Are Art Auction Estimates Biased?

This article considers whether presale auction estimates are unbiased predictors of price when “no‐sales” are considered utilizing a newly constructed sample of over 500 works by eight early twentieth‐century American artists. Unbiased presale auction estimates in predicting price, while expected, are generally not supported in previous work, but these studies (excepting one) do not include no‐sales in the calculations. In order to study the question, we employ a standard approach that uses an inverse Mills ratio arising from a sample selection probit to correct for selection bias. We find that controlling for selection bias, presale auction estimates appear to be biased downward, and we offer possible reasons for this result.