The economic impact of the Olympic Games: evidence from stock markets

Abstract

By means of an event study of stock market reactions to the announcement of the Olympic Games host cities, we find a significant and positive announcement effect of hosting the Summer Games, with a cumulative abnormal return of about 2% within a few days. Yet, we find no significant results for the Winter Games and for losers. Our results differ from those of Mirman and Sharma (2008), who found that the Winter Games have a significantly negative announcement impact, whereas there is no effect for the Summer Games. Our results rely on a larger sample of 15 Olympic events and are obtained by assessing the abnormal returns after the announcement against a ‘business-as-usual’ situation (instead of testing the difference between winners and losers). Our findings are in line with economic intuition: the Summer Games represent a larger event and are thus more likely to make a significant impact. We also find that among the winners, small economies tend to exhibit greater cumulative abnormal returns than their larger peers.

Acknowledgements

We thank Waldemar Rotfuß, Andreas Schrimpf and Michael Schröder for their helpful comments. Frédéric Blaeschke and Balthasar Lackner provided excellent research assistance. Any errors remain the responsibility of the authors.

Notes

¹‘Winner’ refers to the country/the city that hosts the Olympic Games, whereas ‘loser’ refers to a bidding country that does not succeed with its candidacy.

²Their results for the individual events, however, substantially differ from Mirman and Sharma's (2008) findings and from the ones of our analysis.

³Mirman and Sharma (2008) proposed the inclusion of an AR(1) term and day-of-the-week dummy variables. The Durbin–Watson statistic indicates that the errors exhibit no high autocorrelation. Even if an AR term is included, we find the results for small event windows to be qualitatively similar. As event windows larger than [0,1] include different days, day-of-the-week dummies are redundant if the results for different event windows are in line.

⁴The considered indices are available from the authors upon request.

⁵See http://www.ers.usda.gov/Data/Macroeconomics/.

⁶The event window includes the day when the IOC announcement was first absorbed by the market and the subsequent n = τ₂−τ₁ trading days. As the IOC announcement is typically made in the evening (local time), we start the event window on the subsequent trading day unless time zone effects allowed the information to enter the stock exchange on the announcement day itself.

⁷For larger event windows up to [τ₁ = 0, τ₂ = 9], the impact continues to be positive on average. Beyond [τ₁ = 0, τ₂ = 5], however, the results are no longer significant.

⁸To measure which losing candidate was attributed the highest winning probability (ex ante), we use the ex post number of votes in the Committee as a proxy. Accordingly, the city with the second highest number is chosen as the ‘first loser’. Source: http://www.aldaver.com/votes.html.

⁹Ignoring the cross-correlation among losers and basing the results on the individual countries instead of on portfolios yields qualitatively similar results.