to players in the control group.
Figure 8 illustrates that once we consider the average number of goals per minute
played, players in the control group even slightly improved relative to players in the
treatment group in the 06/07 season. In the 07/08 season, on the other hand, players of
Euro 2008 nationalities again improved substantially over the course of the season, while
the goal-shooting performance of players of other nationalities worsened.
It is useful to note that for some output measures, minutes played most notably
but also measures like passes received per minute, there is a substitution effect between
different players. Since the length of each match is limited and there is only one ball
in the game, not all players can simultaneously be on the field or play the ball. If the
average number of minutes played of a subset of players goes up, we would therefore
expect the average number of minutes played of the remaining players to decrease for
example.35 Indeed, the correlation between the minutes played observations of players in
the control group and those of players in the treatment group is about -0.55. For other
output measures, such as shots on goal per minute, the correlation is much closer to zero
but remains negative. We will come back to this when interpreting some of the regression
results.
In spite of the apparent trends in Figures 6 to 8, the regression results in Tables
5 and 6 provide only very weak support for the hypothesis that on average Euro Cup
qualification positively affected the club match outputs of players from qualified nations.
For minutes played, the estimated coefficient of qualif ied is positive but insignificant, as
reported column 1 of Table 5. For the various output per minute measures we employ,
Table 6 shows that again we find only weak support for an average effect of the Euro Cup
treatment. For shots on goal per minute played, the effect is insignificant in the sample
including observations for all field positions, but significant once only observations of
forward players are considered. In the latter case, the effect is about 0.00680 shots on
goal per minutes, which corresponds to a 21% increase in the number of shots on goal per
minute for forward players in the treatment group. We also do not find any significant
effects of qualif ied on the other reported output measures in regression including all field
positions.
35 One needs to be careful with too quick conclusions however. First, in the full sample the proportions of
players from the two different groups need not be constant over time. Second, if we keep the proportions
of player from different groups constant by excluding players who are active for a subperiod only (a
balanced panel), then minutes can be lost to or gained from players outside of the sample.
30