By Lisa Farrell, Jane M. Fry and Tim R.L. Fry

Many studies of art auctions assume that a single statistical model using observed characteristics relating to the artwork and the auction can explain the observed variation in the sample of all artworks and artists. We show that such “pooling” is not always appropriate and may lead to erroneous conclusions.

Empirical studies of the determinants of sales and price in art auctions typically pool together the data on all artworks by all artists and estimate a single statistical model for the determinants of a sale and/or the hammer price of an artwork from the data. Whilst efforts may be taken to ensure that the sample is relatively homogenous, the maintained assumption is that a single statistical model using observed characteristics relating the artwork and auction can explain the observed variation in the data.

In our paper, we examine this assumption using data on the top three selling, by financial value, Australian Indigenous artists: Emily Kngwarreye, Albert Namatjira and Rover (Julama) Thomas. Each of these artists is primarily known for one style of painting – Namatjira for watercolour on paper, Kngwarreye for acrylic (synthetic polymer) on canvas and Thomas for natural earth pigments on canvas. Indeed, they have distinct differences in their body of work in method, in style and technique and in subject matter.

The data used comes from The Australian Art Sales Digest ( and comprises all artworks offered for sale at auction between 1969 and 2014. The data for our chosen artists is summarised in Table 1.

Table 1: Summary characteristics of artists and their auction history

Simple descriptive analysis of both clearance rates and hammer prices clearly indicates that the three artists are different. The key question, however, is can such differences be explained within a single model using a pooled sample and the observed characteristics relating to the artworks and auction?

Our chosen statistical modelling framework is a sample selectivity model. In this framework, we simultaneously model the likelihood (probability or propensity) that an artwork sells at auction and, given that it sells, the hammer price that it achieves. Pooling in our context means that we take all of the data on the three artists as a single sample and estimate a single sample selectivity model. The alternative approach is to take the data for each artist separately and estimate a sample selectivity model for each individual artist. A simple statistical test is then used to determine whether the individual models differ from the single pooled model and thus whether pooling is appropriate.

In our paper, rather than use hedonic characteristics of the artworks we use instead a pre-sale estimate of the hammer price made by the auction house. Pre-sale information on an artwork comes in the form of a lower and upper bound on the likely hammer price. Auction houses claim that such estimates take into account the characteristics of the individual artwork, the market for such artworks and the identity of the artist. Most artworks in our data have such pre-sale estimates (see Table 1). Consistent with previous studies we take the mid-point of these two estimates as our pre-sale price measure. The other observable factors that we use in our paper are the identity of the auction house conducting the auction, the number of Indigenous artworks being offered in the auction and a control for any time varying effects.

The specification of the sample selectivity model that we use is guided by the existing empirical literature. Our selection equation for sold (Yes/No) is non-linear (quadratic) in the pre-sale estimate, contains indicators for the auction house to allow for the fact that the view and conduct of an auction can differ by auction house, and the number of Indigenous artworks in the auction is included as is a time variable. The hammer price equation includes the pre-sale estimate, auction house indicators and a time variable.

Our results show that pre-sale information on the artwork and auction effects are significant in determining sale and, if sold, hammer price for Australian Indigenous artworks offered for sale. Crucially, we find that the three artists studied are very different from each other and thus the data for these artists should not be pooled into a common sample for analysis. We recognise that our study was only possible because we had relatively large samples for each of our individual artists. We, therefore, suggest that our approach could be used whenever a sample size of one hundred or more for each artist exists or that researchers might consider “partial pooling’ where separate models for groups of relatively homogenous artists (e.g. schools of Surrealist artists) could be estimated and tested against the “full pooling” approach.

Finally, our main conclusion that we should not pool may suggest that defining an art movement by the ethnic origin of its creators is misleading when it comes to assigning secondary market value to such artworks. From policy, social and cultural heritage perspectives, such definitions may be useful, but our evidence suggests buyers of Australian Indigenous art assign value to an artwork from information over and above just the ethnic origin of the artist.

This article is based on:

Lisa Farrell, Jane M. Fry and Tim R.L. Fry “Determinants of sales and price at auction for three Australian Indigenous artists: to pool or not to pool?” in Journal of Cultural Economics. Published online: November 2017.

About the authors:

Lisa Farrell is Professor of Economics in The School of Economics, Finance and Marketing at RMIT University.

Jane M. Fry is a PhD student in The Centre for Health Economics at Monash University.

Tim R.L. Fry is Professor of Econometrics and Head of The School of Economics, Finance and Marketing at RMIT University. Prof. Fry is also the Initiator and Head of Organization of the 20th International Conference on Cultural Economics (ACEI) at RMIT University in Melbourne in June, 26-29, 2018.

About the image:

Aboriginal art image. Available on Pixabay under Creative Commons CC0,

May 13, 2024
The Economic Impact Of Unesco World Heritage Designations: Evidence From Italy
April 29, 2024
Rethinking Music Streaming Royalties – How Do We Better Align Consumer Value And Artist Payments?
April 15, 2024
Cultural Capital Externalities
April 1, 2024
Examining The Potential Disruption From Ad-supported Streaming Services

Become a member

Members of the ACEI will be part of a network of scholars, researchers and practitioners interested in advancing cultural economics.

Join Today