Introduction

A very useful laboratory for the study of the informational efficiency in markets is a parimutuel betting market. Participants assess the relative likelihoods that various horses will win and then bet on the basis of this analysis. From this betting we can deduce the market's aggregate assessment of the probability that a particular class of horse will win (the so-called subjective probability) and this can be compared to that class's objective probability of winning. The conventional view is that, if parimutuel betting markets were efficient, these probabilities ought to coincide. Unfortunately, a significant number of empirical studies have found they do not. Most often, favourites are underbet and longshots overbet. However there have been studies which have reported a reverse bias (Busche and Hall, 1988; Woodland and Woodland, 1994). This mismatching of subjective and objective probabilities is termed the ‘favourite-longshot bias’. The instance where favourites are underbet is termed the usual bias; where favourites are overbet, it is termed the reverse bias.

Not surprisingly there have been a number of explanations for the bias. The interested reader is referred to Thaler and Ziemba (1988) and Sauer (1998) for excellent summaries of the literature. One class of explanation appeals to bettor preferences. In particular, they posit that bettors are risk-lovers. This line of research would include the work of Weitzman (1965), Ali (1977), Quandt (1986), and Kanto, Rosenqvist and Suvas (1992). More recently Golec and Tamarkin (1998) have suggested that gamblers prefer return skewness rather than risk.