Replication Data and Code for Underbidding for Oil and Gas Tracts

Common values auction models, where bidder decisions depend on noisy signals of common values, provide predictions about Bayesian Nash equilibrium (BNE) outcomes. In settings where these common values can be estimated, these predictions can be tested. We propose a series of tests, robust to assumptions about the signal structure, to determine whether the observed data could have been generated by a Bayesian Nash equilibrium. In the setting of oil and gas lease auctions in New Mexico, we find evidence that participation decisions are correlated and that participants systematically underbid in light of ex post outcomes.