A multilevel model with autoregressive components for the analysis of tribal art prices

Abstract

In this paper, we introduce a multilevel model specification with time-series components for the analysis of prices of artworks sold at auctions. Since auction data do not constitute a panel or a time series but are composed of repeated cross-sections, they require a specification with items at the first level nested in time-points. Our approach combines the flexibility of mixed effect models together with the predicting performance of time series as it allows to model the time dynamics directly. Model estimation is obtained by means of maximum likelihood through the expectation–maximization algorithm. The model is motivated by the analysis of the first database ethnic artworks sold in the most important auctions worldwide. The results show that the proposed specification improves considerably over classical proposals both in terms of fit and prediction.

Keywords:

• auction prices
• tribal art
• dependent random effects
• hedonic regression model
• multilevelmodel
• repeated cross-sectional data

AMS Subject Classifications:

• 62-07
• 62F10
• 62J10
• 62J05
• 62P20

JEL Classifications:

• C01
• C15
• C58

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

The authors acknowledge the financial support from the grant [RBFR12SHVV] funded by the Italian Government (FIRB project ‘Mixture and latent variable models for causal inference and analysis of socio-economic data’).

ORCID

Simone Giannerini http://orcid.org/0000-0002-0710-668X