What determines enterprise borrowing from Self Help Groups? A Predictive Assessment using Supervised Machine Learning Algorithms

Despite several advantages of borrowing from Self Help Groups (SHGs), why do many enterprises in India, still rely on informal lenders (MoSPI, 2020)? To answer this question, we develop a novel enterprise-village matched dataset and use a variety of Machine Learning methods to predict the choice of an enterprise between SHG and informal lenders as the major source of finance. Among them, XGBoost with an AUC of 94% has the highest predictive power. We conduct several interpretable machine learning techniques to understand village-specific determinants of enterprise borrowing from SHG. Access to urban centers including district headquarter, and socio-demographic factors such as high literacy rates and improved sex ratios in a village, play important roles in credit uptake from SHGs. Absence of financial access points, such as commercial or cooperative bank branches, does not appear to be prohibitive. We also apply XGBoost model to estimate potential demand for SHG loans among self-financed firms. Potential for SHG inclusion among these firms, remains low and skewed toward southern districts.