Frequent Losers as Cover Bidders¶
Screening for Bid Rigging in Public Procurement
Insper Institute of Education and Research, Sao Paulo, Brazil
4--21% Price markup from cover bidding by frequent losers (OLS lower bound to IV upper bound)
Abstract¶
We propose frequent losers (FL)---firms that never win yet participate in abnormally many public tenders---as a screening marker for cover bidding in procurement auctions. Using 4.5 million tender-items from Sao Paulo's BEC platform (2009--2019), we identify 2,735 FL firms via an IQR-based threshold. FL presence is associated with 4--9% higher negotiated prices (OLS with item, year, and purchasing-unit fixed effects). External validation shows 7.1% of FL firms co-participate with CADE-convicted cartelists, 3.5 times the rate expected by chance (\(p < 0.001\)). A leave-one-out instrumental variable yields a 21% price markup (\(F = 396\)), consistent with OLS attenuation. Bajari--Ye tests reject exchangeability and conditional independence for FL bid residuals. The screen requires only participation data, making it deployable by any competition authority with electronic procurement records.
JEL Classification: D44 D73 H57 K21 L41
Keywords: bid rigging cover bidding public procurement cartel screening frequent losers instrumental variables
Key Findings¶
FL presence raises procurement prices
Tenders with frequent loser participation exhibit 4--9% higher negotiated prices (OLS) and up to 21% higher prices (IV), after controlling for item, year, and purchasing-unit fixed effects.
External validation with CADE convictions
7.1% of FL firms co-participate with CADE-convicted cartelists---3.5 times the rate expected by chance (\(p < 0.001\), permutation test). The FL--price association persists after excluding all CADE-involved markets.
Cover bidding is the dominant mechanism
Bajari--Ye tests reject exchangeability and conditional independence for FL bid residuals. FL-present tenders have more genuine competitors (not fewer), ruling out competitive displacement. The evidence points to complementary cover bidding (Regime 1).
Deployable with minimal data
The FL screen requires only participation and outcome data (no bid values), making it operational for any competition authority with electronic procurement records. Estimated welfare loss: R$ 734 million (OLS) to R$ 2.4 billion (IV).