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.