The Effect of “Plus” Services in Microfinance: A Doubly Robust Machine Learning Approach

Abstract Microfinance institutions (MFIs) expand financial inclusion by providing credit and savings services to low-income households excluded from formal finance. Because the poor face multiple needs, many MFIs offer “plus” services—either financial (e.g., insurance, remittances) or nonfinancial (e.g., education, business training, health promotion, gender empowerment). We use a doubly robust, random-forest–based approach to obtain semiparametrically efficient estimates of the average treatment effect (ATE), estimating the ATE of each plus service on both outreach (social mission) and financial performance, while accounting for heterogeneity in MFI characteristics and operating environments. The results show that nonfinancial plus services enable MFIs to both deepen and broaden outreach. By contrast, MFIs that add only financial plus products serve fewer and less-poor clients, consistent with mission drift. The policy implications are that, to advance financial inclusion, stakeholders should prioritize nonfinancial ‘plus’ services and be cautious about promoting only bank-like financial products.