Linde Institute/Social and Information Sciences Laboratory (SISL) Seminar
Measuring the Performance of Large-Scale Combinatorial Auctions: A Structural Estimation Approach
The main advantage of a procurement combinatorial auction (CA) is that it allows suppliers to express cost synergies through package bids. However, bidders can also strategically take advantage of this flexibility, reducing the performance of the auction. In this paper, we develop a structural estimation approach for large-scale first-price CAs. We use bidding data to estimate the firms' cost structure and margins, and use this information to evaluate the performance of the auction in terms of its efficiency and payments to the bidders. The large number of bids commonly observed in large-scale CAs makes it difficult to use the structural methods developed in prior work. To overcome these limitations, we propose a simplified model of bidders' behavior where markups of each package bid are chosen based on a reduce set of package characteristics. We apply our method to the Chilean school meals auction, in which the government procures half a billion dollars worth of mea services every year and bidders submit thousands of package bids. Our estimates suggest that the bidders' cost synergies are economically significant in this application. The current CA mechanism achieves high allocative efficiency and reasonable competitive payments to bidders. We also find that a market share constraint imposed by the government to promote suppliers' diversification, results in only a small efficiency loss. Finally, we conduct some counterfactual experiments to compare the performance of the first-price CA with that of a Vickrey-Clarke-Groves (VCG) mechanism. In contrast to recent theoretical work criticizing VCG for leading to high procurement payments, we find that the total VCG payment is reasonable and close to the payment achieved in the first-price CA. The proposed structural estimation framework is quite general and can be applied to other large-scale CAs applications.
(joint work with Sang Won Kim and Marcelo Olivares)
Contact: Edith Quintanilla at Ext. 3829 email@example.com