Gotham city. Predicting ‘corrupted’municipalities with machine learning

The economic costs of white-collar crimes, such as corruption, bribery, embezzlement, abuse of authority, and fraud, are substantial. How to eradicate them is a mounting task in many countries. Using police archives, we apply machine learning algorithms to predict corruption crimes in Italian municipalities. Drawing on input data from 2011, our classification trees correctly forecast over 70 % (about 80 %) of the municipalities that will experience corruption episodes (an increase in corruption crimes) over the period 2012–2014. We show that algorithmic predictions could strengthen the ability of the 2012 Italy's anti-corruption law to fight white-collar delinquencies and prevent the occurrence of such crimes while preserving transparency and accountability of the policymaker.