Information extraction and artwork pricing

Abstract

Traditional art pricing models often lack fine measurements of painting content. This paper proposes a new content measurement: the Shannon information quantity measured by the singular value decomposition (SVD) entropy of the painting image. Using a large sample of artworks' auction records and images, we show that the SVD entropy positively affects the sales price at 1% significance level. Compared to the other commonly adopted content variables, the SVD entropy has advantages in variable significance, sample robustness as well as model fit. Considering the convenient availability of digital painting images and the straightforward calculation algorithm of this measurement, we expect its wide application in future research.