Using information for 3,562 thoroughbreds listed in Keeneland’s 2011 September yearling sale, hedonic pricing models were estimated using both ordinary least squares regression and a Heckman selection model to test the adverse selection hypothesis that vertically integrated sellers whom breed and race are penalised with bid shading relative to sellers whom only breed, ceteris paribus. Though the null hypothesis of no adverse selection was not rejected in the standard regression model, when taking into account censoring due to sellers withdrawing listed yearlings prior to the auction the null hypothesis was rejected. However, the impact of adverse selection was greatly attenuated by bid shading associated with the reputation effect of sellers, both those whom are vertically integrated and those whom just breed horses, willing to accept final bids less than the stud fee further lessening the winner’s curse in the auction. This seller type is new to the literature and as such is a key innovation of this research. The models controlled for other observable characteristics of the yearlings as well as those of the seller and the design of the auction.