While the possibility of adverse selection is present in many transactional settings, online auctions appear to be especially susceptible to the problem. Unlike buyers in most traditional settings, online auction shoppers are physically unable to inspect the products for sale and must rely on pictures and descriptions provided by the seller. If buyers cannot distinguish quality until after the purchase has been made, there is no incentive for sellers to provide high quality products. As a result, buyers will be unwilling to pay a quality premium, the average quality in the market will decline, and the level of trade will fall to a level below what is socially optimal. Accordingly, if adverse selection exists in online auctions, the quality of items traded would be expected to be subaverage. In addition, we would expect the decreases in online prices to be larger than in offline prices as the variance in the condition of items increases. Using data from completed eBay Motors vehicle auctions, we test both assumptions and examine the ability of online reputation systems to offset the effects of adverse selection. Our results suggest that adverse selection is more pronounced in online auctions compared to traditional marketplaces for used goods, and that reputation systems reduce, but do not fully eliminate, the problem. Consistent with theoretical predictions, we find that as vehicle age and mileage increase (i.e., as the variance of a car’s condition increases), the price that eBay buyers are willing to pay for the vehicle decreases by amounts greater than we would expect offline. In addition, we find that newer cars and those with low mileage are less likely to be sold on eBay. Further, sellers of higher quality vehicles and those with poorer reputations are more likely to protect themselves from the effects of adverse selection by setting reserve prices, and sellers with better reputations are more likely to sell their vehicles and receive higher price premiums.