4 Empirical Strategy
Our empirical analysis investigates the impact of a nation’s qualification for the Euro
2008 on the club match outputs of players from that nation. Assuming for now that the
effect of the Euro treatment is a constant, denoted δ , the basic regression equation we
estimate is:
Ynt = Yi + at + δqualifiednt + X^β + ε⅛t, (8)
where Yint is the output of player i of nationality n on game day t. The treatment dummy
qualif iednt equals 1 if and only if nation n is qualified for the Euro 2008 at time t,
and Xint is a vector of covariates. None of the covariates depends directly on a player’s
nationality, so in fact Xint = Xit ; it includes a dummy variable indicating whether player
i’s club plays at home or as a visitor on day t, player i’s club on day t, player i’s opponent
club on day t, and player i’s position on game day t.24 The player fixed effects γi pick
up any (time-invariant) differences between players, both before and after the Euro Cup
qualifications decisions.25 The game day dummies αt pick up any changes in conditions
affecting all players in the German league. The coefficient of qualif iednt estimates the
average effect of Euro Cup qualification on the output in club matches of players from
the qualified nations.
The identifying assumption is that in the absence of the Euro Cup treatment, play-
ers from qualified countries and players from non-qualified countries would have evolved
similarly over time (given controls), i.e., that δ = 0 in that case. Since players in the
treatment and the control group work in the same environment and are subject to sim-
ilar incentive systems in the absence of international Cups, we think that there is little
reason to doubt this. A player’s eligibility for the Euro Cup treatment, i.e., his nation’s
participation in the qualifications, is determined exogenously by geography.26 Within the
24It is relatively common for players to move between midfield and forward positions, or between
defense and midfield positions, in different matches.
25 The reader may wonder why age is excluded from the list of covariates. This is because in order to
avoid that the age variable picks up a time trend, we use a time-invariant age variable (age on May 17,
2008). In unreported regressions without player fixed effects, we find an inverted U-shape relationship
between most output measures and age.
26In some cases players can legally change nationality. Formerly Brazilian player Deco’s adopted Por-
tuguese citizenship for example, mainly to participate in the Euro 2004 and World Cup 2006. Authorities
and the FIFA have a critical attitude to such steps, however, which are therefore very rare. Nationality
changes by players who grew up in a country other than their country of origin, such as Ghanaian-born
Gerald Asamoah who adopted German citizenship, are naturally more common.
21