Comments
Summary
This was a short, shallow review conducted by SoGive. SoGive is an organisation which provides services to donors to help them to achieve high impact donations.
In this piece we set out the following:
- The existing EA view on microfinance
- Some updates to this based on more recent evidence
- Further considerations relating to business coaching/mentoring/support
- How should this work be funded?
A summary of the conclusions is as follows:
- The existing EA view on microfinance was lukewarm/negative, but this was several years ago.
- More up to date evidence does not seem to materially change this (looking purely at microcredit). Note that the findings (then and now) are mixed.
- We considered the provision of business support (independent of microcredit), and found the evidence to be mixed. So if an organisation suggests that their impact would outperform that of the RCT findings because of the business support they provide alongside the loan, the burden of proof is still on them to demonstrate this.
- We suspect that microfinance should probably be mostly funded by impact investment as opposed to donations, assuming it should be funded at all. This appears to be a departure from earlier EA thinking on the topic, which did not, as far as we are aware, discuss this question.
We currently know of no good, high-impact donation opportunities in the microfinance space that we would recommend, and this review suggests that the burden of proof on such a charity is moderately high.
Long list of assumptions
This was a fairly shallow review. In order to complete it within our time constraints, we made the following scope restrictions/assumptions:
- We considered only microcredit, noting that (a) microfinance tends to focus on the more profitable microcredit first and then branch out to savings and (maybe) insurance and (b) many seem to believe that microsavings and microinsurance are more impactful.
- We made no attempt to analyse the heterogeneities in the beneficiary response to microcredit, which appear to be material; in particular, while much of the evidence pointed to minimal impact of microcredit on income generation, some of the evidence found that it did achieve results. We have not looked into why this is. Similarly for business skills support. It is possible that a charity has identified a way to consistently replicate the positive results; our review has been too shallow to identify such an organisation (this is linked to the next bullet point).
- We have restricted this analysis to the broad area of microcredit; we have not given material consideration to specific microfinance organisations. A useful next step may be to look at specific RCTs where the impact of microfinance appears to have been most promising, and identify the organisation(s) achieving those results, as these may be reasonable candidates for high-impact charity status. We would then want to consider whether the work is likely to be consistently as successful as the work highlighted by the RCT, as well as cost-effectiveness analysis and other analyses.
- Our search for high-quality evidence was limited to J-PAL and AidGrade.
- Our opinion that microfinance should probably be mostly funded by impact investment (if it should be funded at all) is weakly held, and would be improved by more discussions with funders and microfinance organisations, and some other analyses set out in the relevant section below.
- We have performed no analysis of cost-effectiveness. A priori, it seems likely that microcredit which actually works would probably outperform GiveDirectly (i.e. just giving money to the poor) under reasonable assumptions about repayment rates and administration costs; this has not been considered in this analysis.
- In exploring the “existing EA view on microfinance” we did not attempt to carefully explore everything that anyone within the EA sphere has said on the topic. Nor did we attempt to define what is meant by being within or adjacent to the EA sphere. For example the books Poor Economics by Banerjee and Duflo or Portfolios of the Poor by Collins, Morduch et al also refer to microfinance and are (arguably) also within or at least adjacent to the EA sphere, but we did not refer to those books here.
1 The existing EA view on microfinance
Prior to writing this piece, we were aware of the following opinions on microfinance being expressed within or adjacent to the EA sphere:
GiveWell, whose opinion we take seriously, has said in the past:
“We feel that global health is a better area for a donor overall” (source: blog post from 2009)
The book Due Diligence by David Roodman (published in 2011) argues the following:
- Credible evidence that microfinance helps people escape poverty is scarce
- Evidence that microfinance enhances beneficiaries’ sense of freedom is mixed
- The claim that microfinance successfully leads to the building of industries is the strongest, although this does also carry the risk of credit bubbles
Overall, the EA opinion on microfinance appears to have been lukewarm at best.
2 Some updates to this based on more recent evidence
This section sets out our own review of 6 RCTs, followed by another review by Banerjee et al of 6 RCTs (2 of which overlap with our review). These reviews both find mixed, frequently negative results around microfinance (where “negative” is meant in the sense of “no impact” rather than “negative impact”). This seems to suggest no material change in opinion compared to the earlier EA opinion.
2.1 Our review of microcredit studies (no business skills training)
As part of this short review, we found RCTs (randomised controlled trials) on microcredit from the J-PAL website, and reviewed all those which seemed relevant. We also searched on AidGrade and found no further studies beyond those which we had already found.
All 6 of these were “pure” microcredit -- i.e. no mention was made in the papers of accompanying the loan with business training or support.
This resulted in a list of 6 studies, some of which were already considered by Roodman in his book or by GiveWell when they wrote about this several years ago, but some of which had not been published at this stage.
The review found that most high-quality studies of microcredit did not find positive results on outcomes of value, such as women’s empowerment, children’s education, income, or consumption. Possible exceptions:
- One study in Morocco found positive economic outcomes for borrowers, although this was not an unmitigated positive -- the findings were mixed and some were worse off
- One study in South Africa did find a positive impact on economic self-sufficiency (employment and income), and mixed but broadly probably positive impacts on mental health. (note: this study has also been commented by GiveWell).
- One study in the Philippines found that microcredit led clients to shrink their business. While at first glance this may seem like a negative outcome, the paper’s authors argued that it may be positive -- i.e. the smaller business may have been more profitable.
The other papers essentially found zero impact on the outcomes that we might care about.
If the links are too small to read, you can find them in Appendix 1.
* Note that the summary of the 2006-09 Morocco may sound like an unmitigated positive, the actual results are somewhat mixed. Here’s a fairly lengthy quote:
“Across all households, microcredit offers generally increased sales, household consumption, and profits. However, effects on profits were mixed across business types. While more profitable businesses in microcredit villages increased their profits, profits fell for businesses in microcredit villages that already made relatively small profits. While overall self-employment income increased in villages offered microcredit, this increase was accompanied by a decline in income from day labor and salaried positions. This labor tradeoff resulted in no overall change in income levels between treatment and comparison villages.”
** Note: For the 2007-09 Philippines study, we believe the J-PAL interpretation of “improved risk management and sharing” needs further justification. On closer inspection, the paper indicates that clients substituted away from formal insurance towards informal methods such as informal credit from family and friends. The cost of capital as measured in purely financial cost from this form of finance is almost certainly lower than the cost of capital from borrowing from a MFI, however the social cost of capital may be material. Contrary to idealised notions of informal financial risk-sharing being a beautiful expression of financial solidarity, qualitative research also exists to suggest that people may feel a loss of dignity or sense of shame from having to ask their friends for financial help in a way that may not apply for loans from a formal institution.
2.2 Banerjee et al 2015 review
This 2015 review by Banerjee, Karlan and Zinman of 6 microfinance RCTs was very useful, and here’s an extensive quote which we found insightful:
<<What are the key takeaways, in our estimation?
- One is the existence of modest take-up rates of credit among (prospective) microentrepreneurs, which is a prima facie case against microcredit being a panacea (a cure-all) in the literal sense, and presents a statistical power challenge for randomized identification strategies.
- Second is the difficulty of meeting the power challenge by predicting microcredit take-up (as noted above, this speaks to the likely importance of unobserved heterogeneity in borrowing and lending decisions, and hence to the value of large-sample randomization).
- Third is the lack of evidence of transformative effects on the average borrower.
- Fourth is that the lack of transformative effects does not seem to be for lack of trying in the sense of investment in business growth.
- Fifth is that the lack of transformative effects should not obscure other more modest but potentially important effects. If microcredit’s promise was increasing freedom of choice it would be closer to delivering on it.
- Sixth, just as there is little support for microcredit’s strongest claims, there is little support for microcredit’s harshest critics, at least with respect to the average borrower.
- Seventh, the limited analysis of heterogeneous treatment effects in these studies does suggest hints of segmented transformative effects—good for some, bad for others.>>
3 What happens if you include business skills training?
In our experience we have found that some charities have introduced some form of training or support to ensure that the loan is not provided in isolation. A priori, it seems reasonable to believe that this may improve the probability of success, although this should still be treated with some caution given that many interventions achieve relatively little impact.
To illustrate our claim that many microfinance charities include some business training, here are some examples from the SoGive database of microfinance charities:
(If the links are too small to read, just go to https://app.sogive.org and do a search for the charity)
To investigate the extent to which this may improve the success of the intervention, we sought J-PAL papers relating to business skills training.
The studies find that the effectiveness of training is variable. It seems likely, therefore, that any study here which finds positive results in one context is likely to have low generalisability, and so further careful assessments are likely needed when this sort of training is applied in new contexts.
If the links are too small to read, you can find them again in Appendix 2.
Interlude: Tentative conclusions about the effectiveness of microcredit
As the evidence on both microcredit and business skills training appears to be mixed, it seems generous -- maybe foolhardy -- to support a micro credit project without robust evidence to support the claim that the project is effective.
This applies even if the charity counters that the business support they provide will make up for the fact that microcredit, on its own, is insufficient. Business support is also often ineffective, so we need evidence that the particular package being proposed by the charity still works.
It seems that relatively little of the microcredit work conducted by charities is supported by robust evidence of effectiveness, such as a RCT.
4 How should microfinance be funded?
An aspect not considered in any of the GiveWell posts that we’ve reviewed on this topic is how this work should be funded (assuming that it should be funded at all).
Here we outline three reasons for believing that microfinance should perhaps be funded by investors and not donors:
- There may be enough investor appetite to fund microfinance
- Investment may improve targeting of customers
- Investment may avoid bubbles
4.1 There may be enough investor appetite to fund microfinance
Broadly speaking, we at SoGive believe that where it’s possible for an activity to be funded through investment rather than donations, it is preferable for it be funded through investment.
As far as we are aware, the argument for philanthropically funding microfinance is that microfinance for those who are not the poorest of the poor may be fundable by impact investment, however microfinance for the poorest would be too unprofitable. It seems there are some reasons to doubt this:
- Some microcredit charities report very good default rates (example: 1% default + 4% rescheduled http://www.justsmallchange.org.uk/images/Reports/Impact_report_2018-19.pdf; note the explanation for this might be that the default rates stated are simply wildly misleading, as explained in this GiveWell blog post)
- Even if the default rates are higher than they appear, it may still be possible to fund the work through investment. The most compelling finding from Roodman’s book Due Diligence is that microfinance succeeds at industry building, and this argument is much stronger than the argument that microfinance successfully lifts people out of poverty. There is precedent for equity investors opting to invest in risky investments with the possibility that the investment may grow into a large profitable company. As there is precedent for this successfully happening in the past, it may be possible for a company to make the case for equity investment.
This seems to lean us tentatively in the direction that this work should be funded through investment, not donations. However this is a weakly held opinion.
This opinion would be somewhat more robustly held if we knew of at least one microfinance charity which attempted to be funded and provided a meaningful demonstration that this was not possible. However, even if we encountered this there would be an obvious bias on the part of the charity (receiving donor funding is preferable to receiving investment).
To overcome this, and to make this opinion more robustly held still, we would want to hear from (a) multiple microfinance charities who have tried and failed to get funding from investors (b) multiple microfinance organisations that are successfully getting funding from investors (c) multiple investors.
4.2 Investment may improve targeting of customers
Where a microlender (or, indeed, a lender) lends to someone who will be unable to repay, this is contrary to the interests of both lender and borrower. At first glance, it seems likely that the discipline enforced by the requirement to repay the funds will make an investment-funded microlender better at judging this than a philanthropically-funded microlender.
This is a tentatively held opinion which would be more firmly held if we were to conduct a review of philanthropically funded and investment-funded microlenders and compare them for their default rates. This analysis would need great care as there would be a number of confounding factors to adjust for, and it’s obvious at this stage whether it’s even possible to adequately adjust for these confounding factors.
4.3 Investment may avoid bubbles
This is an even more tentatively held opinion, as the world of investment is replete with bubbles. However it seems likely that a bubble would be even more likely to ensue if the funders have no incentive to observe whether the market is already overfunded.
While we could justify this with a general overview of philanthropic propensity to fund something that is already overfunded, we believe that the most useful further analysis here is to consider whether the microcredit market currently *is* overfunded or not.
5. Conclusion
The conclusion of this is to affirm the earlier hesitancy around microfinance. If anything, the question mark about whether donating is the right way to fund the work may add to the hesitancy.
In order to believe that a donation to a microcredit organisation is a high-impact donation opportunity, we would have to have the following:
- Evidence that the particular microcredit charity being considered for funding achieves worthwhile outcomes (because we don’t think that generalising from other
- Evidence that the work is cost-effective (which hasn’t been considered as part of this piece)
- Evidence that donating to this work isn’t displacing possible equity finance, or creating precedents that will make other similar work harder to fund through investment
-------------------------------------------------------------
APPENDIX 1: The microcredit studies (no business skills training)
The 6 studies included in the table above are mentioned again here, together with slightly more information, and links in a not-tiny font size.
2005-2010: https://www.povertyactionlab.org/evaluation/measuring-impact-microfinance-hyderabad-india
Researchers evaluated the impact of increased access to microcredit on the economic and social well-being of women and their families in Hyderabad, India. They found that microcredit had mixed effects on business activities and little to no effect on women’s empowerment or children’s education.
2006-2009:
https://www.povertyactionlab.org/evaluation/impact-microcredit-rural-morocco
Researchers estimated the impact of a microcredit program, which was randomly rolled out in rural areas of Morocco. Thirteen percent of the households in treatment villages took a loan, and none in comparison villages did. Among households identified as more likely to borrow, microcredit access led to a significant rise in investment in assets used for self-employment activities, and an increase in profit, but also to a reduction in income from casual labor. Overall, there was no gain in income or consumption.
2004-2006: https://www.povertyactionlab.org/evaluation/small-individual-loans-and-mental-health-south-africa
An important means of exiting poverty is access to productive resources; yet many poor people lack the capital necessary to invest in higher education, smooth consumption, or start a business. This evaluation examined the direct impact of providing small consumer loans to marginally creditworthy individuals in South Africa on their credit access, investment, and well-being, as well as the profitability of these loans for lenders. Results found approving loans for marginally creditworthy applicants did improve economic outcomes for poor households and generated profits for lenders.
2007-2009:
Microcredit is a widely-used tool that has been proposed as a way to help fight poverty and promote economic growth. Researchers measured the impact of individual-liability microcredit on marginally creditworthy applicants in the Philippines. They found that increased access to microcredit expanded borrowing and improved risk management and sharing, but it also led clients to shrink their businesses.
Note: We believe the J-PAL interpretation of “improved risk management and sharing” needs further justification. On closer inspection, the paper indicates that clients substituted away from formal insurance towards informal methods towards informal credit from family and friends. While the cost of capital as measured in purely financial cost of capital from this form of finance is almost certainly lower than the cost of capital from borrowing from a MFI, the social cost of capital may be material. Contrary to idealised notions of informal financial risk-sharing being a beautiful expression of financial solidarity, qualitative research also exists to suggest that people may feel a loss of dignity or sense of shame from having to ask their friends for financial help in a way that may not apply for loans from a formal institution.
2009-2012:
https://www.povertyactionlab.org/evaluation/impact-microcredit-women-mexico
Microcredit is widely believed to create opportunities for entrepreneurs and families to improve their economic and social well-being. Researchers expanded microcredit offerings in Sonora, Mexico to evaluate the effect of improved credit access on economic and social outcome. They found that microcredit increased access to formal financial services and enabled some businesses to expand, but did not increase household income or prompt new business creation.
2008-2011:
https://www.povertyactionlab.org/evaluation/impact-savings-groups-lives-rural-poor-ghana
In contexts with limited access to formal financial institutions, Village Savings and Loan Associations (VSLAs) could fill a gap by facilitating access to informal financial services. Researchers partnered with CARE to evaluate the impact of VSLAs on rural households. The promotion of these groups led to an improvement in financial inclusion for participants—including, substantial increases in savings and receiving a loan. However, researchers did not find evidence of impacts on average consumption, business outcomes, women’s empowerment, or other welfare outcomes.
APPENDIX 2: The business skills training studies
2002-05
https://www.povertyactionlab.org/evaluation/business-education-microcredit-clients-peru
Microfinance has generated worldwide enthusiasm as a potential answer to economic development and poverty reduction. But high default risk and unproductive use of loaned funds plagues many programs. Researchers worked in Peru to measure the marginal impact of adding business training to a group lending program. The results of this study found business training slightly improved business practices, but had no impact on key business outcomes such as revenue and profit.
2007-15
https://www.povertyactionlab.org/evaluation/boosting-firms-productivity-mexico-consulting-services
Compared to their counterparts in high-income countries, small and medium enterprises (SMEs) in low-and-middle-income countries are often less productive, grow slower, and hire fewer workers. In Mexico, researchers are testing whether this lagging productivity could be due to lower managerial capacity. They found that providing subsidized managerial consulting to Mexican SMEs boosted their productivity and hiring.
2008-11
Microenterprises make up a large portion of employment in the developing world but little is known about constraints on their growth. Researchers partnered with the consulting firm Ernst & Young to test whether providing tailors in Accra with individualized consulting, a sizable cash grant, or both can facilitate growth. Initially, tailors experimented with investments and management practices, but most learned overtime that these new business strategies were not profitable for their ventures. Some tailors, however, did experience positive outcomes—evidence that entrepreneurial talent varies among individuals and an individual’s talent may not be readily known, even to the entrepreneur herself.
2003-08
Governments and donors spend billions of dollars subsidizing entrepreneurship training programs around the world. Some common rationales for these programs include that they can improve access to training services and promote employment among groups affected by discrimination, but it has been difficult to test rigorously whether such benefits exist. Using a randomized evaluation, researchers found little evidence supporting these common rationales for subsidizing entrepreneurship training in the United States.
2013-14
While microfirms can play an important role in the labor markets of developing countries, they tend to have low productivity and difficulty growing. In partnership with Simón de Cirene, a Chilean non-profit organization, researchers evaluated the impact of providing role models and personalized assistance through various delivery methods on the business outcomes of micro-entrepreneurs. One year after the program ended, household income increased for individuals assigned to receive a role model or personalized assistance. The two interventions benefited different micro-entrepreneurs: the role model visit benefited less experienced entrepreneurs, while the personalized assistance benefited more experienced entrepreneurs. Additionally, role models were more cost-effective than technical assistance in improving business participation and income.
