Theoretical and experimental investigations of the performance of keyword auction mechanisms

Abstract

Two keyword auction mechanisms, the Generalized Second‐Price auction (GSP) and the Vickrey‐Clarke‐Groves mechanism (VCG), were compared theoretically and experimentally. The former is widely used in practice; the latter is not, but it has a dominant strategy equilibrium where all participants bid their true values. In the theoretical investigation, by applying the “locally envy‐free Nash equilibrium” to the VCG, we found that the allocations are efficient and that upper and lower bounds of the auctioneer's revenue coincide in the two mechanisms. A laboratory experiment, in which the revenues and efficiencies were similar in both mechanisms, supported this result.

An experimental study of sponsored-search auctions

“…While the dynamic interaction provides bidders with opportunities to learn and adjust their behavior, whether that process will lead them to behave "as if" they play the complete information static game assumed by the theory is far from clear. 13 We thus pose the following question:…”

“…14 13 There is a theoretical argument suggesting that Bayesian learning in repeated play would result in players behaving according to a full-information Nash equilibrium (see Jordan (1991), for instance), but the theory does not suggest how the players would "select" a Nash equilibrium, in case there are multiple Nash equilibria.…”

“…Our approach allows us to study whether the selection of a particular modeling framework is justified separately from whether the equilibrium prediction from the chosen model is empirically valid, and this approach can be useful beyond the current setting. In market design research, some design option entails a strategic environment that is too complicated to tractably analyze, and this forces analysts to settle on 4 The only other experimental work we are aware of is Fukuda et al (2013). Their main purpose is different from ours; they compare the performances between the GSP and VCG auctions (that is, not just the VCG outcome).…”

“…The frequencies of adaptive optimality begin from 0.68 (periods 2-5) to 0.74 (periods 11-15) in the DI-11 treatment and from 0.80 (periods 2-5) to 0.87 (periods [11][12][13][14][15] in the DI-20 treatment. When we examine this more finely according to the value ranking, the highest-value bidders exhibit the lowest level of adaptive optimality and the lowest-value bidders show the highest level of adaptive optimality in each period.…”

Sponsored search auctions: an overview of research with emphasis on game theoretic aspects

“…It should be noted that the comparison of the auctioneer's revenue under the VCG and the GSP mechanisms is further investigated in [35]. As mentioned above, it is established in [30] that the revenue under the VCG dominant strategy equilibrium constitutes a lower bound to the revenue of any SNE of the GSP mechanism.…”