Does European regional competitiveness influence sports? An analysis of three sports

Abstract

The main objective of this work was to test European regional determinants of sports competitiveness. We studied competitiveness in volleyball, basketball and handball. We developed a theoretical model based on the model proposed by Bernard and Busse (2004) to analyse the effect of regional institutions and sport environments that can interfere with sports competitiveness. To test our model, we constructed an enlarged database for all European NUTS2 since 1995, which we analysed using panel data techniques (censored Tobit models). Our results reveal that the regions that are able to maintain player performance do so by persistent effort and external influences. These factors contribute more to regional competitiveness than a region’s central location or political prominence.

Keywords:

• location
• regional economics
• sports
• NUTS

JEL Classification:

• R12
• R39

Acknowledgements

The author acknowledges the two anonymous reviewers for their comments on a previous version of this article. The author also acknowledges the participants of the 17th APDR Congress for their suggestions on a previous version of this article. The author also thanks Joaquim Santos Teixeira for his work on collecting the data used in this article. Remaining errors are author’s exclusive ones.

Notes

¹ NUTS are European statistical descriptors of regional activity. NUTS1 describe a population between approximately 3 and 7 million people; NUTS2 represent a population between 800 thousand and 3 million people; and NUTS3 indicate a population between 150 thousand and 800 thousand people.

² This argument does not contradict our previous argument that a central location with only one modality can drive other modalities to the lower-ranked regions. Observe that when there is only a market for one modality at a central location, other modalities can be successful in the lower-ranked areas and can interact to generate mutually beneficial stimuli.

³ Following Bernard and Busse (2004) and Mourao (2010), we opted for the Cobb–Douglas specification because of its simplicity.

⁴ The rationale for the choice of a log function is that more talent leads to more competitiveness (for example, more talent leads to more points) but with diminishing returns. Therefore, we can state that and .

⁵ According to the model, the independent variables will characterize the economic, central and sports reality.

⁶ We could employ hurdle-type models to estimate specification (5). However, as cautioned by Kristrom (1997) or Clinch and Murphy (2001), hurdle-type models are more appropriate for micro-level data (and not for our macro-level data).

⁷ This index can be considered a notation or a metric choice.

⁸ The points assigned by this Index of Sports Competitiveness have strong correlations with the distribution of ‘prize moneys’ in the respective professional European championship: basketball 0.981, volleyball 0.972, and handball 0.941.

⁹ As a ranking system, this strategy may be subject to many sources of uncertainty, including incomplete information, biases, an incorrect form for aggregating the values or random errors in the direct measurement of the phenomenon (Nardo et al., 2005). We therefore used SimLab to perform an analysis of data sensitivity to validate the influence of sources of uncertainty. We considered the following sources of uncertainty: imputation, weighting schemes, aggregation level and the aggregation method. The input space was then resampled, and we obtained 20 000 combinations of the four previously independent input factors. Team rankings were re-computed for each trial sample. After performing all of the calculations, we observed high correlation between the median of the sample ranks (control ranks) and the sports rank. This robustness was also validated by the small SDs associated with each NUTS2 region relative to the sample ranks. Full results, tables and figures from the sensitivity analysis are available upon request.

¹⁰ To construct these Zipf curves, we used the values of the Index of Sports Competitiveness that will be described in the following sections. A higher Index value indicates that the relevant European NUTS2 region has teams that typically reach the play-offs of the European Team Championships.

¹¹ To test the (causal) effects of these variables on our dependent variables, we also run the estimations with the first lags of these RHS variables. Complete results are available upon request. However, after conducting the conventional statistical tests, we could not reject the hypothesis that the first lags of these RHS variables do not anticipate the dependent variables, favouring the interpretation that these RHS variables cause (anticipate) our dependent variables.

¹² In the cases in which we used the fifth lag, the estimations were from the year 2000 through the year 2009.

¹³ http://epp.eurostat.ec.europa.eu/portal/page/portal/region_cities/regional_statistics/data/main_tables

¹⁴ Based on the Durbin–Wu–Hausman test for the endogeneity of GDPPC, we concluded that this variable can be classified as statistically endogenous with the instruments ‘number of unemployed people’ and ‘Disposable income of private households’. The same conclusion was reached based on the Durbin–Wu–Hausman test for endogeneity of the ‘Population’ variable with the instruments ‘Number of physicians or doctors per capita’, ‘Students at the secondary level’ and ‘Number of households in a densely-populated area’. We also estimated the model as a panel data model with fixed regional effects; although following Hausman tests, we opted again for random-effects tobit models. All of these complete results (including the correlation matrices between the instruments and the residuals of the model) are available upon request.