Econometric fine art valuation by combining hedonic and repeat-sales information

Statistical methods are widely used for valuation (prediction of the value at sale or auction) of a unique object such as a work of art. The usual approach is estimation of a hedonic model for objects of a given class, such as paintings from a particular school or period, or in the context of real estate, houses in a neighborhood. Where the object itself has previously sold, an alternative is to base an estimate on the previous sale price. The combination of these approaches has been employed in real estate price index construction (e.g., Jiang et al. 2015); in the present context, we treat the use of these different sources of information as a forecast combination problem. We first optimize the hedonic model, considering the level of aggregation that is appropriate for pooling observations into a sample, and applying model-averaging methods to estimate predictive models at the individual-artist level. Next, we consider an additional stage in which we incorporate repeat-sale information, in a subset of cases for which this information is available. The methods are applied to a data set of auction prices for Canadian paintings. We compare the out-of-sample predictive accuracy of different methods and find that those that allow us to use single-artist samples produce superior results, that data-driven averaging across predictive models tends to produce clear gains, and that, where available, repeat-sale information appears to yield further improvements in predictive accuracy.