Rule changes and competitive balance in Formula One motor racing

Abstract

This article provides an economic explanation of the frequent rule changes in the Formula One (F1) motor racing series. In a two-stage model, the FIA (the organizer of the F1) first decides whether to change the rules or not, and then the racing teams compete in a contest. It turns out that a rule change reduces the teams’ performances, but also improves competitive balance between the teams. The rule change is implemented, if the FIA's revenue gain from the latter effect overcompensates the FIA's revenue loss from the former effect. We provide empirical evidence from F1 seasons in the period 1950 to 2005, which supports the main implications of the model.

Acknowledgements

The present study was carried out while Camilla Mastromarco was Marie Curie Research Fellow at the University of Munich. We thank Ulrich Woitek and participants of the 20th Annual Congress of the European Economic Association in Amsterdam and seminars at the Universities of Leece and Munich for helpful comments. The usual disclaimer applies.

Notes

¹Day, J., ‘Formula One Teams to Fight Pay-TV Plans’, Media Guardian, February 22, 2002.

²This and the following information can be found on the website http://www.f1technical.net

³All the following information on the F1 series can be found at the website www.fia.com

⁴Another modelling of the success-orientated payments is to explicitly assume that the FIA divides the whole trophy money v into a first prize for the winner and second prize for the looser. See e.g. Moldovanu and Sela (2001) and Szymanski and Valletti (2005). We prefer the above modelling for several reasons. First, the FIA usually does not award differentiated prizes. Second, our empirically analysis focuses on the explanation of the ‘technical’ rule changes, not on the prize policy of the FIA. It is therefore suitable to model the FIA's prize setting behaviour as simply as possible. Finally, our approach approximates the division of the trophy money into a first and second prize and, thus, it seems reasonable that our basic insights will not be changed if several prizes are modelled explicitly.

⁵For simplicity, we ignore other important revenue sources of the teams like sponsoring and advertising. It is straightforward to include these sources in our model. But this would merely complicate the analysis without adding further insights.

⁶Two remarks are in order. First, from Equations 5 and 13 it is also straightforward to show that total/average effort of the teams is increasing in the rule change parameter, i.e. de*/dτ > 0. This insight confirms the result of Baik (1994, 2004) that higher the total effort, the more equal the players are. As explained in the ‘Introduction’, however, in our framework it is more suitable to focus on performance. Second, it is well-known that comparative static results in contest models may not be robust to variations in the functional forms and that some perverse effects may occur. See, for example, Nti (1997). Of course, this is also true for our model, but our empirical results will confirm the prediction of the theoretical model so that the problem is less severe.

⁷Such an argument assumes some kind of myopic behaviour of the FIA and the racing teams. A closer look at the rule changes in the past suggests that this assumption is not unrealistic. Modelling rule changes in dynamic sports contests are left for future research.

⁸The data on points have been collected from the official F1 website at http://www.formulaone.free-online. co.uk/index.html. The remaining data can be found on the websites at http://www.f1technical.-net, http://www.formula1.com, http://www.motorsm.com and http://www.atlasf1.com, which reports on official studies of the FIA.

⁹In the two-player contest, the SD of equals b*/2. The SD of winning percentage is a widely used measure of competitive balance in sports contests (Fort and Quirk, 1995; Szymanski, 2003). Other measures include the SD of winning percentage relative to an idealized SD (Scully, 1989; Quirk and Fort, 1992; Vrooman, 1995), the Gini coefficient (Quirk and Fort, 1992), relative entropy (Horowitz, 1997), the Hirschman–Herfindahl index (Depken, 2002) and the ratio of the sum of SD of team performance through time to the sum of within season SDs of win percentage (Humphreys, 2002).

¹⁰We also test for overdispersion in the Poisson regression model. The hypothesis that E(RC _t ) = Var(RC _t ) cannot be rejected on the basis of the likelihood ratio test.

¹¹Note also that the correlation between safety regulations and other regulations is low as the correlation coefficient between SRC _t and ORC _t amounts to 0.34.