A Bayesian Dynamic Hedonic Regression Model for Art Prices

Abstract

Hedonic pricing models are based on the premise that the prices of marketed goods are related to their attributes. The traditional OLS regression applied to hedonic pricing models assumes that, when using time series, the estimated coefficients with respect to each of the attributes remain constant. This does not have to be the case. We propose a Bayesian dynamic estimation of the hedonic regression model in which the estimated coefficients are time-varying and apply it to art prices. We find that, using a sample of 27,124 paintings sold at auction by 63 Pop artists (2001-2013), the estimated coefficients from the dynamic regression model fluctuate noticeably through time, and also that certain types of artworks, which might be regarded as “safer”, declined in price by less than less safer paintings during the financial crisis (2008-2009). We also estimated a Pop-Art price index, finding that, in the semesters prior to the crisis, Pop-Art prices increased much faster than what an OLS estimation would have suggested.