBN 3-938584-16-5

Read More