In modern football, many teams obtaining disappointing results replace the manager in the middle of the season hoping to improve their performance.1 However, whereas the costs of this strategy are clear (particularly, paying a wage for the new manager, while the old manager is entitled to receive his wage), the benefits are less clear.
Difficulties of assessing the impact of manager turnover
Evaluating whether replacing the manager helps to improve team outcomes is very difficult because it is impossible to know what would have happened if the old manager had continued to lead the team. More specifically, two thorny econometric problems need to be addressed when trying to disentangle the effects of manager turnover on team performance. The problem is that manageres are not randomly fired both because weaker teams tend to replace the manager more frequently and since dismissals are typically decided after a number of repeated negative results. As a manager is dismissed after a series of losses, the probability of an improvement in team performance is large (the so-called “reversion to the mean”).2
Empirical investigations that do not take into account these aspects might erroneously conclude either that the manager change has negative effects if different teams are compared (the manager change is negatively correlated with team quality) or, alternatively, that the manager forced turnover leads to an improvement in team performance even if its actual effects are negligible (since team performance tend naturally to improve after a string of bad outcomes).
Probably also because of these econometric problems, the literature on the effects of manager dismissals shows mixed results. Some works find detrimental effects (e.g. Bruinshoofd and ter Weel, 2003; Audas, Goddard, and Dobson, 1997; Audas, Dobson, and Goddard, 2002), others find no effects (Koning, 2003) and some others provide evidence for performance increase after a dismissal (Dios Tena and Forrest, 2007).
Evidence from Serie A
In a recent paper (De Paola and Scoppa 2012), we have tried to provide some new evidence on manager dismissals considering the Italian “Serie A”, one of the most important championships in the world in terms of revenues produced, supporters’ interest, media coverage.
We use match-level data for the 12 seasons between 1997–1998 and 2008–2009. In our dataset on average 37% of teams have changed the manager during a given season (there were 7 replacements in each season, ranging from a minimum of 4 to a maximum of 9). A total of 21% of matches were played by teams managed by a new manager. Manager change is more frequent among low-performing teams (61% of the lowest placed 10 teams fired the manager) and rare among high-performing teams (only 16% of the first-ranked 10 teams replaced the manager).
We measure team performance using different indicators of the outcome obtained on the pitch: the number of points gained in each match by the team (Points), the number of Goals Scored, and the number of Goals Conceded.
Figure 1 represents the teams’ average performance in matches before and after the change. Based on this figure one might conclude that firing the manager of an underperforming team yields better results. However, the results obtained on the field are influenced by a series of factors, which have to be taken into account in order to have a reliable picture. As explained above, if teams are hit by a series of negative shocks they may recover simply as the result of ‘‘regression to the mean,’’ independently from their decision to dismiss or retain the manager.
Figure 1 Teams’ performance before and after change
In order to handle the problems that may lead to biased estimates of the effect of manager change, we undertake two alternative econometric strategies. Firstly, we estimate a number of regressions trying to take into account the ‘‘regression to the mean,’’ by controlling for team fixed effects, team quality indicators, and lagged match results. Secondly, we use a matching estimator to estimate the effect of changing the manager comparing teams that have changed the manager only with teams that have not dismissed the manager but have very similar observable characteristics.
Independently from the model used we find that changing the manager does not produce any statistically significant effect on team performance. We also evaluate the impact of manager turnover on the offensive (Goals Scored) and defensive capability (Goals Conceded) of the team. We find a weak positive effect only on the number of goals scored.
So why are managers replaced?
This finding confirms results obtained by some recent studies and suggests that the firing of the manager has to be explained in relation to other reasons rather than for the expected improvement in team performance. For example, team boards may overestimate their own ability to undertake optimal replacement decisions, or as suggested by the scapegoating theory, firing the manager may represent a convenient tool for owners in order to placate frustrated stakeholders and supporters and displace blame for the poor performance away from themselves.