Identifying corruption risk profiles in public procurement over emergency periods: a latent class - Item Response Theory approach

Frequent crises require adapting quality assessment systems for public and private services to ensure timeliness, equity, and sustainability while addressing corruption risks. Public procurement, a key component of crisis response, is particularly vulnerable to corruption due to regulatory relaxations, increased spending, and dynamic market conditions. This work relies on an approach that integrates traditional red flag detection with a supervised machine learning approach to identify corruption risks based on historical data and statistical testing. Using a big data source on Italian tenders and a multidimensional Item Response Theory model, we analyze public procurement during the Covid-19 crisis. Awarded companies are classified into subgroups based on their corruption risk profiles, accounting for latent and multidimensional risk factors. The findings contribute to improving corruption risk assessment in crisis-driven procurement systems, supporting more effective and transparent governance.