Who won the contract? An analysis of bidder networks and collusion in public procurement

Abstract

This research analyses procurement markets as co-bidding networks, in which firms that sell goods and services to the government are linked if they bid together on the same tender, to detect suspicious bidding patterns, potential bid-rigging and collusive structures in public contracting. I argue that groups that have similar bidding patterns and interact frequently in the market are more likely to present collusion markers. The method is applied to public tenders in Paraguay from 2014 to 2018. Using a community detection algorithm based on the co-bidding similarity of firms and the frequency of interactions, suspicious groups were found in 23 of 76 markets studied. I applied collusion screens to each group, based on price differences of tenders won, market shares and the success rate of firms. The groups detected present different patterns of interaction that in some cases do not align with collusion practices, or do it partially. For instance, some of the flagged markets had a high concentration, stable market shares and were captured by members of the suspicious groups. Also, firms can use other strategies to win contracts, such as diving the market according to the geographic location, or bidding under different legal names. The method can be useful to detect anomalous behaviour and filter suspicious cases for further analysis in procurement markets, where no previous data about bid-rigging is available.