Principal investigator: Eduarda Miller Figueiredo
Original title: Collusion in Auctions with Constrained Bids: Theory and Evidence from Public Procurement
Authors: Sylvain Chassang and Juan Ortneer
Location of the Intervention: Ibaraki Prefecture (Japan)
Sample Size: 10.553
Sector: Public sector
Primary Variable of Interest: Bids at auctions
Type of Intervention: Policy change in the distribution of winning bids in auctions.
Methodology: Game Theory and Diff-in-Diff
Summary
This study investigates the impact of minimum price restrictions on collusion in the Japanese construction industry. Researchers analyzed project data from public works auctions held between 2007 and 20116 in Ibaraki Prefecture, Japan. The results indicate that keeping contracts less complete than necessary plays a significant role in practice and can be used to reduce the occurrence of collusion among companies. It concludes that implementing a minimum bid price is always more effective than the absence of such a measure.
- Policy Problem
By studying the mechanics of cartel application and its interaction with constraints in the context of repeated auctions, this article showed that in the presence of colluding bidders, attempts to extract surpluses can foster collusion and reduce the auctioneer's surplus. Conversely, providing minimum surplus guarantees can limit collusion and improve the auctioneer's surplus.
Literature on cartels considers scenarios in which cartel members may compromise through mechanisms and, therefore, argues that proper auction design can successfully limit collusion, provided participants have the financial resources and are able to make payments. ex-ante (Pavlov, 2008; Che and Kim, 2009). When studying complete implementation in repeated environments using dynamic mechanisms, Lee and Sabourian (2011), as well as Mezzetti and Renou (2012), show that implementation in all equilibria can be achieved by restricting the set of continuation values available to players to support repeated game strategies.
McAfee and McMillan (1992) already show that collusion makes lower maximum prices desirable, and, following this reasoning, the authors of this article analyzed here argue that higher minimum prices can help weaken cartels.
- Implementation and Evaluation Context
Ishii (2008) and Kawai and Nakabayashi (2014) provide evidence of widespread collusion in Japanese procurement auctions. This suggests that procurement in Japan is an environment where minimum price restrictions could have a plausible effect on the issue of collusion in the Japanese construction industry.
Based on the reasoning presented above, the authors have three main objectives:
- To provide transparent insight into how bidding restrictions (minimum prices) can affect cartel behavior and bid distribution;
- To empirically assess whether enforcement restrictions are a significant determinant of cartel behavior;
- Explore this understanding of cartel behavior to develop a collusion test.
- Policy/Program Details
It is known that governments need to procure construction services on an ongoing basis, facing a limited and stable pool of firms that can potentially perform the work, and a subset of which participate regularly. Legislation often requires participants to register, and governments disclose bids and results after the conclusion of each auction. The repeated and public nature of the interaction makes collusion a realistic concern.
For the research, the authors used available data from public works projects auctioned between May 2007 and March 2016, corresponding to 10.553 projects, which took place in the 30 most populous cities of Ibaraki Prefecture (Japan). The control cities did not undergo political change in the distribution of winning bids during the period, while the treatment cities did.
- Assessment Method
To perform the evaluation, the authors used a game theory model. In this model, at each period (tA buyer acquires a single unit of a good through a first-price auction. In each period, a subset of companies is eligible to participate in the auction. And in each period (tEach participating company can deliver the goods at a cost. Companies can send transfers to each other regardless of whether or not they participate in the auction. And it is assumed that all companies belong to the cartel and monitor each other's production costs.
The acquisition contract is awarded according to the first-price auction, with restricted bids; that is, bids outside a price range are discarded. The winner is the bidder with the lowest bid, but they do not deliver the goods for the price they offered.
The interaction is repeated, and firms can use the promise of continued collusion to enforce compliant bids and transfers. The bids and transfers must be part of a perfectly balanced subgame of the repeated game between firms. Given the entire game, under complete information, the unique outcome of competitive equilibrium is such that the winning bid equals the maximum between the second-lowest cost and the minimum price. The contract is awarded to the lowest-cost bidder whenever the winning bid is above the minimum price and is randomly allocated among all bidders with costs below the minimum price when the winning bid equals the minimum price.
To measure the impact of a policy change on the distribution of winning proposals at the city level, the authors used the methods “change-in-changes"[1] or estimation by differences in differences.
- Main results
Under collusion, the introduction of a small minimum price should lead to a drop in the distribution of the right tail of winning bids. Under competition, such a change was not expected.
The results of estimations using quantile regressions show that policy change is associated with a decrease in dominance in the right tail of winning bids. The implication is not only that collusion exists, but that restrictions on the application of the cartel are binding and that the sustainability of the collusion is limited by price constraints.
In studying who is affected by the policy change, the authors realize that in the absence of minimum prices, long-term firms obtain contracts at higher prices. The introduction of minimum prices has a disproportionately greater impact on long-term winners than on new winners.
Difference-in-differences analysis assumes that control cities are unaffected by the policy change. A potential concern is that some of the long-term bidders active in a treatment city may also be active in control cities. If this is the case, the introduction of minimum bids in a treatment city could also cause a shift in the bid distribution in the control cities.
- Lessons in Public Policy
The conclusion of this article is that implementing a minimum price in the observed bids always outperforms the absence of a minimum price. When there is no collusion, this does not impact the distribution of bids. However, in the presence of collusion, it may only reduce the dispersion of bids.
The authors also provide empirical evidence that the mechanism of keeping contracts more incomplete than necessary plays a significant role in practice and can be used effectively to influence collusion between companies.
References
Athey, S. and GW Imbens (2006): “Identification and Inference in Nonlinear Difference-in-Differences Models”, Econometrics, 74, 431-497.
Che, Y.-K. and J. Kim (2009): “Optimal collusion-proof auctions“, Journal of Economic Theory, 144, 565-603.
Ishii, R. (2008): “Collusion in repeated procurement auction: A study of a paving market in Japan”, Tech. rep., ISER Discussion Paper, Institute of Social and Economic Research, Osaka University.
Kawai, K. and J. Nakabayashi (2014): “Detecting Large-Scale Collusion in Procurement Auctions”, Available at SSRN 2467175.
Lee, J. and H. Sabourian (2011): “Efficient Repeated Implementation”, Econometrics, 79, 1967-1994.
McAfee, R.P. and J. McMillan (1992): “Bidding rings,” The American Economic Review, 579-599.
Mezzetti, C. and L. Renou (2012): “Repeated Nash Implementation”, Available at SSRN 2096184.
[1] Athey and Imbens (2006).
