Auction houses' confidence, bility and biases in predicting artworks' prices. Evidence from the auction sales of the Italian paintings.

2 1. Foreword The purpose of this study is to determine whether there are any bias in the pre-sale estimates of the paintings that are put up for sale at auction. In particular, I concentrate on paintings by Italian artists. For my analysis, I refer to a dataset of 1.975 repeated sales of Italian artists internationally. I want to test whether there is any “Country effect” (some Country providing better estimates than others) or any “auction house effect” (one auction house being more precise). In particular, I want to confront the Italian sale houses with the foreign ones, since one would expect auctioneers are better at predicting prices for the artists of their same home Country. The hypothesis I want to verify is whether Italian auctioneers know more about Italian art. My research shows that this seems to be true. I also check for any differences between artistic categories. The methodology that I use to verify all this is to define some variables that represent the reliability of the estimates (the most important is the difference between the sale price and the mean estimate) and then to run some regressions to determine which are the factors that influence these variables (whether they are the sale houses, the sale locations…). I inserted all data in an Spss database1, which is the same software that I use to run the regressions. My work is organized as follows. Before presenting the data I work on specifically, in chapter 2 I present a literature review of all studies that have been carried on till this moment about the art market. This field has been covered extensively, but most of the studies concentrated on building some indices to compute the yield on art investments. It is quite a different approach compared to determining the reliability of the estimates. Ultimately, this means to assess whether some auctioneers are better than others. By the way, nobody as conducted this kind of study considering one specific artistic production (like the Italian one in our case). A few studies that have been carried on about the estimates focused on geographical markets (single Countries), whose sale data are easier to collect. Instead, my dataset comprises a vast number of sales from everywhere in the world and hence I believe it leads to much more interesting observations. 1 Spss (Software package for social sciences) is a statistical software which is very convenient to use when dealing with huge quantities of data.