A0509
Title: From red flags to red alerts: Prioritizing irregularities in public procurement through predictive modeling
Authors: Ivan Pastor - Universidad Internacional de La Rioja (Spain) [presenting]
Felix J Lopez Iturriaga - Universidad de Valladolid (Spain)
Abstract: While red flag indicators are commonly used to detect potential integrity risks in public procurement, their widespread presence often limits operational effectiveness. A high volume of flagged contracts can overwhelm oversight capacities and dilute attention from truly anomalous or suspicious cases. A complementary approach is proposed to enhance the practical utility of existing red flag systems. A neural network model is introduced to jointly predict two key outcomes for public contracts: The number of bids received and the post-award integrity score assigned by the open tender platform, a composite risk indicator based on procedural transparency, competition, and potential irregularities. By comparing the models' predictions with actual outcomes, contracts whose results deviate significantly and negatively from expected norms are flagged as high-priority cases for audit or further investigation. The empirical application covers over 200,000 high-value public contracts awarded in Spain between 2022 and 2024. The model achieves strong predictive performance (R2 = 71.96\% for bids; 59.15\% for integrity score), enabling a reliable benchmark of normal behavior. A small but relevant subset of contracts shows unexpected outcomes, such as receiving a single bid when healthy competition is expected. The approach supports smarter oversight and more efficient allocation of audit resources.
