Bogus Joint Liability Groups in Microfinance -- Theory and Evidence from China

Survey data from CFPAM, the leading joint-liability microfinance lender in China, indicate that nearly 70% of all borrower groups in the sample are `bogus' -- that is, one person uses all loans given to group members in a single investment project while all other group members act as unproductive cosigners. This practice not only violates CFPAM rules but is also inconsistent with the majority of the theoretical literature on group lending, a basic tenet of which is that each borrower uses their own loan to implement their own investment project (what we call `standard' group). We therefore extend the classic model of group lending under joint liability by explicitly allowing for both standard and bogus groups in a setting with the possibility of strategic default due to limited enforcement. The optimal choice between standard and bogus groups is endogenous and depends on the borrowers' characteristics (project productivity and probability of success). We analyze the optimal group loan contract (or menu of contracts) and show that bogus groups optimally arise when either the productivity differential between the projects in a group is high (in heterogeneous groups), or when the absolute level of project productivity is high (in homogeneous groups). Explicitly allowing for the possibility of bogus groups not only helps the lender avoid losses which would occur if their presence is ignored, but also enhances productive efficiency and borrower welfare in the economy. We test the model predictions with data from rural China and evaluate the welfare gains from implementing the optimal contract (or menu) relative to the benchmarks of: (a) lenders operating unaware of bogus groups or (b) lenders using a contract that endogenously rules out bogus group formation.