You are on page 1of 15

This article was downloaded by: [UQ Library]

On: 21 June 2012, At: 00:08


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Development Studies


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/fjds20

Assessing the Welfare Effects of


Microfinance in Vietnam: Empirical
Results from a Quasi-Experimental
Survey
a b b
Son Nghiem , Tim Coelli & Prasada Rao
a
Centre of National Research on Disability and Rehabilitation
Medicine, the University of Queensland, Australia
b
School of Economics, the University of Queensland, Australia

Available online: 18 Apr 2012

To cite this article: Son Nghiem, Tim Coelli & Prasada Rao (2012): Assessing the Welfare Effects
of Microfinance in Vietnam: Empirical Results from a Quasi-Experimental Survey, Journal of
Development Studies, 48:5, 619-632

To link to this article: http://dx.doi.org/10.1080/00220388.2011.638051

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-


conditions

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
Journal of Development Studies,
Vol. 48, No. 5, 619–632, May 2012

Assessing the Welfare Effects of Microfinance


in Vietnam: Empirical Results from a
Quasi-Experimental Survey
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

SON NGHIEM*, TIM COELLI** & PRASADA RAO{


*Centre of National Research on Disability and Rehabilitation Medicine, the University of Queensland, Australia,
**{School of Economics, the University of Queensland, Australia

Final version received June 2011

ABSTRACT. This article analyses the effects of NGO microfinance programmes on household welfare in
Vietnam. Data on 470 households across 25 villages were collected using a quasi-experimental survey approach
to overcome any self-selection bias. The sample was designed so that member households of microfinance
programmes were compared with non-member households with similar characteristics. The analysis shows no
significant effects of participation in NGO microfinance on household welfare, proxied by income and
consumption per adult equivalent.

1. Introduction
Microfinance involves the provision of financial services to the poor. The principle, which
underpins its operation, is to use group lending with peer monitoring to secure loans instead of
the traditional reliance on physical collateral used elsewhere in the finance industry. During the
last 30 years, microfinance has evolved from a grassroots movement to a global industry with
more than 150 million clients worldwide (Harris, 2009). This emerging form of finance has
attracted the attention of international development agencies, commercial finance institutions
and governments.
Over the last two decades the microfinance movement has developed rapidly in Vietnam, but
there has been little research into the contribution that microfinance has made to the welfare of
its Vietnamese clients. Our current knowledge about the impact of microfinance in Vietnam has
primarily been reliant on anecdotal evidence; the limited empirical literature has not fully
addressed the self-selection issue that can introduce a bias into the results. Moreover, there is no
consensus in the microfinance literature that has attempted to control for selection bias (e.g., Pitt
and Khandker, 1998; Coleman, 1999; Khandker, 2005; Chemin, 2008; Roodman and Morduch,
2009), or about the significance of microfinance impacts. Pitt and Khandker (1998), Khandker
(2005) and Chemin (2008) found that microfinance significantly improved the income of
participants in Bangladesh, but Coleman (1999), for Thailand, and Roodman and Morduch
(2009) for Bangladesh find that the impacts are either small or statistically insignificant.

Correspondence Address: Son Nghiem, Centre of National Research on Disability and Rehabilitation Medicine, the
University of Queensland, QLD 4006, Australia. Email h.nghiem@uq.edu.au

ISSN 0022-0388 Print/1743-9140 Online/12/050619-14 ª 2012 Taylor & Francis


http://dx.doi.org/10.1080/00220388.2011.638051
620 S. Nghiem et al.

We conduct a quasi-experimental survey design first described by Coleman (1999) to


investigate the effects of microfinance programs provided by non-governmental organisations
(NGOs) on the welfare of clients in Vietnam. The survey design offers a simple and efficient way
to mitigate the self-selection issue. Our main contribution is a proposed estimation of the
spillover effect using the cross-section quasi-experimental survey. We also integrate semi-
structured interviews of other stakeholders (e.g., programme officers, village leaders and donors)
and secondary data in the analysis of the household survey data. Our findings are consistent with
Coleman (1999) who reports that in Thailand, microfinance created no statistically significant
effect on the income or consumption of its clients.
The article is organised into five sections. Following the introduction, Section 2 presents a brief
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

review of the selection bias issue and an overview of microfinance in Vietnam. Section 3 describes
the sampling design used in the household survey, choice of variables and descriptive statistics.
The econometric analysis is in Section 4 with concluding remarks in Section 5.

2. A Brief Review of the Literature


2.1 Selection Bias and Microfinance Impacts
The evaluation of microfinance impacts may be confounded by selection bias, due to
voluntary participation in programmes (i.e., self-selection) or the selective placement of
programmes (i.e., non-random placement). Unfortunately very few studies published in the
past two decades have addressed this issue adequately (Armendariz and Morduch, 2010: Ch.
9). The main methods selected by previous researchers to address these issues include the
utilisation of instrumental variables (Pitt and Khandker, 1998), analysing panel data
(Khandker, 2005), propensity score matching (Chemin, 2008), conducting randomised control
trials (Banerjee et al., 2009; Karlan and Zinman, 2010), and the implementation of quasi-
experimental surveys (Coleman, 1999; Kaboski and Townsend, 2011). Instrumental variable
(IV) techniques can be an effective method to control for the selection bias issue, but finding
an appropriate instrumental variable in microfinance studies has proved to be notoriously
difficult. Pitt and Khandker (1998) used the exogenous eligibility criterion of microfinance
programmes in Bangladesh (i.e., households who own less than half an acre of land) to
correct for selection bias. Morduch (1998) pointed out problems with this and subsequent
attempts to address or correct for the problem yield conflicting conclusions regarding the
impact of microfinance in Bangladesh; see Pitt (1999), Chemin (2008), Roodman and
Morduch (2009). Heckman and Urzua (2010) show that parameters estimated by IV methods
can be different from the true causal effects, while it is well known that IV estimates are
sensitive to the choice of valid instruments. Alternative approaches are necessary to identify
treatment effects of microfinance.
Advocates of randomised control trials (RCT) argue that they can produce unbiased results.
As eligible households are randomly selected in a properly designed trial, microfinance
programmes should be the cause for any differences between outcomes of programme
participants and that of non-participants. Although RCTs have been regarded as the gold
standard for empirical impact evaluation, they face various challenges. First, impact assessment
involves a multi-level process (e.g., regional and individual) while in reality randomisation is
often done only at the individual level (Deaton, 2010). Second, as observed by Armendariz and
Morduch (2010: 302), spillover effects remain an obstacle for RTC studies when randomisation is
conducted at household or individual level. Finally, RCTs are often more costly and intensive
than alternative methods such as quasi-experimental surveys.
Quasi-experimental surveys are designed to ensure that the characteristics of control (non-
member) and treatment (member) groups are similar. One of the first microfinance studies to use
a quasi-experimental survey design was Coleman (1999), who compared outcomes of households
included in the programme with those who remained on the waiting list. In both programme and
The Welfare Effects of Microfinance in Vietnam 621

non-programme villages, eligible and ineligible households were selected, and impacts were
measured by controlling for the duration that a household participated in the microfinance
programmes. The author also used a fixed effect estimator to control for the possibility that the
order in which eligible villages received microfinance programmes was not random. Coleman
found that the effect of microfinance on household income was not significant. The author
argued that the small loan size of microfinance programmes might explain their insignificant
impact. Coleman recommended that further research, including analysis of microfinance in other
nations, is required to form a clearer view of the impacts of this financial tool.
Quasi-experimental surveys offer several advantages. First, they can control for endogeneity
without imposing the large research costs associated with RTCs. Second, the use of structural
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

models can explain how microfinance affects the livelihood of clients. Third, one can apply
econometric methods such as fixed effects to control for any residual endogeneity. Finally, a
quasi-experimental survey has the ability to address spillover effects, even in a cross-section
setting.
A small literature (e.g., Pham and Izumida, 2002; Barslund and Tarp, 2008) has examined
microfinance in Vietnam but the focus has been on the demand for credit and credit rationing,
rather than the consequences of credit provision per se. To the best of our knowledge, no
previous Vietnamese studies have examined the impact of microfinance on the non-government
(NGO) sector, or used a quasi-experimental survey. The aim of this study is to examine the
effects of microfinance on the welfare of participating households using a quasi-experimental
survey following Coleman (1999). The survey offers a more holistic approach by including semi-
structured interviews and discussions with programme managers, local government officials and
donors to provide context to the data collected. In addition, a simple way to control for spillover
effects is proposed by using the village type dummy variable. We also address zero-value data
using the Battese (1997) approach with a translog production function, an approach that has not
previously been used in microfinance studies.

2.2 Microfinance in Vietnam


Microfinance is relatively new in Vietnam, having started in the early 1990s to restore the rural
credit system after the collapse of the traditional credit cooperatives in the late 1980s. This
section reviews the main microfinance service providers in Vietnam, focusing on formal and
semi-formal sectors.

2.2.1 The formal sector. Banks and formal financial institutions share a large proportion of the
microfinance market in Vietnam, providing services to some 5.9 million households, including
2.7 million poor households (McCarty, 2001). The main providers in this sector are the Vietnam
Bank for Agriculture and Rural Development (VBARD), the Vietnam Bank for Social Policy
(VBSP) and the People’s Credit Fund (PCF). The VBARD was established in 1988 and is one of
the largest banks in Vietnam with some 2500 branch offices spread throughout the country,
serving seven million households and accounting for 58 per cent of rural households in Vietnam
(Wolz, 1997). In 2004, the size of the average loan lent by the VBARD was $US1300, reflecting
the fact that its target clients are mainly micro-enterprises, rather than poor households. The
VBARD also offers savings services with total deposits of $US7.5 billion from some seven
million saving accounts in 2005.
The VBSP, initially named the Vietnam Bank for the Poor, was established in 1995 with the
objective of contributing to poverty reduction. The VBSP did not set up its own infrastructure
but used the existing branch network and staff of the VBARD. The target clientele of the VBSP
were poor households defined by the national poverty line. The funds used by the VBSP are
mobilised from various sources within the government and therefore does not rely solely on its
own savings. While funds provided by the government and its agencies are interest free, the
VBSP also receives government subsidies to cover the gap between the interest rate paid to
622 S. Nghiem et al.

commercial banks and the lending rates charged to its clients (World Bank, 2002). By 1997, the
VBSP delivered credit to 1.3 million households with a total loan volume of 1800 billion Dong
(at the time of this survey $US1 ¼ 15,000 Dong), averaging 1.37 million Dong per household
(Wolz, 1997).
The PCF was established in 1993 with the main function of mobilising local idle funds to
provide credit to households and individuals. The main difference between PCF and the credit
cooperatives that collapsed in the late 1980s is that it pays greater attention to savings
mobilisation. The PCF followed the operational model of the Caisse Populaire system in
Canada, which gives members a sense of ownership through the provision of shares with
accompanying voting rights in the decision-making process. The PCF model also developed
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

rapidly, with more than one thousand PCFs serving 1.2 million members by 2008 (Bateman,
2010: 193).

2.2.2 The semi-formal sector. The semi-formal players in microfinance include non-bank
institutions, which are authorised to provide credit but are not allowed to accept savings from
non-members. The main service providers in this sector include NGO microfinance programmes
(NMPs), which started operating in Vietnam in the early 1990s (UNDP, 1996). Most NMPs were
components of integrated development programmes conducted by NGOs, and thus the number
of NMPs has varied with the number of NGO projects in operation. When their development
projects were completed, only a few NMPs continued and these were transformed into member-
based financial institutions, such as cooperatives or community banks. The target groups of
NMPs had been primarily women and the poor in the remote and disadvantaged regions.
Because of limited funding and a focus on the poorest sector, the outreach from NMPs has
remained modest, with estimated outstanding loans of $US2.1 million serving 67,000 clients
(McCarty, 2001).
Despite their small scale, NMPs have several positive characteristics when compared to formal
service providers. The microfinance component of most NMPs was integrated with other
development activities such as health care and agricultural extension in order to enhance the aim
of achieving the ultimate goal of reducing poverty (UNDP, 1996). More importantly, NMPs
often operated in partnership with local mass organisations, especially the Women’s Union
(WU), in which all stakeholders are likely to gain from the partnership. For example, NMPs can
have access to a nationwide network of WUs to reach their main target clientele (i.e.,
poor women) while WU can be more attractive to its members owing to these microfinance
services.
Most NMPs followed the Grameen Bank model, in which members of an NMP organise
themselves into groups of five persons with no family or marriage relationship. Each person then
acts as a guarantor for the other persons in the group to obtain loans. Group members meet
periodically (e.g., weekly, fortnightly or monthly) to contribute savings, and repay interest and
principal by instalments. The choice of meeting frequency depends on the characteristics of
members. For example, in groups where members are small traders they usually prefer to meet
weekly since they have more frequent cash earnings while farmers prefers to meet monthly
because they have less frequent cash transactions.
In short, the rural finance market in Vietnam is heterogeneous with the main players being
VBARD, VBSP, PCF and NMPs (UNDP, 1996). This study examines NMPs in Vietnam and
compares their performance with the comparable Thai microfinance institutions examined by
Coleman (1999).

3. The Survey Design and Data


The study is based on a multi-level survey of institutions and households. The institutional
survey includes semi-structured interviews with managers of microfinance service providers, local
officials and donors to explore their perspectives on the effectiveness of microfinance on poverty
The Welfare Effects of Microfinance in Vietnam 623

reduction, and to examine their operational efficiency. In addition, the survey of microfinance
managers collected information on characteristics of NMPs that were used to select the sample
for the household survey.
The household survey followed Coleman (1999), whereby the control group consists of those
eligible households who would be able to receive the services when the programme expanded.
Although there are various providers of microfinance services in Vietnam, (e.g. VBSP and
VBARD), this study focused on NMPs because of the lack of comprehensive research on this
sector. Due to restrictions on time and resources, the survey focused on NMPs in nine provinces
in the north (Hanoi,2 Ha Tay, Hoa Binh, Phu Tho, Yen Bai, and Son La) and central (Thanh
Hoa, Ha Tinh, and Quang Binh) Vietnam, where most of the poverty reduction programmes,
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

including microfinance, operate.3


All NMPs are funded from one or several international NGOs such as Oxfam, Save the
Children, and Action Aid. Although the overall goal of the programmes surveyed is poverty
reduction, their specific objectives differ, ranging from improving childhood nutrition to rural
infrastructure development. Therefore, microfinance is just one component used by NGOs to
alleviate poverty. A factor common to all these programmes is that implementation is assisted
through partnerships with a local Women’s Union, one of the five mass organisations in
Vietnam.4 Membership of these NMPs includes households that are poor5 and meet the selection
criteria of other NGOs projects (e.g., families with malnourished children, having no latrine for
sanitation, or no clean water sources). In addition, all NMPs apply group lending with
instalments and compulsory saving requirements. The operational scale of NMPs differs greatly;
some have only several hundred members (they operate in only a few villages) while others have
more than 10,000 members. We purposely selected NMPs that have operated for at least three
years. The maximum loan size also varies considerably among NMPs, ranging from 0.5 to 15
million Dong (about $US33 to $US1000). Loan interest rates from NMPs vary from 0.8 to 1.7
per cent per month. In the NMPs surveyed, all members had borrowed at least once, and 95 per
cent of members had rolled over their loans at least once.
Although preparation started in January 2004, the survey was conducted from March to June
2004 using a two-stage sampling approach. In the first stage, the pool of control and treatment
villages were identified by selecting 10 NMPs that planned to expand their operations and which
had been in operation for at least three years. The sampling frame in this stage was constructed
from the list of some 80 programmes that operate in the north and central regions selected from
the NGO directory in Vietnam (NGOR, 2003). Eighty NMPs were contacted and 44 were found
to be active and willing to participate in the study. All 44 eligible NMPs were surveyed at the
programme level (i.e., interviews with managers of headquarters offices and collection of
secondary data from their records) for a separate study to examine the efficiency of their
operations.
The current study focuses on the effects of microfinance at the household level, where a sub-
sample of 10 eligible NMPs (in operation for at least three years with a plan for possible
expansion) were selected. From each of these 10 NMPs, we randomly selected one member
village and at least one non-member village with similar characteristics (for location,
infrastructure and poverty rate), which had not received microfinance services from NMPs. At
the end of this stage, 10 control villages and 15 treatment villages in the project sites of 10 NMPs
in the north and central regions of Vietnam were selected.
In the second stage, households were sampled randomly from a list of eligible households in
each village. For member villages, these eligible households were selected from member lists of
NMPs while ineligible households were not included on the member list. For non-member
villages, eligible households were those meeting the poverty and additional criterion of the
NGO projects while ineligible households did not meet the criteria (see Table 1). It was
planned to select 20 households per village, however some households could not be found or
had no adult at home and hence could not be interviewed. Thus, the total number of
households interviewed was 471, or 19 households per village on average. The sampling rate is
624 S. Nghiem et al.

Table 1. The sampling frame for the household survey

Villages that meet the selection criteria of microfinance programmes


(poor villages and lack of access to microfinance services)
Have received microfinance services Have not received microfinance services
(treatment villages) (control villages)
Eligible Ineligible Eligible Ineligible
Households Households Households Households
(Group 1) (Group 2) (Group 3) (Group 4)
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

20 per cent overall, where eligible households were over-sampled at 26 per cent while the
average sampling rate of ineligible households was six and nine per cent in member and non-
member villages, respectively.6
The household survey applied a choice-based sampling technique so eligible households were
over-sampled. This approach allows the researcher to obtain reliable results while minimising
data collection costs (Lancaster and Imbens, 1991; Imbens, 1992). With the available informa-
tion on the populations of eligible and ineligible households, the conversion of estimates to full
population estimates is straightforward using sampling weights, which are the inverse of the
probabilities of being sampled in each stratum.

3.1 Choice of variables


The effects of the contribution of microfinance to household welfare were identified by examining
the relationship between the duration in microfinance and welfare indicators, after controlling
for characteristics of households and villages. The choice variables representing these factors are
discussed briefly below.

3.1.1 Welfare indicators. As the ultimate goal of microfinance is to improve the livelihood of the
economically active poor, the effects of microfinance should be measured by changes in the
welfare of clients. Many variables can be used to measure a change in welfare, such as income
and consumption (proxies for economic wellbeing), education and health spending (proxies for
human capital) and spending on social events (a proxy for social capital). We focus on economic
indicators, including income and consumption per adult equivalent, taking into account age and
sex differences.

3.1.2 Household characteristics. The household characteristics considered include the depen-
dency ratio, number of people of working age (16–60 years), arable land, average education level
of each person eligible to participate in the workforce and age of the household head. The
dependency ratio is the number of people outside the working age range divided by the number
of people of working age; households with a higher dependency ratio may have more difficulty
improving their living standards. Households with a higher capital stock, especially non-
residential capital stock, may be able to generate higher income. A dummy variable representing
a shock (e.g., illness, burglary and fire) in the survey year is included as such events could have a
considerable impact on household income and consumption.

3.1.3 Village characteristics. The village characteristics that are likely to influence household
welfare include the availability of electricity, road quality and the distance to a township.
Electricity and road conditions may have positive influences on income and consumption. The
distance to the nearest township, obtained from the village information sheet filled by the village
head, captures the influence of location on household welfare. Villages located closer to a
The Welfare Effects of Microfinance in Vietnam 625

township may have easier access to a marketplace and off-farm job opportunities, and hence
higher income and consumption.
The economic performance of households could also be affected by market prices. The
price of rice (the main product of most households surveyed) and the casual wage rate
were selected.7 However, the direction of price effects on household income and
consumption is not clear. If households sell rice and/or work as hired labour, then a rise
in prices will have a positive effect upon household income, otherwise negative effects may
occur.

3.1.4 Eligibility and treatment variables. The eligibility dummy variable (equal to one for
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

eligible households and zero otherwise) is a function of the unobservable characteristics that
permit households to join microfinance programmes. This will have a negative sign if micro-
finance targets poor households.
The duration of member participation in microfinance is used as the treatment variable
because it has advantages over using the amount of funds loaned. If the total amount of the loan
is used one should use total loans of the family from all sources, as money is fungible (one cannot
distinguish the contribution of microfinance loans and loans from other sources to household
welfare). In addition, NMPs not only provide credit but also other development services such as
literacy and health care. Hence, using total loans will implicitly assume that there is no effect
from other integrated services. The loan size variable faces further difficulties because it requires
a choice between the current outstanding loan and the total amount of lending since joining the
NMP. The former option does not reflect the cumulative impacts of microfinance while the latter
was difficult to obtain because rural households often did not keep an accurate record of loans,
especially those from other sources and NMPs. Therefore, this study uses the duration (in
number of months) that households are members of NMPs as a treatment variable; this is easier
to collect, reflects the cumulative nature of microfinance effects, and can mitigate the issue of
money fungibility.

3.2 Descriptive statistics


The sample means (see Table 2) show that between the four groups many household
characteristics are similar. For example, the household heads surveyed are all approximately 40
years of age and live in a family of five with two people employed in the labour force. Other
summary data revealed that the majority of households owned their houses and around two-
thirds of households own at least one television set (TV) and one-third own at least one
motorcycle.8
All four groups use loans from external sources for their financial needs, whether or not they
participated in microfinance programmes. In addition, the amount borrowed among the four
groups did not differ considerably. Since ineligible groups in this study were able to access other
form of credits, it is possible that impacts of NMP credit are minimal. Mean income and
consumption suggests that eligible groups are likely to be poorer than ineligible groups, which
reflects the fact that NMPs target poor households.
One observation of interest is that in member villages the total loans (from all sources such as
microfinance, banks and relatives) of eligible households was larger than that of ineligible
households while the reverse applies in non-member villages. The availability of microfinance
services may have improved the creditworthiness of members; they appear to access credit from
other sources to the same extent as ineligible households in the same village.9 Although
microfinance loans comprise only a small proportion10 of the total debt of member households
they can create some cumulative positive effects.
The means of the main household welfare indicators also suggest that access to microfinance
may create positive effects for its clients. For example, income and consumption of eligible
households in member villages is respectively 18.8 and 16.6 per cent higher than the relevant
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

Table 2. Means of main variables from the household survey

Total Group1 Group 2 Group 3 Group 4


(n ¼ 471; (n ¼ 237; (n ¼ 41; (n ¼ 164; (n ¼ 29;
626 S. Nghiem et al.

N ¼ 2538) N ¼ 919) N ¼ 665) N ¼ 624) N ¼ 340)


Variables Unit/Description
Mean Std. Mean Std. Mean Std. Mean Std. Mean Std.

Age Years 40.66 8.82 40.21 8.09 39.88 8.84 41.60 10.00 40.10 7.30
Education level of household head Years 7.13 1.89 7.19 1.77 7.24 1.85 6.96 2.08 7.38 1.76
Average education level of people in labour age Years 7.12 1.57 7.18 1.49 7.29 1.35 6.95 1.73 7.47 1.61
Household size Persons 4.73 1.45 4.86 1.44 4.54 1.47 4.61 1.50 4.66 1.29
Number of labourers Persons 2.36 0.94 2.38 0.93 2.20 0.84 2.40 1.02 2.17 0.60
Dependency ratio Dependents/labour 1.18 0.80 1.22 0.83 1.25 0.84 1.08 0.78 1.23 0.68
Land per capita M2/person 1200 481 1214 286 762 258 1270 625 1298 728
Number of loans Loans 1.44 1.32 1.76 1.22 1.12 1.05 1.07 1.35 1.31 1.65
Total borrowings Thousand Dong 2888 3663 3163 3529 2246 2443 2525 3867 3603 4703
Microfinance participation Months 23.52 32.71 51.77 29.87 0.00 0.00 0.00 0.00 0.00 0.00
Total household income Thousand Dong/year 13922 11057 14491 10298 18206 17818 12198 10125 12952 7749
Income per adult equivalent Thousand Dong/year 4226 3327 4272 2995 5942 6262 3763 2708 4042 2423
Total household consumption Thousand Dong/year 10817 7585 10847 5793 15746 12151 9306 8114 12159 6156
Consumption per adult equivalent Thousand Dong/year 3307 2144 3244 1719 5086 4157 2849 1640 3901 2385
Household net income Thousand Dong/year 3104 8365 3645 9296 2460 10322 2893 6326 793 7224
Dummy variables
Sex 1 ¼ male; 0 ¼ female 0.38 0.48 0.34 0.48 0.32 0.47 0.41 0.49 0.52 0.51
Profession 1 ¼ farmer; 0 ¼ others 0.94 0.23 0.94 0.24 0.90 0.3 0.95 0.22 1.00 0.00
Owned a house 1 ¼ yes, 0 ¼ no 0.95 0.22 0.95 0.21 0.98 0.16 0.93 0.25 1.00 0.00
Owned a TV 1 ¼ yes, 0 ¼ no 0.68 0.47 0.73 0.45 0.78 0.42 0.62 0.49 0.55 0.51
Owned a motorcycle 1 ¼ yes, 0 ¼ no 0.34 0.47 0.30 0.46 0.41 0.5 0.36 0.48 0.41 0.50

Note: Group 1 ¼ Eligible households in member villages; Group 2 ¼ Ineligible households in member villages; Group 3 ¼ Eligible in non-member villages; Group
4 ¼ Ineligible households in non-member villages.
The Welfare Effects of Microfinance in Vietnam 627

households in non-member villages. The relative figures for income and consumption per adult
equivalent are 13.5 and 13.9 per cent, respectively. The descriptive results also show a skewed
distribution, suggesting that logarithmic transformations could be a sensible choice in our
regression analysis (Singh and Maddala, 1976).

4. Econometric Analysis
4.1 Model Specification
As a logarithmic functional form is selected to address the skewed distribution of income and
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

consumption, the presence of zero values for variables such as duration in NMPs can be
problematic. Battese (1997) proposed a clever application of dummy variables to obtain
unbiased estimates, using dummy variables for the regressors with zero values and replacing the
undefined logarithms of zero with zeros. Applying this treatment, the effect of microfinance on
household welfare is specified as:

ln Yij ¼ b0 þ b1 Eij þ b2 VTj þ b3 ln Tij þ b4 Xij þ b5 Vj þ b6 Dij þ mij ð1Þ

where:

lnYij is the log of household welfare indicators such as income or consumption of household i in
village j;
Eij is the eligibility dummy variable, equal to one for eligible households (in both the control and
treatment villages) and zero for ineligible households;
VTj is the dummy variable for village type, equal to one for member villages and zero otherwise;
Tij ¼ Max(Tij, Dij) where Tij is the duration (months) that households participated in NMPs;
Xij ¼ ln Xij if Xij40 and Xij ¼ 0 otherwise, where Xij is a vector of household characteristics such
as age, sex, education and labour;11
Vj ¼ ln Vj if Vj40 and Vj ¼ 0 otherwise, where Vj is either a vector of variables (for village


characteristics model), or a vector of village dummy variables (for village fixed-effects model);
Dij is the vector of Battese (1997) dummy variables for household and village characteristics with
zero values before taking the logarithm;
mij is the idiosyncratic error term; and
b1, b2 and b3 are parameters, and b4, b5 and b6 are vectors of parameters to be estimated.

The use of the eligibility dummy variable Eij can mitigate the self-selection issue because it
captures unobservable characteristics that made eligible households decide to join microfinance
programmes. Likewise, the village type dummy variable VTj captures unobserved characteristics
that make some eligible villages selected to join NMPs before others. In the village-fixed effect
model, where distinctive characteristics (observed and unobserved) of each village are captured
by the village dummy, VTj proxies for spillover effects because it captures the difference in
outcomes of ineligible households between members and non-member villages (see Table 3). One
source of this difference is the spillover effect from members to non-member households in
programme villages. This difference can be positive (e.g., non-member household can learn
financial management skills from member households) or negative (e.g., a new business set-up by
member households takes customers from non-member households).
Note that there are still two choices of specification: village characteristics or village-fixed
effect, depending on whether or not NMPs selected eligible villages randomly. If NMPs did so
then the village characteristics specification of Equation (1) can be estimated efficiently and
consistently using ordinary least squares (OLS). Otherwise, OLS estimates are inconsistent due
to unobservable characteristics of member villages, leading to a correlation between Tij and mij in
Equation (1). In the village fixed-effects model, however, village dummies can capture all
628 S. Nghiem et al.

Table 3. An approximate decomposition of microfinance effects

Member villages Non-member villages


(VTj ¼ 1) (VTj ¼ 0) Difference

Eligible b0þb1þb2þb3 b0þb1 b2þb3 Effects measured


households without ineligible
(Eij ¼ 1) households
Ineligible b0þb2 b0 b2 Spillover effects
households
(Eij ¼ 0)
Difference b1þb3 b1 b3
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

Effects measured Unobservable ‘Pure’ microfinance effects


without characteristics of
control villages eligible households

observable and unobservable characteristics within each village, and hence provide consistent
estimates, regardless of whether or not eligible villages received microfinance services randomly.
A generalised Hausman test was applied as proposed by Weesie (1999) and extended by Creel
(2004) to determine whether Tj and mij are correlated. The test results rejected the null hypothesis
of no correlation; therefore the village fixed effects model is selected in this study.12 A limitation
of this choice is that the factors at village level that contribute to income and consumption of
households were not determined. However, the village fixed effect model allows us to estimate
spillover effects, which is one of the factors that make outcomes of ineligible households in
member villages differ from those in non-member villages.

4.2 Effects of Microfinance


The results in Table 4 show that access to NMP microfinance has no statistically significant effect
on income and consumption of member households. One possible interpretation is that the
contribution of microfinance services concentrates on smoothing rather than increasing income
and consumption levels. Another possible reason is that NMP microfinance is irrelatively small
compared to the total income or assets of a household. The average loan size from NMP
microfinance programmes accounts for only four per cent of the average income of member
households (derived from Table 3).13 Finding a significant effect of microfinance on household
income may be an ambitious exercise under these circumstances because we focus only on NMP
credits while other forms of credit may be prevalent and more important.
To test for the sensitivity of variable specification another estimate defines the duration in
NMPs as a categorical variable (the duration variable was coded arbitrarily as 0 ¼ no
participation, 1 ¼ participation for 1 year, 2 ¼ participation for 2–3 years, 3 ¼ participation for
4–5 years, and 4 ¼ participation for more than 5 years). This yielded a similar result –
participants in NMP microfinance programmes did not have significantly higher income or
consumption compared with non-participants.
This insignificant result may be because we control for selection biases and non-random
programme placement using an eligibility dummy at both household and village levels. As a
comparison, a ‘naive’ model was estimated with a village characteristics model ignoring the
survey design. In other words, the ‘naive’ model did not take any measure to control for the
selection bias issue. The results are more promising in the sense that income per adult equivalent
increases significantly (at 5%) by 7700 Dong for an additional month of participation in an
NMP. In the categorical specification of duration we also obtained significant results in income
for those participating in microfinance for 4–5 years and in consumption for those participating
in microfinance for more than five years.14 Nevertheless, the magnitude of the impact is minimal:
compared with non-participants, the income of those who participated for at least three years
The Welfare Effects of Microfinance in Vietnam 629

Table 4. Effects of microfinance on income and consumption

Log of income per Log of consumption per


adult equivalent adult equivalent
Independent variables Coefficient t-ratio Coefficient t-ratio

Log of duration 70.06 70.59 0.01 0.15


Eligible households 70.12 71.33 ***70.46 74.22
Participating villages 0.22 1.23 *0.28 1.70
Log of dependency ratio 0.05 0.93 0.04 0.97
Log of household size ***70.43 73.82 ***70.51 74.66
Log of productive land ***0.26 4.94 ***0.16 5.21
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

Log of physical capital ***0.27 5.36 **0.13 2.47


Log of labour education ***0.36 2.91 ***0.27 3.35
Log of age ***0.39 2.69 ***0.40 4.23
Female-headed household *70.14 71.69 **70.14 72.19
Shock dummy 70.08 70.90 0.04 0.61
No dependent dummy 0.20 0.78 0.11 0.47
No land dummy ***1.22 3.35 0.30 1.06
No participating dummy 70.18 70.47 70.02 70.06
Constant **1.86 2.29 ***4.22 6.00
R-squared 0.54 0.63
N 471 471

Note: Robust standard errors are used. ***, **, and *represent 1, 5 and 10 per cent significant level,
respectively. Village dummies were dropped to reserve space. Variables with zero values are treated using
the approach of Battese (1997).

increased by 0.35 per cent, and the consumption of those participating for at least five years
increased by 0.19 per cent.
The negative sign of the household eligibility dummy confirms that these households are
relatively poorer than ineligible households, although it is only significant for consumption.
The positive sign of the village type dummy (only significant at 10% in consumption
estimate) suggests some minor spillover effects from member to non-member households in
participating villages. Due to the dummy nature of these variables, these coefficients can be
interpreted as the percentage difference between the two groups concerned after controlling
for household characteristics. For example, the consumption per adult equivalent of
eligible households is, on average, 0.46 per cent significantly lower than that of ineligible
households.
The coefficient on the eligibility dummy is larger (in absolute value) on the consumption side,
suggesting that eligible households have more modest consumption although their income is not
far below other households. This finding supports the theory that the poor are more risk-averse,
and hence poor eligible households in NMPs may apply a more modest consumption bundle to
save some resources as a precaution. However, squeezing consumption too much may dampen
future income, due to a decrease in productivity (e.g., due to ill heath, malnutrition) and lack of
investment, meaning that the poor tend to be trapped in poverty.
Other explanatory variables are significant determinants of income and consumption. Larger
households have significantly lower income and consumption per adult equivalent. One possible
explanation is the high seasonality of agricultural production, thus the average return may be
lower for large families when it is difficult to find off-season employment. Production land is
positively related to income and consumption, confirming that land is the primary input of rural
households. Since most rural households are involved in agricultural production, it is reasonable
that having more land could generate higher income and consumption. Another important input
for rural households is physical capital,15 which has a positive and significant effect on both adult
equivalent income and consumption.
630 S. Nghiem et al.

The two proxies for human capital, the average education level of people of labour age
(proxies for knowledge) and age of household head (proxies for experience), have a significant
and positive influence on income and consumption. Households with more educated labour have
a higher capacity or ability to learn and apply new skills, which can lead to an improvement in
income and consumption. Older household heads appear to be in more productive households.

5. Concluding Remarks
This study provides an analysis of the effects of NGO microfinance on household welfare in
Vietnam, in terms of income and consumption. The duration of participating in NMP
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

microfinance is the treatment variable; this is expected to capture the cumulative effects of
microfinance and other development activities provided by NMPs. The quasi-experimental
survey aims to compare the welfare of NMP microfinance members with non-member
households of similar characteristics. A generalised Hausman test showed that the order in
which eligible villages received microfinance might not be entirely random so a village fixed-
effects model was utilised (this can provide consistent estimates, regardless of the order in which
villages receive services).
The contribution of NMP microfinance to household welfare is statistically insignificant. We
hypothesise that the small sizes of NMP microfinance loans (relative to the total loans and total
income of households) has constrained the potential contribution to household welfare. Overall,
NGO microfinance programmes may not be as effective as some anecdotal evidence has
indicated. However, we only considered NGO (NMP) microfinance; this may have no significant
impact because it targets poorer households whereas other (non-eligible or control) households
have access to other sources of microfinance (possibly in larger amounts). Thus, NGO
microfinance may play a role in providing the poor with access to credit for consumption
smoothing even if not at the scale required to impact on incomes. Also, this study did not
examine impacts of NMPs on non-monetary welfare, which may be created by, for example,
smoothing the household consumption pattern.

Notes
1. It is possible to build a simple analytical model of agricultural households as a basis for the econometric model. As
the focus is on a suitable survey framework, the model is not presented (it is available from the authors on request).
2. Although Hanoi is the capital city of Vietnam, the survey site covers only Soc Son district, which is still a poor and
agriculture-dominant area.
3. This study was in combination with another study involving the sensitive issue of ranking operational performance of
NMPs, so names of these NMPs are not revealed.
4. The other four mass organisations are Farmers’ Union, Youth Union, Elderly Association and Veterans’
Association. Most were established around the early 1940s to mobilise all sectors of the population to participate in
the struggle for the independence of Vietnam. All mass organisations have a complete network from central to village
level. Although mass organisations are classified as non-governmental civic organisations, they have strong links with
the government (e.g., leaders of mass organisations receive salaries from the government’s budget).
5. Most NMPs applied participatory wealth ranking (PWR) technique to identify poor households. Essentially, this
technique classified all households in the village into several wealth groups (e.g., rich, middle and poor) by taking the
average ranking results of 5–7 representative villagers. Most village heads interviewed confirm poor households
ranked by PWR often also belonged to the poor group as ranked by the national poverty line.
6. See the sample (n) and population (N) in the first row of Table 2.
7. These prices were obtained from the ‘village information sheet’ completed by village heads.
8. Rental property is almost a missing market in rural Vietnam because it is relatively cheap and easy to construct a
rural shelter while land is allocated by the government. TVs and other housing accessories do not account for quality
(e.g., TVs may be old cheap brands), so households are not as ‘rich’ as one may think.
9. The average loans from all sources of the two groups in the treatment villages were 3.163 (Group 1) and 2.246 (Group
2) million Dong, respectively. The average amount of loans mobilised outside NMP microfinance of member
households is roughly the total loans of eligible households in non-member villages (Group 3 ¼ 2.545 million Dong,
similar to the figure of Group 2).
The Welfare Effects of Microfinance in Vietnam 631

10. An estimate of this proportion is (3.163–2.545)/3.163 ¼ 19.5 per cent as the difference between total loan of Group 1
(3.163) and Group 3 (2.545), roughly the loan from microfinance.
11. Note that no transformation is needed for dummy variables in the household and village characteristics.
12. Details of the test results are available on request.
13. From Table 3, the ratio of average microfinance loan and household income is roughly calculated as the difference
between the average total loan of households in Group 1 and Group 3, divided by household income of Group 1
(3.163–2.525)/14.491 ¼ 4.4 per cent.
14. The results are available on request.
15. Physical capital is measured by the estimated current value of production equipment (e.g., oxcart, tractor, threshing
machine, and shed), livestock (e.g., bullock and sow), and transport means (e.g., motorcycles and bicycles). The
evaluation of capital stock was by counting capital stock items at the household survey and price information
collected from the survey with village heads.
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

References
Armendariz, B. and Morduch, J. (2010) The Economics of Microfinance (Cambridge, MA: MIT Press).
Banerjee, A., Duffo, E. Glennerster, R. and Kinnan, C. (2009) The miracle of microfinance? Evidence from a randomized
evaluation, Working Paper, Abdul Latif Jameel Poverty Action Lab, Massachusetts Institute of Technology, USA.
Barslund, M. and Tarp, F. (2008) Formal and informal rural credit in four provinces of Vietnam. Journal of Development
Studies, 44(4), pp. 485–503.
Bateman, M. (2010). Why Doesn’t Microfinance Work? The Destructive Rise of Local Neoliberalism (London: Zed Books).
Battese, G.E. (1997) A note on the estimation of Cobb-Douglas production functions when some explanatory variables
have zero values. Journal of Agricultural Economics, 48(2), pp. 250–252.
Chemin, M. (2008) The benefits and costs of microfinance: evidence from Bangladesh. Journal of Development Studies,
44(4), pp. 463–484.
Coleman, B.E. (1999) The impact of group lending in northeast Thailand. Journal of Development Economics, 60(1), pp.
105–141.
Creel, M. (2004) Modified Hausman tests for inefficient estimators. Applied Economics, 36(21), pp. 2373–2376.
Deaton, A. (2010) Instruments, randomization, and learning about development. Journal of Economic Literature, 48(2),
pp. 424–455.
Harris, S.D. (2009) The State of the Microcredit Summit Campaign Report 2009 (Washington, DC: Microcredit Summit
Campaign).
Heckman, J.J. and Urzua, S. (2010) Comparing IV with structural models: what simple IV can and cannot identify.
Journal of Econometrics, 156(1) pp. 27–37.
Imbens, G.W. (1992) An efficient method of moments estimator for discrete choice models with choice-based sampling.
Econometrica 60(5), pp. 1187–1214.
Kaboski, J. and Townsend, R. (2011) A structural evaluation of a large-scale quasi-experimental microfinance initiative.
Econometrica 79(5), pp. 1357–1406.
Karlan, D. and Zinman, J. (2010) Expanding microenterprise credit access: using randomized supply decisions to estimate
the impacts in Manila. Working Paper, Abdul Latif Jameel Poverty Action Lab, Massachusetts Institute of
Technology.
Khandker, S.R. (2005) Microfinance and poverty: evidence using panel data from Bangladesh. World Bank Economic
Review, 19(2), pp. 263–286.
Lancaster, T. and Imbens, G.W. (1991) Choice-based sampling – inference and optimality, Manuscript, Department of
Economics, Brown University, USA.
McCarty, A. (2001a) The Institutional and Legal Structure of Microfinance in Vietnam (Hanoi: DFID and the State Bank
of Vietnam).
McCarty, A. (2001b) Microfinance in Vietnam: A Survey of Schemes and Issues (Hanoi: DFID and the State Bank of
Vietnam).
Morduch, J. (1998) Does microfinance really help the poor? New evidence on flagship programmes in Bangladesh.
Mimeo, McArthur Foundation Project on Inequality Working Paper, Princeton University.
NGOR (2003) NGO Directory (Hanoi, Vietnam, NGO Resource Centre).
Pham, B.D. and Izumida, Y. (2002). Rural development finance in Vietnam: a microeconometric analysis of household
surveys. World Development, 30(2), pp. 319–335.
Pitt, M.M. (1999). Reply to Jonathan Morduch’s ‘Does microfinance really help the poor? New evidence from flagship
programs in Bangladesh’. Mimeo, Department of Economics, Brown University, USA.
Pitt, M.M. and Khandker, S.R. (1998) The impact of group-based credit programmes on poor households in Bangladesh:
does the gender of participants matter? Journal of Political Economy 106(5), pp. 958–996.
Roodman, D. and Morduch, J. (2009) The impact of microcredit on the poor in Bangladesh: revisiting the evidence.
Working Paper No. 174, Washington, DC: Center for Global Development.
Singh, S.K. and Maddala, G.S. (1976) A function for size distribution of incomes. Econometrica, 44(5), pp. 963–970.
632 S. Nghiem et al.

UNDP (1996) Microfinance in Vietnam: A Collaborative Study based upon the Experiences of NGOs, UN Agencies and
Bilateral Donors (Hanoi, United Nations Development Program).
Weesie, J. (1999) Seemingly unrelated estimation and the cluster-adjusted sandwich estimator. Stata Technical Bulletin,
52, pp. 34–47.
Wolz, A. (1997) The transformation of rural finance system in Vietnam. Discussion Paper No. 60, Research Centre for
Agrarian and Economics Development, Heidelberg, Germany.
World Bank (2002) Vietnam Banking Sector Review, Financial Sector, East Asia and Pacific Region (Washington, DC: The
World Bank).
Journal of Development Studies 2012.48:619-632. downloaded from www.tandfonline.com

You might also like