ABSTRACT
Research question
In recent years, European soccer leagues have strengthened their internationalization efforts by relocating games abroad. In this context, we empirically investigate the impact of the political relations between the importing and the exporting countries on on-site demand for such games. While there is a rich literature addressing this issue in more general settings, it has never been explored in sports.
Research methods
We use a stated-preference approach and survey data of a representative sample of soccer-interested US residents, which we combine with data from the American Community Survey. The effects are estimated using Tobit regressions and an instrumental variable approach.
Results and findings
We find that the willingness-to-pay for admission to relocated games is higher if the country where the competing teams come from is perceived as friendly.
Implications
We provide preliminary empirical evidence that the political relations between importing and exporting countries may be important for international demand of spectator sports. As such, the public perception about political relations between countries may constitute a crucial component for sport leagues’ expansion strategies.
Acknowledgements
Previous versions of the paper were presented at the 24th Annual Meeting of the German Association of Sport Economics and Sport Management in 2021; the 2nd North American Association of Sports Economists (NAASE) Virtual Conference in 2021; the 12th Conference of the European Sport Economics Association in 2021. We thank NAASE discussant Brad Humphreys and conference participants for their valuable comments and suggestions. The usual disclaimer applies.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 For a media discussion on relocating games of La Liga to the US see: https://www.theguardian.com/football/2018/aug/16/la-liga-stage-match-us. For Serie A discussions to relocate games to China see: https://www.reuters.com/article/uk-soccer-italy-china/italy-may-discuss-staging-serie-a-match-in-china-idUKKCN1R320B. For Premier League discussions adding a 39th game in the calendar to be played abroad see: https://www.dailymail.co.uk/sport/football/article-5207145/Manchester-City-plans-39th-Premier-League-game.html
2 Product ethnicity describes the stereotypical associations of a product with particular countries: for example, red wine is typically associated with Italy and France (Usunier & Cestre, Citation2007).
3 The survey was administered by the globally operating market research company Questback. The online panel was provided by Lightspeed GMI.
4 Since the Italian cup final (Coppa Italia) was rescheduled on short notice and staged during our survey period, it was excluded on purpose from the cup finals sample ex post. To check whether the absence of that game and country affects the findings in any form, (i) we also excluded the corresponding Italian Serie A game from the league games sample. Likewise, we checked, (ii) whether including the Italian cup final in the cup finals sample affects our main findings. Since our main conclusions hold for both re-estimations, we are not concerned about any bias due to omitting the Italian cup final in our main specification (results are available upon request).
5 See Nalbantis and Pawlowski (Citation2016) for a detailed description of the survey design, data quality, and representativeness of the samples (e.g. geographical distribution). In total, 6590 persons were invited to participate in the first survey and the response rate was 47%. In total, 5805 persons were invited for the second survey and the response rate was 54%. In total, 44% of respondents from the second survey also participated in the first survey. Moreover, our survey includes US residents, i.e., both US and non-US citizens. Excluding those with no US citizenship (<4% of the sample) does not affect our findings (results are available upon request).
6 The value corresponding to the 99th percentile is assigned to all values above this threshold, which is the case for 98 and 131 observations in the cup finals and league games samples, respectively. This approach (also called winsorizing) reduces the impact of outliers, while keeping all observations in the analysis (Ghosh & Vogt, Citation2012). We prefer this method as larger stated values usually contain more hypothetical bias (Murphy et al., Citation2005). Outlier reduction via trimming produces similar results (results are available upon request).
7 The rather small variation of perceived friendliness between countries (see ) may suggest that our findings are only driven by the different perceptions of the UK. We therefore re-estimate our main models excluding the UK games (i.e., the FA Cup final and Premier League game). Our findings remain (results are available upon request).
8 To check whether our findings differ by geography, we run further models including state dummies. Our main findings remain. Moreover, state dummies are jointly statistically insignificant (results are available upon request).
9 The survey respondents were assigned a value that indicates the share of veterans in the county they live in according to their ZIP code statements. To identify the county for each ZIP code, we use data from https://www.unitedstateszipcodes.org/. In case ZIP codes cross county borders, the website reports the most common county considering the percentage of population of a ZIP code area in the county areas as well as the percentage of the county areas in the ZIP code.
10 A caveat of the CMP estimator is that the software package lacks common tests to further assess the instrument strength. We try to circumvent this limitation with the use of two-stage least squares (2SLS) estimations. While this approach yields less efficient estimates, as it does not consider the limited nature of our data, it can provide commonly accepted diagnostic tests, the results of which can be seen as indicative (Angrist & Pischke, Citation2009). We consulted the effective F-statistic and the weak-instrument-robust Anderson-Rubin confidence set, which suggests that when using a linear model, the IV is only valid for the cup finals estimations. Therefore, we are more cautious when interpreting results for the league games sample.
11 In case of the binary specifications, the CMP estimator estimates the first stage of the IV-Tobit models with a probit regression.