Duvendack, M., Palmer-Jones, R., Copestake, J.G., Hooper, L., Loke, Y. and Rao, N. (2011) What is the evidence of the impact of microfinance on the well-being of poor people? London: EPPI-Centre, Social Science Research Unit, Institute of Education, University of London.Link to Source
The authors include 58 papers, some of which are based on the same datasets. The included studies provide data from 19 countries in Sub-Saharan Africa, the Middle East and North Africa, South Asia, East Asia and the Pacific, and Latin America and the Caribbean. They utilise a variety of study designs, such as randomised control trials, pipeline designs, with/without comparisons (in panel or cross-section form), natural experiments and general-purpose surveys.
The authors examine the included studies by intervention and outcomes, as well as by research design and analytical methods, and find the following:
- Two randomised controlled trials (RCTs) conducted in India and South Africa yield no convincing evidence on the positive impact of microfinance interventions on well-being. Microfinance interventions have an impact on business activities, but they affect well-being only indirectly. These RCTs find very few impacts on health, education and subjective well-being, or indirect indicators of well-being such as income or consumption expenditures. In conclusion, the evidence from the RCTs on the impact of microfinance interventions on the well-being of the poor is limited and inconclusive.
- Nine studies report findings from pipeline designs without random allocation of units of study, showing no significant impact of microfinance on the vast majority of welfare outcomes. Overall, the pipeline studies test a large number of outcome variables but provide only limited evidence of strong positive impacts of microfinance. In addition, the greatest impacts are reported by studies with the weakest designs. Thus, the authors conclude for pipeline designs that microfinance has little or no statistically significant effect on well-being outcomes.
- Sixteen studies report findings from with/without, before/after and panel designs.These studies include the replications of the Pitt and Kandker (1998) study, several of which conclude that Pitt and Kandker overstate the impacts of microcredit on the well-being of poor people. Several studies which are based on their data produce mixed results, ranging from significantly positive to significantly negative impacts.
- The authors also revisit longitudinal data provided by USAID and particularly focus on a study by USAID conducted in India. Although follow-up publications broadly confirm the findings of this study, the authors identify problems with controlling for unobservables which reduce the credibility of its positive estimated impacts for microfinance.
- In a similar vein, the remaining with/without studies lack a common thread and provide mixed results, which range from significantly positive to significantly negative impacts of microfinance. In addition, the authors raise doubts about the reliability of outcomes presented by most with/without studies.
The authors conclude that impact evaluations of microfinance tend to have low internal validity, not allowing for strong conclusions. While acknowledging that they treated the included studies very critically, they argue that there is no good evidence for the beneficent impact of microfinance on the well-being of poor people.
Microfinance institutions worldwide are growing and have become an important tool in the fight against poverty, often targeting women. Existing evidence on the impact of microfinance is contradictory, and at the time of writing only one systematic review on the impact of microfinance interventions on the poor in Sub-Saharan Africa had been published, along with a review protocol on the impact of microcredit on the empowerment of women.
The objective of this systematic review is to assess the evidence on the impact of microfinance on the social and economic well-being of poor people in developing countries who are excluded or marginalised within their own society.
The authors included randomised control trials, pipeline designs, with/without comparisons (in panel or cross-section form), natural experiments and general-purpose surveys, assessing the effect of microfinance on the well-being of poor people. The minimum sample size (treatment and control group combined) had to be larger than 100.
The included population was defined as individuals from low and middle income countries classifiable as poor, excluded, or marginalised within their society. Microfinance interventions were included, including microcredit and/or “credit plus” programmes, which may include savings components. Studies purely examining micro-savings were excluded.
Primary outcomes labelled as “well-being” were income, health and education. Secondary outcomes included microenterprise profits and/or revenues, expenditure, labour supply, employment, assets, housing improvements, education, health and health behaviour, nutrition and women’s empowerment.
The authors conducted a search for formally published studies, studies available in abstract, web-based studies, PhD theses and organisational reports from 1970 onwards. Searched databases included EconLit, Embase and Informaworld, as well as systematic-review databases, relevant NGO websites, and search for PhD thesis abstracts, via Google Scholar. The authors also searched the bibliographies of relevant books, journal articles, PhD theses and grey literature.
The authors assessed study quality and synthesised the evidence in a narrative format, differentiating the evidence according to different study designs.
The review has clear inclusion criteria and the authors searched a wide range of databases, search engines and websites. The authors include a very comprehensive quality assessment of the included studies using Cochrane and EPPI guidelines and report the results of this assessment. However, the systematic review has some limitations. The authors only searched the English language literature, and while this was due to budget constraints, it is a limitation of the search. Data extraction was only carried out by the first author, although this was checked by other members of the team. The authors tap into the problem of unit of analysis errors, but do not fully address it with reference to their synthesis.