Association for Public Policy Analysis & Management

Open electronic auction is the most popular form of public procurement in many countries. Public organisations often lack the resources and expertise to run a tender themselves. Instead, they delegate responsibility for conducting procurement auctions to an auctioneer with industrial experience. This decentralised approach, and the amount of money involved, make public procurement especially prone to collusion and corruption which can reduce competition. In fact, many procurement auctions attract only a single bidder. Policymakers often view transparency as a panacea for collusion in procurement. But, paradoxically, policies meant to increase transparency often lead to more successful collusion agreements.

To help explain this phenomenon, I formulate a model of procurement auctions which includes bid tailoring. Bid tailoring happens when a procurement employee with responsibilities tailors auction specifications to give an unfair advantage to a favoured bidder. In my model, the auctioneer is dishonest and cares about her expected share of the winner's rents. She increases a winning probability of her favoured bidder by disqualifying the competitors on the bases of rigged specification and in exchange for a kickback. The favoured bidder thus has an incentive to overstate his costs of procurement and bid higher to increase his rents from procurement contracts. The model also predicts that bid tailoring leads to welfare loss by a procurer who incurs higher costs.

My model builds on auction theory. However, it adds several new pertinent features to existing auction models. First, the favoured bidder is better informed about auction specifications and his chances to get a contract and thus has a different equilibrium bidding strategy than the other bidders. Second, to model the auctioneer's behaviour, I use the discrete choice approach and allow the estimated auction rents to affect the auctioneer's utility. Once the parameters of the model are recovered, they will be used to figure out how the auction rent is split between an auctioneer and a bidder. This approach also allows to estimate the effectiveness of the bonus reward system that can be implemented instead of the existing fixed-salary system to motivate the integrity of procurement auctioneers. Finally, I propose a method of model identification which is based on an observed bid evaluation decision of an auctioneer and missing bids. These three properties help advance both the auction and corruption literatures.

I apply the model to the large dataset containing over 2 million of procurement purchases from a developing country where bid tailoring has been proven to be a concern. To determine which auctions in the data are suspicious of collusion, I develop a statistical test to detect bid tailoring which uses red flags of collusion, in particular, the unusually low number of reported. The reduced-form evidence suggests that two effects present: the procurement prices are 46.1% higher in auctions where some bids were not reported; and though the average number of bidders in such auctions is considerably higher the actual number of admitted and reported bidders is 20% lower. And these patterns are not explained by a non-collusive bidding model.