Big Loans to Small Businesses : Predicting Winners and Losers in an Entrepreneurial Lending Experiment

We experimentally study the impact of substantially larger enterprise loans, in collaboration with an Egyptian lender. Larger loans generate small average impacts, but machine learning using psychometric data reveals dramatic heterogeneity. Top-performers (i.e., those with the highest predicted treatment effects) substantially increase profits, whereas profits for poor-performers drop. The magnitude of this difference implies that an individual lender’s credit allocation choices matter for aggregate income. Evidence on two fronts suggests large loans would be misallocated: top-performers are predicted by loan officers to have higher de- fault rates; and, top-performers grow less than others when given small loans, implying that allocating larger loans based on prior performance is not efficient. Our results have important implications for credit expansion policy and our understanding of entrepreneurial talent: on the former, the use of psychometric data to identify top-performers suggests a pathway towards better allocation that revolves around entrepreneurial type more than firm type; on the latter, the reversal of fortune for poor-performers, who do well with small loans but not large, indicates a type of entrepreneur that we call a “go-getter” who performs well when constrained but poorly when not