A Heckit Model of Sales Dynamics in Turkish Art Auctions: 2005-2008

Abstract

Unsold artworks are excluded from a traditional hedonic price index as no observable price can be attached to them. The question is whether the exclusion of unsold artworks lead to a sample selection bias in the traditionally constructed art price indices. In this paper, we examine the art auction sales performance in Turkey for the period between January 2005 and February 2008 using a unique database which contains 11,212 sales records including unsold items. We employ the two-stage Heckit model. Our empirical model combines demand-side influences with supply-side characteristics as well as the auction microstructure. We find that there is no sample selection bias created by unsold works. This finding also provides an explanation for why the attempts in the literature to identify which works are (not) sold turned out to be largely unsuccessful. On the behavioural side, we confirm the existence of the âafternoon effectâ in both sales rates and in sales prices in Turkish art auctions. There is also some evidence for the âdeath effectâ and âmaster effectâ in both sales rates and sales prices. Finally, we find that the returns in the Turkish art market serve as a hedge against inflation in our sample period.