Jump Bidding in Public Procurement Descending Auctions

Advisor: Nathalie Gimenes

Co-advisor: Lucas Lima

I study why bidders place jump bids—price cuts larger than necessary to take the lead—in descending procurement auctions and how these cuts affect contracted prices. Using 2,961,454 Brazilian federal procurement auctions from 2015 to 2018, I document that jump bidding is widespread among bidders but used selectively: 91.7% of usual bidders jump at least once, yet jumps account for only 11.1% of bids. I evaluate alternative explanations by considering auction rules, within-auction variation across phases and price states, and responses to jumps. The evidence is most consistent with a response to the risk that the auction ends before a bidder can bid again. Monitoring costs may also play a role. Rival reactions are inconsistent with deterrence, and bidder anonymity and standardized goods reduce the plausibility of reputation and common-value signaling as general explanations. Finally, I estimate the price effect by instrumenting jump bidding with random-phase duration. Under exact exclusion, each additional duration-induced jump is associated with a 4.25 percentage point reduction in the contracted-price-to-reference-value ratio, a 6.3% reduction from the mean of 0.68. Allowing for a direct effect of the instrument to construct sensitivity ranges places the reduction between 1.59 and 2.11 percentage points (2.3–3.1% relative to the mean). In a sample where non-credible standing bids obscure jump size, the estimate falls to a statistically significant but economically negligible 0.15 percentage points. This pattern is consistent with the effect operating through the information that a jump conveys rather than through the price cut itself.