Like the media belief in a lockdown baby boom (see here), there is a belief in the media that sports team performances affect fertility and birth rates (e.g. see here, or here). Most stories like 'Super Bowl babies' have been proven to be a myth. However, throwing more data at a question like this is often good. That's what Luca Fumarco (Masaryk University), and Francesco Principe (University of Padova) did in this new article published in the journal Economics Letters (ungated earlier version here). Specifically, Fumarco and Principe looked at how national football (soccer) team performances at international competitions (FIFA World Cup and UEFA European Football Championship) affected the number of births nine months later, for 50 European countries.
National team performance was measured using the weighting of each match used in FIFA's Elo rating system (more on that later). Births were monthly counts. Fumarco and Principe find that:
Across all of the specifications, we see that, on average, an increase in performance by one standard deviation is associated with a reduction in monthly births by 0.3% nine months after the event.
They also perform a robustness check looking at the effect on other numbers of months after the event, and find that:
The effect of performance on monthly births is statistically significant nine months after the tournament... while the effect after ten and eleven months is not significant....
And the results were also statistically insignificant for 1-8 months after the event. So, on the surface, this seems to support the idea of a 'baby slump' rather than a baby bump from better national team performance. Fumarco and Principe conclude that:
...an increase in national team performance in international football competitions is associated with a drop in births nine months after the event...
We hypothesize that these results might be explained by individuals’ time allocations choices... the attendance of live events (e.g., from late afternoon to late night, on TV, at the stadium, on big screens in public places...) may reduce the time spent on physical intimacy...
The mechanism they propose is speculative. However, there is good reason to doubt the headline results in this study. First, I'm not convinced that their measure of national team performance is valid. They claim to use national teams' performance "as measured by the ELO rating system", but clearly they do not. The Elo rating system that FIFA uses takes into account the strength of the opposition and goal difference (see here), neither of which make an appearance in Fumarco and Principe's measure. [*] Fumarco and Principe take into account only the weighting of the match, which increases as the tournament progresses. That is a fairly crude measure of team performance, and not a whole lot better than the number of matches played, or the number of matches won. It would be interesting to see how the results panned out simply using the number of games.
Second, on a related note, Fumarco and Principe appear to use the full time series of monthly births for each country in their analysis (~17,000 observations). However, the tournaments only happen every two years, and most teams don't play in every tournament (or even any tournament). In those cases, Fumarco and Principe set the team performance variable equal to zero, which is not so different from a team that lost all of its games (which would be assignment 3 points, as they assign a minimum of 1 point per game). Including a bunch of months where there is no tournament and every country has a zero for team performance will seriously skew the results. Now, Fumarco and Principe use a variety of fixed effects, including month fixed effects, and month x year fixed effects. That will reduce some of this problem, but won't eliminate it entirely. It would be interesting to instead see how robust the results were to including only the month that is nine months after each tournament (i.e. April of each year), and applying a difference-in-differences format using countries that did not participate in each tournament as controls.
Third, there is no control for population in their model. It should be obvious enough that the number of births depends on the number of women of childbearing age. So, by excluding population size from the model there is a serious omitted variable bias. They do include country fixed effects, but that will simply reduce the size of this bias, not eliminate it.
This is a study that started with an interesting research question, but I don't think we can really take their results as given (even notwithstanding that they are correlations rather than causal). This is the sort of research that a good student could easily follow up on and improve upon.
****
[*] A side note: For a number of years, I generated Elo-type ratings for a number of international sports, along with Super Rugby and the NFL (see here). So, I have a bit of experience with these systems.
No comments:
Post a Comment