Biased Auctioneers
Summary
A neural network that predicts art auction prices from images and catalogue data reveals that auction houses' pre-sale estimates are informationally inefficient: when the algorithm's valuation exceeds the estimate, prices come in higher relative to the estimate and lots are less likely to go unsold, and auctioneers' errors are persistent enough to be predictable from their past errors.
Abstract
We construct a neural network algorithm that generates price predictions for art at auction, relying on both visual and non-visual object characteristics. We find that higher automated valuations relative to auction house pre-sale estimates are associated with substantially higher price-to-estimate ratios and lower buy-in rates, pointing to estimates' informational inefficiency. The relative contribution of machine learning is higher for artists with less dispersed and lower average prices. Furthermore, we show that auctioneers' prediction errors are persistent both at the artist and at the auction house level, and hence directly predictable themselves using information on past errors.
Previously circulated as "Machine Learning, Human Experts, and the Valuation of Real Assets."
How to cite
Aubry, Mathieu, Roman Kräussl, Gustavo Manso, and Christophe Spaenjers (2023). Biased Auctioneers. Journal of Finance 78, 795–833.
@article{aubry2023biased,
author = {Aubry, Mathieu and Kräussl, Roman and Manso, Gustavo and Spaenjers, Christophe},
title = {Biased Auctioneers},
journal = {Journal of Finance},
volume = {78},
number = {2},
pages = {795–833},
year = {2023},
doi = {10.1111/jofi.13203}
}