Using Repayment Data to Test Across Models of Joint Liability Lending

Abstract

Spurred by its successful delivery of credit to poor borrowers in diverse areas of the developing world, joint liability lending has caught the imagination of development theorists and practitioners. Various theories have arisen to explain why joint liability group-based lending can be an improvement over traditional individual-based lending. Here we exploit the idea that if a model were true, the repayment rate would vary in a systematic way with various covariates. We thus use observed repayment rates to test across four representative and oft-cited models of joint liability lending. The theoretical part of this paper develops the models' implications for repayment and derives new ones, signing the derivative of a project choice, monitoring, default, or selection equation. For example, we find that several models imply that higher correlation of output can raise the observed repayment rate, and in some the ability to act cooperatively leads to lower repayment rates. More generally, the models agree on some dimensions and conflict on others. The empirical part uses survey data from 262 Thai joint liability groups of the Bank for Agriculture and Agricultural Cooperatives (BAAC) and from 2880 households of the same villages. Nonparametric, univariate tests and multivariate logits are used to evaluate the predictions of the models. A hybrid, partially linear procedure is used to understand any differences in the results between the two types of tests. We find that the Besley and Coate model of limited enforcement is strongly supported in the more rural, poorer region of Thailand covered by the data. In the more prosperous region, closer to Bangkok, support is found for the Stiglitz model of moral hazard and the Ghatak model of adverse selection.