We analyse credit default swap settlement auctions theoretically and evaluate them empirically. In our theoretical analysis, we show that the current auction design may not result in the fair bond price and suggest modifications to the auction design to minimize mispricing. In our empirical study, we find support for our theoretical predictions. We show that an auction undervalues bonds by 10%, on average, on the day of the auction and link this undervaluation to the number of bonds that are exchanged during the auction. We also document a V-shaped pattern in underpricing during the days surrounding the auction: in the days leading up to the auction, the extent to which bonds are underpriced declines, while after the auction, the extent to which they are underpriced increases, with the smallest underpricing coming on the day of the auction.