The Value of Art

Abstract

This thesis studies how non-monetary signs of social, historical, and artistic significance are translated into monetary value in the global fine art auction market. Using a dataset of fine art auction sales from 1986 to 2025, we analyze two outcomes: realized hammer price and whether a lot is bought in. We estimate hedonic regressions with fixed effects, difference-in-differences and event study models around moments of institutional recognition, and neural network models for auction outcome prediction. The results show that institutional markers of recognition are strongly associated with higher prices and greater sale success. In the fully specified hammer price model, recorded provenance, exhibition history, and literature references are all linked to large price premia, while in the bought in model provenance and literature are associated with lower probabilities of failing to sell. Event study and difference-in-differences results further suggest that institutional validation can shape market outcomes, though unevenly: works by female artists experience a modest relative improvement after 2007, non-European artists show heterogeneous post-1989 responses across regions, and artist death is associated with a short-run posthumous price premium. The neural network results reinforce these findings while clarifying the limits of prediction. The hammer price model modestly outperforms the fixed-effects regression, while predictive importance is driven primarily by artist identity and medium rather than by institutional variables alone. The bought in model performs much more weakly, suggesting that sale outcomes depend more heavily on unobserved factors. Auction prices, then, should not be read as transparent measures of the value of art. Rather, they show how judgment becomes price, and how price in turn shapes the making of artistic meaning.