Abstract:
This paper uses a panel data set of microfinance institutions (MFI) across the world to compare several identification strategies of cost efficiency and economies of scale. Concretely, we contrast the non-parametric Data Envelopment Analysis (DEA) with the Stochastic Frontier Analysis (SFA) and a distribution-free identification based on time-invariant heterogeneity estimates. Furthermore, we analyze differences of production functions across regions and investigate the relevance of accounting for unobserved heterogeneity across countries. The results suggest that efficiency rankings of MFIs are robust across identification strategies, but highlight the relevance of accounting for unobserved heterogeneity. We further find substantial economies of scale for a pure financial production process. However, accounting for the multi-dimensional production process of MFIs by including a measure of outreach lowers the estimated extent of economies of scale for the parametric estimations, suggesting that producing outreach creates high transaction costs and requires exploitation of local knowledge.
