Bidders’ risk attitudes have important implications for sellers seeking to maximize expected revenues. In ascending auctions, auction theory predicts bid distributions in Bayesian Nash equilibrium does not convey any information about bidders' risk preference. We propose a new approach for inference of bidders’ risk attitudes when they make endogenous participation decisions. Our approach is based on the idea that bidders' risk premium -the difference between ex ante expected profits from entry and the certainty equivalent - required for entry into the auction is strictly positive if and only if bidders are risk averse. We show bidders' expected profits from entry into auctions is nonparametrically recoverable, if a researcher observes the distribution of transaction prices, bidders' entry decisions and some noisy measures of entry costs. We propose a nonparametric test which attains the correct level asymptotically under the null of risk-neutrality, and is consistent under fixed alternatives. We provide Monte Carlo evidence of the finite sample performance of the test. We also establish identification of risk attitudes in more general auction models, where in the entry stage bidders receive signals that are correlated with private values to be drawn in the bidding stage.