MINISTRY OF EDUCATION AND TRAINING
UNIVERSITY OF ECONOMICS HOCHIMINH CITY
----------

VU DUC CAN

This thesis
is PREFERENCES,
completed at University
Economics
Ho Chi
RISK
SOCIALof
CAPITAL,
AND
Minh City.

MICROCREDIT RISKS: AN EXPERIMENTAL

STUDY
IN THE
DELTA
Supervisor
1: Assoc.
Prof.MEKONG
Dr. Truong
QuangREGION
Thong OF
VIETNAM

Supervisor 2: Dr. Nguyen Duc Quang



3

ABSTRACT
The aim of this study is to empirically analyze social and
demographic factors related to microfinance borrowers in order to
measure their effects on microcredit risks as undergone by
microfinance institutions (MFIs) in the Mekong Delta Region of
Vietnam. Further, the study looks at risk preferences, social capital,
and others with respect to microfinance borrowers’ behavior to
estimate their impact on microcredit risks facing MFIs.
In this study, a series of economic experiments was conducted
with the participation of microfinance borrowers in six provinces
of the Mekong Delta Region to capture the effects of risk
preferences and social capital on microcredit risks, which was well
justified by the findings. Specifically, those who seek more risk are
less likely to have bad debt, while those being more risk averse
suffer more. Given social capital, mutual support in the community
and trust impact positively on microcredit risk. These results form
a firm basis for devising feasible policies in direct relation to
microfinance lending activities involving MFIs.
Keywords: microfinance, risk preferences, social capital, risk
seeking, risk averse.


Chapter 1: INTRODUCTION
1.1. Problem statement.
Microfinance has come into existence and gone through a long
history of development, thus establishing its significant role and

preference is a tendency toward risk decisions as can be made by
individuals and investors to obtain the highest possible
profitability. Handa (1971) argued that risk preference is the choice
between a high-risk asset and low-risk asset to gain higher returns.
Charness et al. (2013) and Eckel et al. (2010) concluded that in
economics emphasis can be put on suggestive methods in
analyzing risk preferences, and the suggested preferences can be
affected by the measures used. According to Stiglitz and Weiss
(1981), borrowers are motivated, and have a tendency, to invest in
risk-ridden projects. This means that borrowers with bad debts are
willing to take high risks. The experiments carried out by Zeballos
et al. (2014) showed that borrowers having no bad debts
demonstrate more risk-seeking behavior than those with bad debts.
Stiglitz and Weiss (1981) hypothesized that people investing in less
risky projects are those suffering bad debts. The poor cannot repay
their loans because they refuse to face risks, so the effectiveness is
low (Zeballos et al., 2014).
In Vietnam, Vieider et al. (2013) concluded that farmers are on
average risk neutral and that income is negatively associated with
risk aversion. The studies of Nguyen et al. (2016) and Tanaka et al.
(2010) in northern and southern villages accentuated the impact of
risk attitudes and risk and time preferences on trust and reliability,
and risk aversion and patience. So, which specific behavioral
characteristic of microfinance borrowers has influence on
microcredit risks? Do risk preferences differ in microfinance


practices between rural and urban regions? These are also the
research gap for this paper to fill.
1.1.2. Social capital and microcredit risks.

borrowers’ risk preferences and other social, demographic factors
affect microcredit risk of MFIs? and (ii) How do borrowers’ social
capital and other social, demographic factors influence microcredit
risks of MFIs?
1.4. Research methods.
Several research hypotheses are proposed based on the review
of related literature and relevant theoretical framework as well as
the results of previous studies. Field experimental method is
adopted to collect the data. Regression analysis with Binary
Logistic is performed to process the data, along with the use of
Probit technique to test the robustness of the results obtained. The
results are then screened, discussed, and interpreted to put forward
policy implications, and limitations are also outlined as a basis for
future research.
The technique proposed by Eckel and Grossman (2002) is
adopted to conduct the experiment of eliciting risk preferences.
Concerning social capital, the study suggests the Game of
contributions to community, and Camerer and Fehr’s (2003)
method is employed to investigate trust and reliability.
1.5. Research participants and scope.
- Participants: microcredit borrowers of microcredit providers,
including both formal and semi-formal institutions.
- Scope: A total of 176 microfinance borrowers residing in both
rural and urban areas in six provinces of the Mekong Delta Region,
namely Kien Giang, Hau Giang, Vinh Long, Tien Giang, Ben Tre,
and Long An. All six surveys and experiments were undertaken
from May through October 2017.


1.6. Contributions of the study.


study.
Introduction.

Chapter 2: Risk preferences, social capital, and microcredit risks.
Chapter

3:

Research

design.

Chapter 4: Analysis of the effects of different factors on
microcredit risks: Surveys and experiments in the Mekong Delta
Region.
Chapter 5: Result discussion and policy implications.


Chapter 2: RISK PREFERENCES, SOCIAL CAPITAL, AND
MICROCREDIT RISKS
2.1. Risk preferences and microcredit risks.
In human society risk is perceived to exist in all kinds of
activities. Attitudes of people toward risk remarkably differ;
therefore, it can be used to speculate on their economic behavior
and decisions. The impact of risk is direct and diverse, from
borrowers’ behavior and activities to investment, manufacture, and
consumption and behavior toward risk. Other effects arise from
demographic, financial, or physical factors and social capital.
2.1.1. Prospect theory.

getting high-interest loans. According to Zeballos et al. (2014),


borrowers without non-performing loans seek more risk than those
who have. Eckel and Grossman (2008) found that female students
are more risk averse than their male counterparts. While
Binswanger (1980) detected no difference in risk compared to the
scope of investment between the rich and poor, Vieider et al.
(2015) showed that unmarried people are less risk averse, whereas
women and the elderly are more risk averse.
In brief, studies on risk preferences concern a wide range of
subjects and areas. The results also indicated many differences in
subjects, areas, and fields of study. However, the impact of risk
preferences

on

risks

involving

microcredit

lending

and

microfinance activities in Vietnam has not been studied at length.
So, how do risk preferences as well as other social, demographic
factors of microfinance borrowers affect the risks in microfinance

the better the ability to repay loans and the more economical it is.
Greiner and Wang (2009) exhibited an information asymmetry
between lenders and borrowers.
2.2.3. Studies on social capital in Vietnam.
In a study by Nguyen Tuan Anh and Thomése (2007), social
capital was shown to well handle troubles involving land
consolidation in agriculture. Nguyen Van Ha and Kant (2004)
indicated that social capital has a strong and positive influence on
household income. Tran Huu Dung (2003) pointed out the
relationship between social capital and economic policy; between
social capital and economic growth. Social capital has a profound
impact on the quality and rate of human capital accumulation. Dinh
Hong Hai (2013) noted the dark side of social capital. Ngo Thi
Phuong Lan (2011) believed that social capital helps minimize
risks in the transition from rice cultivation to shrimp farming in the
Mekong Delta. Nguyen Hong Thu (2018) concluded that


microfinance has an impact on the income of poor households,
which, as argued by Mai Thi Hong Dao (2016), is affected by such
factors as age, household size, rate of dependency, total assets,
microcredit, and regions affect the income of poor households.
Phan Dinh Khoi (2013) concluded that working for local
authorities, being members of loan groups, education level, skilled
labor, and inter-commune roads affects the accessibility to
microfinance. Dinh Phi Ho and Dong Duc (2015) asserted that
household characteristics, residential locations, and shocks of
environmental risks have influence on farm households’ income
and expenditure.
Accordingly, it is conceivable that social capital has a

environment, legal environment.
 Subjective risks:
⁕ MFIs: administration capacity, lending procedure and
policy, loan inspection and supervision, service and ethical quality
of credit staff.
⁕ Microfinance borrowers: education level, production and
business capacity, ethical issues.
2.3.3. Measuring microcredit risks in the study.
2.3.3.1. Definition of bad debt and related views.
Bad debt can be referred to as ‘doubtful debt’ or ‘nonperforming loan’. WB defined it as substandard loans that may be
overdue, or doubts arise over the repayment capacity as well as the
recoverability of capital, frequently occurring in the event that the
debtor has been declared bankrupt or detected with property
dispersion. According to IMF, “a loan is nonperforming when
payments of interest and/or principal are past due by 90 days or
more, or interest payments equal to 90 days or more have been
capitalized, refinanced, or delayed by agreement, or payments are
less than 90 days overdue, but there are other good reasons—such
as a debtor filing for bankruptcy—to doubt that payments will be
made in full.’
Risks in microfinance lending practices
Prospect theory.
- Risk preferences.
- Personal behavior.
- Risk averse.
- Risk seeking.
- Risk neutral.

Bad
debt

3.1.3. Basis for location and sample selection.
This is a key area of the southern region of Vietnam, full of
distinct characteristics of ecosystems, industries, and ethnic groups.
The author has over 20 years’ experience in the banking industry
for proper selection of participants. Given the total number of 176
participants, 33.5% was selected from Vietnam Bank for Social
Policies (VBSP), and the remaining 66.5% from local commercial
banks.
3.2. Selecting experimental methods in economics.
3.2.1. Eliciting risk preferences.


Several experimental techniques have been developed to delve
into attitudes toward personal risks, including Balloon Analogue
Risk Task (BART), questionnaires, methods as proposed by
Gneezy and Potters, Eckel and Grossman, and selection based on
price lists.
3.2.2. Measuring social capital.
The methods applied comprise Trust Game and Public Goods
Game.
3.2.3. Assessing and selecting methods.
The author employed Eckel and Grossman’s technique, Trust
Game, and Public Goods Game.
3.2.4. Organization and role assignment.
The author first collected demographic information and
decided on the locations for Risk Game (Game 1). In Games 2 and
3, each assigned group is composed of one group leader, one
secretary, and one assistant.
3.2.5. Basis for determining rewards.
Table 3.2: Average income and expenditure per capita per day



Each participant received VND100,000 and would have to
make a random selection of one of the six Scenarios as detailed in
Table 3.3. Then, they were instructed to cast either of the two lots
labelled “win” and “lose”.
Table 3.3: Game options
Unit: VND
Scenario
1
2
3
4
5
6

Amount
subtracted
0
-20.000
-40.000
-60.000
-80.000
-100.000

Amount
received

Selection


debt

and

0

otherwise.

participant’s selection;

Independent
represents age;

variables:

represents

represents gender;

represents education level.
- Z is a control variable denoting living area.
3.4.2. Regression equation for public goods Game experiment:
where is a dependent variable, and is an independent variable
representing the participant’s intention to donate.
3.4.3. Regression equation for trust Game experiment:
where is a dependent variable, is an independent variable denoting
percentage of the amount given to his partner by the participant.
3.4.4. Regression equation for all three experiments (robustness
check):
where is a dependent variable, represents participant’s selection,

4.1.1. Overall statistics on participants’ characteristics.
4.1.1.1. Bad debt.
Most of the survey participants have no bad debt (81.8%,
standard loans).
4.1.1.2. Education level.
The general level is rather low. 18.2% are found not to have
completed primary education. While 33% have completed primary
education but not secondary high school education, 29.5% have
finished secondary high school, yet not high school, education.
4.1.1.3. Living areas.
63.1% and 36.9% of the participants live in rural and urban
areas respectively.
4.1.1.4. Mortgages.
37.5% of the households are found with property mortgages
and 62.5% with unsecured loans.


4.1.1.5. Locations of households’ bank loans.
33.5% of the loans are requested at Vietnam Bank for Social
Policies and 66.5% at local joint stock banking institutions.
4.1.1.6. Principal income sources.
There is a relatively even distribution of households’ sources of
income (ranging from 23.3% to 29.5%).
4.1.1.7. Quantitative indicators.
The average age is 46.9 along with the most advanced of 85
and the earliest of 20. Standard deviation is 12.4. Average
household size is 4.5 people (max is 12, and min is 1). The average
percentage of household members with employment is 68.7%. The
average loan rate is VND23.58 million per household. The duration
of loans repayment is about 15.8 months, ranging from 1 to 60

mostly completely risk averse (Scenario 2), besides the trends
toward risk seeking behavior (Scenario 6) and risk neutral behavior
in rural areas.
4.1.3.2. Participants’ characteristics and their options
in Public Goods Game.
81.1% of the participants choose to contribute; males
and females account for 85.3% and 79.2%, respectively.
Urban and rural citizens with contribution make up 81.5%
and 82%, respectively.
4.1.3.3. Participants’ characteristics and their options in Trust
Game.
The give-money option reaches 92%. Most participants agree to
give money to their partners, but little difference exists between
males and females as well as between urban and rural inhabitants.
4.2. Testing differences in certain criteria based on participants’
characteristics.
4.2.1. Differences in bad debt.


•

Gender: 75 male participants have bad debt, whereas this figure for
females is 101. At 10% significance level, there exists a difference
in characteristics of loans between males and females.

•

Living areas: 65 urban participants have bad debt, whereas this
figure for their rural counterparts is 111. At 10% significance level,
no difference can be detected in characteristics of loans in terms of

Game.
4.2.3.1. Between borrowers’ characteristics and their options in
Public Goods Game.
At 10% significance level, the results are as follows:
 Gender: Participants’ options are different.
 Living areas: Participants’ options are not different.
4.2.3.2. Between borrowers’ options and their bad debt in
Public Goods Game.
A certain difference exists: Most participants with contribution
do not have bad debt, and conversely, those without do have.
4.2.4. Differences in certain characteristics in Trust Game.
4.2.4.1. Intention to give money according to roles of
participants.
A difference exists in decision to return money to partners.
4.2.4.2. Amounts of money given to partners.
Amounts of money given by number 1 participants to their
number 2 partners are larger than those given by number 2
participants to their number 1 partners.
4.2.4.3. Amounts of money given to partners according to bad
debt.
Those having no bad debt give their partners larger amounts of
money than those who have.


4.3. Regression results of the effects of different factors
on bad debt.
4.3.1. Regression results of the effects of different
factors on bad debt in Risk Game.
At 10% significance level, the results are as follows:
 Riskier options have negative associations with bad debt. That is,

At 10% significance level, the results are as follows:
•

Those who give more money to their partners will less likely have
bad debt.

•

Considering merely the effects of age and gender, these are not
attributed to bad debt.

•

The higher the education level of a participant, the less likely he
bears the burden of bad debt. For young people, more advanced
age is associated with less bad debt. Still, the older a participant
becomes, the more likely he suffers bad debt.

•

Larger loans have a link with less bad debt suffered. No substantial
conclusion can be drawn with respect to the difference in bad debt
level between urban and rural areas. The number of household
members with employment history does not make any difference to
bad debt, and neither do loan terms.
4.3.4. Regression results of the effects of different
factors on bad debt by combining Risk Game and
Public Goods Game.
At 10% significance level, the results are as follows:


conclusion can be drawn with respect to the difference in bad debt
level between urban and rural areas. The number of household
members with employment history does not make any difference to
bad debt, and neither do loan terms.
4.3.6. Regression results of the effects of different
factors on bad debt by combining Public Goods Game
and Trust Game.
At 10% significance level, the results are as follows:

• Age, gender, and education level all have no impact on the
likelihood of having bad debt.
•

Larger loans have a link with less bad debt suffered. No substantial
conclusion can be drawn with respect to the difference in bad debt
level between urban and rural areas. The number of household
members with employment history does not make any difference to
bad debt, and neither do loan terms.