Abstract: The Demand for Pari-mutuel Wagering on NYRA Thoroughbred Horse Races
We develop multivariate time-series regression models with fixed effects to estimate the demand for win, place, and show (WPS), exacta, and trifecta wagering on 30,032 individual New York Racing Association Thoroughbred races contested at Aqueduct, Belmont Park, and Saratoga racetracks over the period January 2000 through December 2012. Seemingly unrelated regression (SUR) estimation methods which allow for correlation between disturbances to the three wagering pools produce substantially more accurate estimates than ordinary least squares (OLS) estimates that do not; interdependence among the three pools is important. The December 2010 closure of the New York Off Track Betting Corporation parlors is associated with average reductions of 7.4 and 4.2 percent in WPS and exacta pools respectively, but has no statistically significant effect for trifecta wagering. Ceteris paribus, races contested a) by the highest and the lowest quality participants, b) over shorter distances, c) on fast main dirt and on firm main turf courses, d) in clear and cloudy weather conditions, e) at Saratoga, f) by non-juvenile horses, g) without sex restrictions, h) after the first race on a card, i) in May and June, and j) before 2001 are associated with statistically significant increases in handle in all three wagering pools relative to races that are not. Betting is purse inelastic; a 1 percent increase in the purse of a race is associated with no more than a one fifth of 1 percent increase in any wagering pool. We find that 12.52, 11.62 and 10.88 betting interests maximize WPS, exacta, and trifecta wagering respectively. The optimal number of betting interests is independent of racing surface for WPS and trifecta wagering, while turf races are associated with a small but statistically significant increase in the exacta pool maximizing number of betting interests. The infrequency of takeout rate changes in the data means we can accurately estimate only short-run price effects for wagering, using regression discontinuity methods. We find that wagering demand is negatively related to own price, as expected, exacta and trifecta betting is own-price elastic in the short-run, and there are statistically and economically significant cross-price effects for betting in all three pools.