The use of control groups in impact assessments for microfinance

Table of Contents Foreword 1 1. Why use control groups in impact assessments of microfinance projects? 3 2. Biases of the control-group method and how to avoid them 7 2.1 Sample selection bias 7 2.2 Misspecification of underlying causal relationships 9 2.3 Motivational problems 10 3. Conclusions and recommendations 15 General recommendations 18 Recommendations for evaluators and donors 18 Recommendations for programme managers 18 The Microcredit Summit Campaign aims at reaching 100 millions of the world's poorest with credit by the year 2005. As microcredit is increasingly claimed to be a powerful tool for alleviating poverty, it is naturally interesting to find out whether it actually has this effect. While there is substantial first-hand evidence of the empowering effect of microfinance on the poor, the effects, ways and causes are less known — a call for more thorough impact assessments in microfinance. For the ILO in particular, impact assessments are essentially intertwined with its objective of social justice, because they help understand how some people can be better off than others, as a result of the same intervention. This paper by Paul Mosley was prepared as a contribution of the ILO to the CGAP working group on Microfinance Impact Assessment Methodologies. CGAP members decided in April 1996 to establish a working group on this issue, recognizing that assessments of the impact of microfinance services are complex and entail several methodological challenges. A virtual meeting was held April 7-19th 1997, reviewing several background papers 1 and a synthesis report by David Hulme 2. Paul Mosley is professor at the University of Reading, and the head of its International Development Centre. He recently published with David Hulme Finance Against Poverty (London: Routledge, 1996) that sparked off a lively discussion among microfinance specialists on the trade-off between financial sustainability of microfinance institutions (MFIs) and outreach to the poor. As Paul Mosley states in his introduction, the evaluation of any project or activity, including microfinance, aims ideally at knowing the effects and who is affected, but also the causes. Social sciences have developed a number of methodological approaches that attempt to simulate the situation which would have prevailed if there had been no project. One of them is the control group method, which compares a population that had benefitted from a microcredit scheme to another group which had not. While being increasingly used in microfinance impact assessments, not least due to its limited data …