Media Briefings

AUCTION DESIGN: NEW EXPERIMENTAL EVIDENCE

  • Published Date: April 2009

Sealed bid auctions typically raise much higher revenues than open auctions when several units of the same good are sold simultaneously. That is the one of the findings of experimental research by Dirk Engelmann and Veronika Grimm, published in the April 2009 Economic Journal. The results have important implications for current debate about the ideal design for electricity markets.

Auctions play an important role in everyday business life. An increasing number of important markets are organised as auctions. Prominent examples are electricity spot markets, private and public procurement, and spectrum auctions.

Auction rules are often developed by economists who use theoretical models to analyse the strategic incentives of the bidders. But when it comes to ‘multi-unit auctions’, where several units of the same good are sold simultaneously, theoretical models often do not allow exact conclusions on bidder behaviour and outcomes. Bidders’ behaviour may also differ systematically differ from that of a ‘rational and payoff-maximising agent’.

Engelmann and Grimm’s study contrasts the theoretical prediction of behaviour in five different multi-unit auctions with the actual behaviour of bidders whom they recruited to play those auctions in a controlled laboratory environment.

The results show that some of the predictions of theoretical models of auctions are accurate and are thus helpful in designing auctions. But they also demonstrate where bidders systematically deviate from the predicted outcome – and why.

The auction formats the research examines are different in two dimensions: first, the auctions are either open (where bidders can observe each other’s behaviour) or sealed bid; and second, the auctions differ in the pricing rule.

The results of the laboratory experiments show that low price (or collusive) outcomes are indeed more often achieved when this is predicted by theory: if the auction is open – so that bidders can observe each others’ behaviour – and all bidders pay the same price for each unit they receive.

Contrary to what theoretical models predict, however, the auction revenue depends on whether an auction is open or sealed bid than on the particular pricing rule. Sealed bid auctions in general raise much higher revenues.

The main reasons are that bidders cannot observe each other’s behaviour and thus, coordination on low prices is more difficult. Moreover, bidders in sealed bid auctions often bid as if they focus on getting a unit rather than on the price they pay and do not learn that this is not optimal because – in contrast with open auctions – they typically receive little feedback on the bidding process.

These results have relevance, among other markets, to a current debate in electricity market design about whether auctions should use a uniform-price or a pay-as-bid rule. The uniform-price auction is susceptible to collusion, leading to inefficient

allocations and low auction revenues, as the study has shown in line with theoretical results.

For pay-as-bid auctions, for the setting considered in this study, theoretical considerations suggest that each bidder should bid the same price on all units. But this is not what bidders do: in fact, in pay-as-bid auctions, the different bids of individual bidders typically differ a lot – implying substantial efficiency losses compared with the theoretical prediction.

The study demonstrates that changes in market design have to be carefully evaluated beforehand – and in particular that purely theoretical considerations are not always sufficient.

Among the auctions analysed in the study is an auction format proposed by Larry Ausubel, economics professor at the University of Maryland. The ‘Ausubel auction’ is an open format that does not suffer from incentives to coordinate on low prices. The auction performs quite well – achieving efficient outcomes and relatively high prices, in particular, after experimental participants have gained some experience with the mechanism.

In future research, Engelmann and Grimm plan to investigate behaviour in more complicated auction environments in order to contribute further to the advancement of auction and market design.

ENDS

Notes for editors: ‘Bidding Behaviour in Multi-unit Auctions – An Experimental Investigation’ by Dirk Engelmann and Veronika Grimm is published in the April 2009 issue of the Economic Journal.

Dirk Engelmann is at the Department of Economics, Royal Holloway, University of London. Veronika Grimm is at Friedrich-Alexander-Universitaet Erlangen-Nuernberg, Lehrstuhl für Volkswirtschaftslehre, insbesondere Wirtschaftstheorie, Lange Gasse
20, 90403 Nürnberg.

For further information: contact Dirk Engelmann on +44 1784 44 3968
(email: dirk.engelmann@rhul.ac.uk; website: http://www.rhul.ac.uk/Economics/For- Staff/engelmann.html); Veronika Grimm on +49 (0)911/5302-224 (email: veronika.grimm@wiso.uni-erlangen.de; website: www.wirtschaftstheorie.wiso.uni- erlangen.de); or Romesh Vaitilingam on 07768 661095 (email: romesh@vaitilingam.com).