High quality evidence on the state of financial access around the world is advancing rapidly. A happy consequence of increasing knowledge is the ability to better recognize what we don’t yet know. Today, FAI is launching a series on the ten questions, some micro, some macro, that need answers if we are to make informed decisions on how to improve financial access. These questions will be available as a framing note at the end of the series on the FAI site and later as part of a collection of studies to be published in a forthcoming book.
Question 1: Does financial access--evaluated in typical settings with a long enough time horizon to see change--substantially improve the well-being of customers?
The most fundamental, unresolved question concerns impact. Does expanding financial access really make a notable difference to families and communities? And, if so, how and when?
Muhammad Yunus (1999) and other early microcredit advocates took us down a narrow path. Yunus’s stress on “microcredit for micro-enterprise” continues to play well with the public, but we’re learning that the rhetoric does an injustice to the complicated reality of how low-income families actually use financial services (e.g., Collins et al 2009). Not only that, but the early microcredit rhetoric does an injustice to what innovators, including Grameen Bank itself, are doing on the ground. In practice, they often provide savings services, insurance, and some training, not just business loans. And even when it comes to credit itself, the notion that loans are sought exclusively for business investment fails in the data, whether when asked directly (Collins et al 2009) or when derived indirectly (Karlan and Zinman 2011).
That takes us to the question of what borrowers are in fact doing with their loans when they’re not funding business. One important use is to pay for big, lumpy expenses, including healthcare costs, school fees, and home repair. A second is to pay down more expensive loans. A third is to help smooth seasonal ups and downs of consumption. The uses make perfect sense from the standpoint of economic theory, but they make microcredit advocates nervous.
Advocates worry that borrowers cannot repay loans if there’s no major business investment in the picture, but they miss the larger picture. The evidence in Collins et al (2009) suggests that borrowers repay microcredit loans with money earned in various wage and self-employment activities. Even if a particular loan is not used to fund self-employment, income from self-employment may nevertheless provide an important way to repay the loan. The distinction too often gets lost. From the bank’s perspective, the central question should be whether the household can generate the cash flow to service the loan (and whether it will be in a position to apply that cash flow to loan repayments), not whether the loan is used for a particular purpose.
Given the complexity of what access to finance means for households, decent evaluations with clear bottom lines are needed to anchor conversations. But so far, evaluations have done little to settle the question of whether microfinance “works” or not. We have an older set of studies that hoped to measure the impact of microfinance on consumption or income but which were compromised in one way or another, usually by not making comparisons to credible control groups (Armendáriz and Morduch 2010 provide an overview). Randomized trials do far better in terms of credibility, but researchers are often forced to grab opportunities where they arise and thus tend to investigate narrow populations and short-term outcomes. One thing we have learned from randomized trials is that internal validity is crucial. We’ve also learned that internal validity is no substitute for external validity (Cartwright 2007), and it is questions around how – and if -- results extend to other contexts that leave the literature unsettled.
While the new evaluation literature focuses rightly on study design, a complementary concern rests with data accuracy. The work of Robert Cull and Kinnan Scott (forthcoming paper on “Question Format and the Identity of the Respondent: Experimental Evidence Measuring Use of Financial Services”) point to important practical steps for collecting sharper data on financial variables—especially with regard to informal mechanisms and the broad range of “semi-formal options.
Together with others, including Samphantharak and Townsend (2010) and de Mel et al (2009), the studies give us a better grounding for mapping theory and data. Improved data on cash flows, microenterprise finance, and the use of financial services will be one part of answering the impact question.
The series has been compiled as a framing note on the FAI site now and will later appear as part of a collection of studies to be published in a forthcoming book.