n 2009, the results rom two microcredit impactstudies in Hyderabad, India, and Manila, thePhilippines were released to mixed responses(Banerjee, Duo, Glennerster, and Kinnan 2010;Karlan and Zinman 2011). Some media declared mi-cronance a ailure (Bennett 2009). Many in themicronance community dismissed these random-ized studies as too limited to be a true reection o the entire sector.
These rst randomized studies caused a sensa-tion because they challenged the dominant impactnarrative or microcredit—a narrative that rests onloans to capital-constrained microentrepreneurswho earn a steep return on marginal capital andthus can repay a relatively high interest rate and re-invest to grow out o poverty—and the way in whichthat narrative had been universalized in the popu-lar imagination. In act, the results were more nu-anced. What the microcredit studies really showedis that this model o microcredit works or somepopulations—those who successully grow busi-nesses—but not or others.Many now agree that the expectations or micro-credit in the popular discourse were overblown.For some, the pendulum had swung: ar rom a pan-acea against poverty, some argued that microcreditwas actually doing harm. The evidence supportsneither extreme view. In act, the results o thestudies aligned with and conrmed some o the evi-dence rom nonrandomized methods already in themicronance research literature that ound modestbut neither revolutionary nor deleterious impactsrom credit. While the concept o capital that willallow poor people to unleash small business oppor-tunities remains valid or some poor clients, not ev-ery borrower is a microentrepreneur—take-up ratesor credit products are oten surprisingly low, andnot all economic activities that poor people engagein yield high returns. Microcredit is not transorm-ing inormal markets and generating signicantlyhigher incomes on average or enterprises. And yetthe industry has ocused almost exclusively on therhetoric o entrepreneurship and has overlookedthe many important benets to households that areusing loans to accelerate consumption, absorbshocks, or make household investments, such as in-vestments in durable goods, home improvements,or education or their children.Combined with other evidence, randomizedevaluations are contributing to an emerging body o knowledge that is creating a new narrative aroundhow nancial services or the poor really work. Asthe results rom new studies have been released,the discussion has evolved, and randomized evalua-tions are being used to examine when particularproducts and designs work, or what segments o people, and why.Today researchers are using randomized tech-niques to better understand the underlying nan-cial services needs o poor clients and what impactsare achieved when appropriate nancial servicesare oered. Building on evidence rom earlier non-randomized studies, researchers are increasinglyable to work with micronance providers to applythese techniques to product innovation and totweak product design. In this way, randomizedtechniques can make a signicant contribution tothe eld by clariying our understanding o precise-ly how, and under what conditions, nancial ser-vices benet poor people.
(See Box 1.)Poor households clearly have other nancialneeds that go beyond working capital loans to mi-croentrepreneurs. They use a variety o inormaland semi-ormal mechanisms to cope with risk,
Latest Fidis rm RadmizedEaluatis Micrface
1. See, e.g., Helms (2010).2. Naturally, not all settings are appropriate or randomized eval-uations. This paper does not discuss such methodologicalissues in detail, but it does share ndings rom settings whererandomized evaluations were easible and illuminating.