Expert analysis and insider information in horse race betting: regulating informed market behavior.

• Author: Peirson, John
• Position: Symposium

1. Introduction

   The setting of odds on gambling events has been the subject of much academic attention over the last 30 years: See the reviews by Sauer (1998), Thaler and Ziemba (1988), Vaughan Williams (1999), and Coleman (2004). In particular, the impact of the use of insider information by gamblers has been examined in a large number of theoretical and empirical studies: See the work of Dowie (1976); Crafts (1985); Henery (1985); Shin (1991, 1992, 1993); Schnytzer and Shilony (1995, 2003, 2005); Vaughan Williams and Paton (1997); Paton, Vaughan Williams, and Fraser (1999); and Schnytzer and Snir (2008). Additionally, the regulation of the use of information in gambling and prediction markets has become a subject of great interest: See the work of Paton, Siegel, and Vaughan Williams (2009).

   The purpose of the present study is to develop a realistic model of insider and expert gambling on horse racing and the odds offered by U.K. bookmakers. The impact of insider trading and betting by experts leads to an increase in the uncertainty faced by bookmakers and leads them to reduce the odds offered on events for which there is greater likelihood of such bets. In the current study, this effect is quantified in order to measure directly the effect of insiders on the odds offered to gamblers without access to superior information. Unlike many earlier studies, we make a clear distinction between the actions of gamblers with inside information and those who process publicly available information to give objectively accurate estimates of the probabilities of different horses winning (i.e., expert bettors).

   The setting of odds by bookmakers on often-repeated sporting events provides an excellent opportunity for economics to investigate the pricing by a market maker of a financial asset with an uncertain outcome in the face of the possibility of insider knowledge. The analysis presented here allows the extent of the impact of insider trading on asset prices to be measured more directly than did previous models. This direct impact is different from the measurements of, for example, Shin (1991, 1992, 1993), whose studies indicate the frequently observed tendency for the odds on outsiders to be of poorer value than those for favorites. Shin suggests that the so-called favorite-longshot bias is determined only by insiders' gambling. In contrast, our analysis shows that insider models equally well describe the outcomes following from the behavior of well-informed gamblers with no access to privileged information. Thus, the empirical investigation of this article employs a data set in which the impact of expert opinion is unlikely to be marked and that of insider trading to be of importance. The need for regulation of insider gambling is examined in the light of our model and the quantification of its effect.

   The article comprises four additional sections. Section 2 explains the gambling on horse racing in Britain and reviews the literature on the setting of odds in the face of insider trading. Section 3 develops a new model for how bookmakers who maximize expected profits set odds in the face of insider and expert gambling. Section 4 tests the predictions of the model against extensive U.K. data sets of 2-year-old horses that have never raced before compared to those that have raced previously. The final section gives a conclusion and analyzes the need for regulation of insider gambling and trading in light of the theoretical and empirical evidence presented here.

2. Insider Information

   Before considering the theory and empirical investigation of the effect of the use of insider information on the setting of bookmaker odds on horse races, it is necessary to consider the arrangements for betting on races in Britain. Gamblers can bet on horse racing with bookmakers at the racecourse, off course, online, and on the telephone. It is also possible to bet on the pari-mutuel (known as the Tote) and on internet betting exchanges; the latter are internet-based technological platforms facilitating double-auction person-to-person markets based on sporting events such as horse races.

   Unlike the Tote and other pari-mutuel systems, bookmakers do not necessarily set odds proportional to stakes bet on the various outcomes of a race. They may, for example, take a view on a race and set odds that do not balance their liabilities. Insider trading activity may expose them to a loss-making position if a particular horse wins. In fact, a balanced book, in which the same amount is earned regardless of which horse wins, is a rarity in British bookmaking.

   U.K. bookmakers offer odds capturing the terms of the wager on an event, usually expressed on a fractional basis, the language of which requires explanation. An outcome for which the odds are 4 to I, or "four to one against winning," for example, indicates an implied probability of 1 in 5, or 0.2. Conversely, odds of 1 to 4, or "four to one on," represent a winning chance of 4 out of 5, or 0.8. The convention is that "a four to one shot" is implicitly understood in U.K. racing parlance to be "against," unless otherwise specified. This can be disconcerting for economists used to encountering odds in other contexts, in which odds of 4 to 1 would be routinely understood to be the chance of the event occurring (i.e., 0.8 or "four to one on" in betting market terms).

   The terms of bets to win (which are the focus of this article) placed with bookmakers can be agreed upon at the time of the transaction by requesting "fixed odds." If, for example, I wish to bet a horse to win, and the bookmaker's current advertised odds are 14/1, I can lock in this odds value, which will be honored by the bookmaker should the horse win, regardless of subsequent market fluctuations.

   In the absence of odds fixed at the time of the transaction, winning bets with bookmakers are settled at the starting price. The starting price, or SP, is a unique odds value for each horse determined by official on-course odds inspectors reflecting the most commonly occurring odds that bettors can obtain in the betting ring. SPs can therefore be seen as the final consensus of bookmaker odds at the time the race commences.

   Betting at the racecourse starts about 20 minutes before the race. A bet made during the live market in betting is by default struck at current odds offered by the bookmaker, although SP can be requested. It is commonly accepted that the initial odds that are offered by bookmakers reflect their evaluation of the past form of horses (see Royal Commission [1978, p. 468]). The weight of on-course betting alters these odds. Heavy support for a horse will shorten its odds, and lack of support will lengthen the odds. This tendency is particularly marked in the final 10 minutes before the race, as the rate of betting increases. Bookmakers may attribute strong support for a horse to insider information and expert analysis and may reduce the offered odds sharply. Until shortly before a race, off-course betting is made without knowledge of the starting prices that will be used to settle the bet. Approximately 15 minutes before the start of a race, the odds from the on-course market in betting are transmitted to off-course betting offices, and they become the basis for fixed odds betting. Off-course bookmakers can hedge their position and affect the on-course odds by acting as bettors and placing wagers to win with other bookmakers or, increasingly, on the betting exchanges. As betting at racecourses represents a small proportion of all betting on horse racing (see Mintel 2007), this can be an important effect. Racecourse bookmakers have to compete with the odds offered by online bookmakers and betting exchanges.

   We now consider the academic study of bookmaking on horse races. Given the explicit importance of probabilities in relation to odds in betting markets, it is useful to adopt a definition of bettor rationality suggested by All: "... bettors are rational in the sense that no one prefers a bet with a smaller winning probability and the same or lower return, or with a lower return and the same or lower winning probability, to that available to him." (Ali 1977, p. 809, footnote 9)

   Ali goes on to distinguish rationality from sophistication, asserting that "... bettors are sophisticated in the sense that the objective winning probabilities are known to them." (Ali 1977, p. 809, footnote 9)

   The existence of betting markets is paradoxical in that they yield, in aggregate, negative expected returns. This characteristic is inconsistent with rational behavior, as defined by Ali, in the sense that there exist markets for other financial assets that yield positive returns with ostensibly lower risk to capital.

   Further, many empirical studies of odds structures have observed a favorite-longshot bias (i.e., where, on average, the odds offered are such that the rate of return to betting on favorites is better than on longshots [in relation to bookmaker markets, see, for example, Dowie 1976; Crafts 1985; Shin 1991; Cain, Law, and Peel 2003]). While most studies have shown that negative returns were made across the probability range, these can be reduced by concentrating on favorites or near-favorites at bookmaker odds.

   If such odds structures reflect the revealed preferences of bettors, they clearly imply irrational behavior, as defined by Ali. Early studies of the favorite-longshot bias did, in fact, attribute the phenomenon to irrational bettor behavior, not least Ali himself (also see, for example, Snyder [1978], who surveys six early empirical studies of bias in pari-mutuel markets in which such behavior was assumed).

   In contrast, information-based models of odds bias have received much attention in recent empirical studies of betting markets. Hurley and McDonough (1995) explored the asymptotic behavior of "informed" bettors by assuming that, in contrast to the uninformed bettors, they know the true probabilities of horses...