Testing the uncertainty outcome hypothesis using data from second tier soccer in Ireland

ABSTRACT

This study estimates a demand relationship for second tier soccer in Ireland to test the uncertainty outcome hypothesis (UOH). Using data from three recent playing seasons, the UOH is found to be upheld. In addition, well-determined effects for fixture quality, recent team performance and travel distance between the competing teams’ stadia are also obtained.

KEYWORDS:

• Football
• spectator demand
• uncertainty outcome hypothesis

JEL CLASSIFICATION:

• D12
• L83

Acknowledgement

Veronica Masubo is thanked for providing research assistance. However, the authors remain entirely responsible for any remaining errors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ In addition, Reilly (2015), using a quadratic parameterisation for the win probability, also reports evidence of a U-shaped relationship with log attendance minimized at a probability value of 0.57 and rising thereafter.

² In the 2014 season, four clubs had one missing attendance figure each, while two others had two missing figures each. In the 2015 season, the two missing figures were for different clubs. Overall, there appears to be no systematic pattern to the absent attendance data.

³ The use of gambling odds to test the UOH is contingent on the gambling market being information efficient. In order to investigate this, we used a three-category ordered logit model to regress the actual result on the log of the implicit forecasted probability over one minus this probability. The specification also included an array of publicly available fixture characteristics prior to the match including the league positions of the home and visiting sides, team dummies for the home and visiting sides, distance between stadia, the number of home team wins in the last three fixtures, and the day of the fixture. Although the coefficient on the log odds variable was well determined, none of the other estimated effects were found to be statistically significant at a 0.05 level or lower (either individually or jointly). We thus feel the assumption that this football gambling market is information efficient has some empirical support. The results of this exercise are available from the authors on request.

⁴ There is little or no variation in ticket prices either across clubs or over the three years covered by our data. In addition, income variation across the geographical regions within which the clubs are located is limited. Therefore, neither price nor income effects can be identified. These omitted factors are thus assumed to be absorbed within the club-specific fixed effects.

⁵ The null hypothesis of homoscedasticity was tested using the Koenker test and found to be upheld. The chi-squared value with one degree of freedom was computed as 0.07 with a p-value of 0.793. In addition, we choose not to cluster the variance–covariance matrix by team given the small number of teams present in the panel.