This paper analyzes risk aversion in discriminatory share auctions. I generalize the k‐step share auction model of Kastl (2011, 2012) and establish that marginal profits are set‐identified for any given coefficient of constant absolute risk aversion. I also derive necessary conditions for best‐response behavior, which allows determining risk preferences from bidding data. Further, I show how the bidders' optimality conditions allow computing bounds on the marginal profits that are tighter than those currently available. I use my results to estimate import rents from Swiss tariff‐rate quotas on high‐quality beef. Rents are overestimated when ignoring risk aversion, and rent extraction is underestimated. Small bidders (small, privately owned butcheries) are more risk averse than large bidders (general retailers). Best response violations are few and uniform across bidder sizes.