Analysis of a subsidised microcredit programme for Uruguayan entrepreneurs: the role of structural arrears

This study evaluates artificial neural networks (ANN) as credit-scoring tools for microloan allocation in a public non-profit microfinance institution. Using data from vulnerable Uruguayan entrepreneurs, the model incorporates borrower, business, and loan characteristics available at application. Results show that ANN substantially outperform traditional approaches in predicting structural default, defined as repayment of half or fewer scheduled instalments, rather than minor or temporary delinquency. The findings demonstrate that low-cost, readily available information can improve default-risk assessment. ANN can complement loan officers’ judgement, optimise resource allocation, and reduce structural default, thereby strengthening the institution’s financial sustainability and resilience.one