Yet, there is not much rigorous evidence on the borrower impact of both types of microcredit to either substantiate or challenge such a strategic shift. We present some such evidence from a randomised field experiment in Mongolia.

The ability of microcredit to combat poverty remains hotly debated. After years of rapid growth, various microfinance institutions (MFIs) are currently struggling with repayment problems and, in some cases, a political backlash.

Scepticism has been further fuelled by several randomised field experiments which show that the capacity of microcredit to lift people out of poverty might be smaller than previously thought. In a nutshell, the evidence suggests that microcredit may reduce liquidity constraints, help families cope with shocks, and encourage entrepreneurship. The ultimate impact on poverty indicators such as income and consumption nevertheless remains ambiguous. Impacts on health and education are difficult to substantiate too.

Learning about the impact of microcredit is also important because the microfinance industry itself is changing. A number of leading MFIs have moved from joint-liability lending, as pioneered by Grameen bank in the 1970s, to individual lending. Under joint liability, small groups of borrowers are responsible for the repayment of each other’s loans. Group members are treated as being in default when at least one of them does not repay and all members are denied subsequent loans. Group lending often involves committing to repayment meetings and can exploit social pressure, making it onerous for borrowers. This is a key reason why MFIs are moving from joint to individual lending.

Somewhat surprisingly, there exists very limited evidence on the relative merits of individual and group lending in terms of borrower impact. Armendáriz and Morduch (2005), p. 101-102) note that: “In a perfect world, empirical researchers would be able to directly compare situations under group-lending contracts with comparable situations under traditional banking contracts. (…). The best evidence would come from well-designed deliberate experiments in which loan contracts are varied but everything else is kept the same.” This blog post discusses such evidence (Attanasio, Augsburg, De Haas, Fitzsimons, and Harmgart, 2011).

The experiment

Mongolia is the most sparsely populated country in the world and this makes disbursing, monitoring, and collecting small loans very costly. The aim of our experiment, conducted in cooperation with Mongolia’s XacBank, was to analyze whether group lending can be an effective and efficient way to lend. Mongolian microcredit has traditionally been provided as individual loans, reflecting concerns that the nomadic lifestyle of indigenous Mongolians had impeded the build up of social capital.

Our experiment took place in 40 villages (Figure 1).

Overview of participating villages and provinces

XacBank was interested in expanding access to poor and female borrowers, an underserved market segment. A total of 1,148 women from the poorest parts of the population participated and a detailed face-to-face survey was administered to each of them during March-April 2008 (baseline survey, see first photo.

Participant being interviewed, photo taken by Britta Augsburg

We measured variables that reflect households’ living standards and that could in principle be affected by the intervention during a 1.5 year interval: income, consumption, and savings; entrepreneurial activity and labour supply; asset ownership and debt; and informal transfers.

After the baseline survey, we randomised at the village level: women in 15 villages received access to individual loans, to group loans in another 15 villages, while in 10 control villages XacBank did not lend to the women during the experiment. The randomisation removed selection bias, allowing us to attribute post-treatment differences in outcomes to the two lending programs.

The ‘treatment period’ during which XacBank disbursed loans lasted 1.5 years – from April 2008 to September 2009 – with some variation across villages. During this period, 57 (50) per cent of the respondents in the group (individual) lending villages borrowed from XacBank. The probability of receiving a microloan during the experiment was 24 percentage points higher in treatment than in control villages.

In October-November 2009 we conducted a follow-up survey to measure the poverty status and economic activity of all women again. We use the data of both survey rounds to measure the impact of the programs on poverty by comparing all women who initially signed up in treatment villages, irrespective of whether they borrowed or not, with those who signed up in control villages.

Group lending versus individual lending: similarities…

Although XacBank’s loans were intended to finance business creation, about half of all credit was used for household rather than business purposes in both the group- and individual-lending villages. For instance, we find that at the end of the experiment the probability of owning a VCR or radio was 17 and 14 per cent higher in the group- and individual-lending villages, respectively (compared to control villages). For large household appliances the corresponding figures are 9 and 7 per cent.

A second finding that holds for both treatment programs is that women with lower education seem to benefit more. We take education as a proxy for long-term poverty, more reliable than a wealth indicator as it is easier to measure and more stable over time. The results suggest that it is the poorer part of the targeted population that benefits most from microcredit, regardless of how it is delivered.

Third, we find no differences in repayment behaviour between both lending programs. Giné and Karlan (2010) also compare repayment rates between group and individual lending – both with mandatory weekly repayment meetings – and find no significant differences. In our case neither loan programme included mandatory repayment meetings.

… and differences

We also find important differences between the impact of group and individual loans, to suggest that the former were more effective. For group loans we find a positive impact on female entrepreneurship, one of the main intermediate objectives of the programs. This is largely driven by less-educated women who at the end of the experiment had a 29 per cent higher chance of operating a business compared to similar women in control villages. This difference is 10 per cent for highly educated women. Enterprise profits increase over time as well.

Did increased entrepreneurial activity feed through to improved household well-being? To answer this question we use detailed information on household consumption elicited in the surveys. We find a significant and robust increase, relative to control villages, in food consumption in group-lending villages.  Access to group loans led to more and healthier food consumption, in particular of fresh items such as fruit, vegetables and dairy products.

Total food consumption was 17 percentage points higher. Over time we also see an increase in the use of combustibles and felt for the isolation of gers – traditional Mongolian felt tents (Photo 2) – as well as other non-durable and total consumption.

Two group-borrowers in front of one of their gers, in Ikhtamir village in Arhangay province, photo by Ralph De Haas

Our findings for individual lending suggest that this form of lending was simply not as effective. We find no impact on female entrepreneurship or on consumption, not even with increased exposure to credit. We do find, however, that over time there is an increase in the probability that women operate a business jointly with their spouse – and that these joint enterprises gradually also become more profitable. Nevertheless, it is not clear whether these longer-term effects translate in the same way into higher consumption as they do for group borrowers. We find no evidence that food consumption goes up with exposure in individual-lending villages.

There is at this stage no evidence of changes in income as a result of either of the programs, though it may simply be too early to observe such effects. The more sustained and more generalised increase in consumption in group-lending villages seems to indicate that these loans are more effective at increasing permanent income. Why?

One possibility is that joint-liability ensures better discipline. Group discipline may not only prevent the selection of overly risky investment projects, it may also ensure that a substantial part of the loans is actually invested in the first place. We document results on informal transfers that seem to support this hypothesis: women in group-lending villages decrease their transfer activities with families and friends, the opposite to what we find in individual-lending villages. This could reflect that groups replace some of their informal financial networks but further analysis is needed to explore this.

Our weaker results for individual loans may also reflect that borrowing at baseline (i.e. pre-program) was somewhat higher in individual compared to group-lending villages. Moreover, since group-lending was an innovation in Mongolia, the unmet demand for this product – and its marginal impact – may have been larger. Loan take-up was indeed higher in group-lending villages. This could indicate that some women, in particular the less-educated, had not been comfortable with borrowing alone but were willing to borrow as part of a group. This would imply that group and individual lending are complementary services for which the demand differs across borrower types. The process of liability individualisation by MFIs may therefore run the risk that certain borrowers, those who are not able or willing to borrow on their own, may gradually lose access to finance. It is too early to write off group lending just yet.

B. Armendáriz and J. Morduch (2005), The Economics of Microfinance, MIT Press, Cambridge.

O. Attanasio, B. Augsburg, R. De Haas, E. Fitzsimons, and H. Harmgart (2011), Group lending or individual lending? Evidence from a randomised field experiment in Mongolia, EBRD Working Paper No. 136, London.

X. Giné and D. Karlan (2010), Group versus individual liability: Long-term evidence from Philippine microcredit lending groups, mimeo.

Share on Facebook

Ger – a portable, felt-covered dwelling – in rural Mongolia

Although microfinance has grown rapidly over recent years, hard evidence on its socio-economic impact is only emerging slowly. To what extent does microfinance lift people out of poverty by allowing them to generate income from small-scale enterprises? And is group lending (‘joint-liability’) or individual lending the best way to reach out to borrowers? These are questions that XacBank, a leading microfinance institution in Mongolia, has been grappling with as well. The bank wanted to expand its outreach to indigent rural borrowers, in particular female ones, who hitherto had only limited access to financial services. But what is the best way to expand lending to such ‘difficult’ customers?

On the one hand, individual loans may be more suitable in a country in which the nomadic lifestyle may have limited the build up of social capital outside of the family structure. On the other hand, group lending may work well if the looser ties within groups reduce the risk of collusive behaviour. Moreover, since monitoring costs are particularly high – loan officers have to travel extremely large distances to reach remote clients – the reduction of such costs through a group lending structure may be particularly valuable.

To help XacBank with its strategic decision making and to assess empirically the impact of access to microfinance on small business development and poverty reduction, a project team at the EBRD (Ralph De Haas and Heike Harmgart) and the Institute for Fiscal Studies (Orazio Attanasio, Britta Augsburg and Emla Fitzsimons) designed a so-called randomised field experiment. The design entails an experimental set up involving 40 soums (villages) across 5 aimags (provinces). Together with the Mongolian Women’s Federation (MWF) a list was drawn up in each village with the names of relatively poor women who were interested in a XacBank loan to expand or set up a small business. These women were also asked to form preliminary borrowing groups. All 1,148 of them were then interviewed by a survey company in March 2008 (the baseline survey).

Session of MWF representatives involved in the field experiment

The 40 soums were then randomly divided into 10 ‘control soums’, 15 ‘group lending soums’ and 15 ‘individual lending soums’. Information from the baseline survey confirmed that the randomisation worked well: the participating women in all three types of villages were on average very similar across a wide range of observable characteristics.

In a next step, all women in group-lending soums were visited by a XacBank loan officer and groups that were deemed credit-worthy were offered a group loan. In the individual lending soums, women were offered an individual loan, while in the control soums XacBank delayed the roll-out of lending for the duration of the experiment. Importantly, when the women signed up to the project it was carefully explained to them that they would only have a 75 per cent change of actually obtaining a loan during the first year (since XacBank would delay the introduction of lending in 10 out of 40 villages).

A follow-up survey was conducted in October/November 2009, about 20 months after the baseline survey. Four interview teams re-interviewed 982 of the initial respondents; a re-interview rate of 86 per cent. This means that for 982 respondents we have detailed information from both the baseline and follow-up surveys on income, consumption and saving patterns, asset ownership, (in)formal enterprises, and exposure to shocks. Respondents were also asked about their income expectations and attitudes towards risk. Finally, novel questions were asked to gauge how well group members knew their co-borrowers, with the aim of allowing us to make inferences about the ‘information asymmetries’ within joint-liability groups. Detailed information was also gathered on the characteristics of villages and loan officers that participated in the experiment, while XacBank provided the project team with comprehensive repayment data on all loans.

Respondent being interviewed. The stones are used to answer questions about the respondent’s own expectations about her future income.

The project team is currently combining and analysing all of these data. This should allow us to compare how the respondents in the control soums (no loans offered) and the treatment soums have developed over time. Since the women in the treatment and the control villages were on average very similar before the experiment, differences in their subsequent development and outcomes will only be related to whether or not they received a loan.

This comparative analysis is of particular interest given the global financial crisis, the impact of which was felt in rural Mongolia in the period between the two survey rounds. Cashmere prices dropped by more than one third over a short period of time, adversely affecting many herder families. We hope that our results shed light on the question of whether the availability of microfinance has alleviated – or maybe even increased – rural households’ financial vulnerability during the crisis. Complete results are expected to be available in June of this year and will be summarised on this blog. For now, the data collected during the baseline survey already provide some insights into the state of rural microfinance in Mongolia. Three observations stand out:
First, we find that only 44 per cent of respondents had no outstanding debt at the time of the first interview while 46 per cent already had a loan. Almost ten per cent of respondents even had two or thee loans. Contrary to popular perception, penetration of microcredit was already quite advanced across rural Mongolia, even among our sample of relatively poor women who were selected because of their supposedly limited access to finance.

Second, virtually all respondents with a loan took out that loan in 2007 or 2008. Almost half of the respondents had not had another loan (whether repaid or not) in the last five years. Competition for rural customers – in particular between Khan Bank, XacBank and Mongol Postbank – had intensified in recent years. In our sample, Khan Bank has by far the largest market share: we find that just over one half of those with an outstanding debt owe it to Khan Bank.

Third, we find that between 70 and 80 per cent of the debt outstanding at the start of the experiment was used for consumption purposes and not for financing micro-entrepreneurial activities. This is an important finding since it shows that even though microfinance in rural Mongolia has advanced rapidly in recent years, the vast majority of loans has been used for consumption rather than income-generation. That is not to say that these loans have not been ‘useful’: the ability to smooth consumption is particularly important at low income levels. But it will be interesting to see whether the crisis has impacted households with varying debt levels differently. Moreover, it will be of interest to find out to what extent the individual and group loans disbursed during our experiment, which were intended to finance businesses, have indeed been used for such income generating purposes.

Stay tuned for more…

Share on Facebook