that, unlike in LS they observe their signals before paying the entry cost. This reverses the LS predictions that the optimal reserve price is zero and entry fee positive (zero) in a common (private) value framework. Also Menezes and Monteiro (2000) change the LS setting so that bidders learn their valuations before paying the entry cost. Ye (2004) generalizes the LS model so that bidders can update their beliefs after entry. In an ongoing work Harstad (2005) extends the endogenous entry model further by introducing affiliation and generalizing the information flows. Cox et al. (2001) test experimentally a model of endogenous entry, exit and bidding in common value auctions. They find that observed entry is lower than predicted by the model and that winner’s curse occurs among inexperienced bidders. Gal et al. (2007) extend the LS model by allowing the entry costs to differ across bidders. The resulting game is bi-dimensional, because bidders receive signals on both the value and the entry cost. They find that partially reimbursing the entry costs of high entry cost types increases revenue.
Auctions where bidders have both private and common components in their valuations have drawn some attention in the recent literature. More generally they belong to a class of auctions where the essential feature is that signals on valuations are multidimensional, because there is a signal for both the private and common component. Maskin (1992), Dasgupta and Maskin (2000) and Jehiel and Moldovanu (2001) have
60
studied these auctions. The main feature is that no auction format is efficient when sig- nals are multidimensional although Pesendorfer and Swinkels (2000) are able to derive some conditions for restoring efficiency in a uniform price auction with many bidders. These auctions are inefficient because a bidder with an overly optimistic conjecture about the common component may outbid a bidder with a higher private valuation. Jackson (2005) shows that equilibrium fails to exist when valuations have both com- ponents in second price and English auctions. Goeree and Offerman (2003) are able to derive the equilibrium by aggregating the multidimensional signals into a single statis- tic by relying on the independence assumption. They study the effects of competition and information disclosure in this set up and find that increasing both reduce the inefficiency. Goeree and Offerman (2002) use an experimental setup to study efficiency in these auctions and find observed efficiencies close to the predicted ones. Compte and Jehiel (2002) propose a model where bidders know their private component and are differently informed about the common component. They study the welfare effect of adding one bidder to one-object sealed-bid second price and ascending auctions. They find that with symmetric bidders, an extra bidder is good for welfare whereas with asymmetric information about the common element this is not the case. Compte and Jehiel (2002) tackle bidder asymmetry in their model, but that particular asymmetry is different from what I am interested in. They assume that bidders draw their signals from different distributions. In my auction data, the bidders are asymmetric also in the sense that they put different weights on the private and common value component. In the extreme case, one bidder could be a pure common value bidder and another have independent private values. Asymmetry in this dimension has not been addressed in the literature before. The theoretical analysis of a model with both common and private values that allows for this particular form of asymmetric and endogenous entry is beyond the scope of this paper and possibly beyond the reach of existing analytical tools. I focus on an empirical analysis on how differences in the information paradigm affect the entry choices. Although I do not impose this complex information structure on the estimations explicitly, this is still the first empirical analysis of auctions that studies bidder behavior when values have both private and common component.