A Machine Learning Approach to Analyze and Support Anti-Corruption Policy

Can machine learning support better governance? In the context of Brazilian municipalities, 2001-2012, we have access to detailed accounts of local budgets and audit data on the associated fiscal corruption. Using the budget variables as predictors, we train a tree-based gradient-boosted classifier to predict the presence of corruption in held-out test data. The trained model, when applied to new data, provides a prediction-based measure of corruption that can be used for new empirical analysis or to support policy responses. We validate the empirical usefulness of this measure by replicating and extending some previous empirical evidence on corruption issues in Brazil. We then explore how the predictions can be used to support policies toward corruption. Our policy simulations show that, relative to the status quo policy of random audits, a targeted policy guided by the machine predictions could detect almost twice as many corrupt municipalities for the same audit rate. Similar gains can be achieved for a politically neutral targeting policy that equalizes audit rates across political parties.