1
To What Extent Does Manager Nationality Influence
Performance Following Managerial Turnover, Using Football
Match Data from The English Premier League Between 2009 And
2019.
George Reynolds
11376015
Supervisor: mr. PR (Patrick) Stastra MSc
Bsc Economics and Business
Specialisation: Finance and Organisation
June 20, 2020
2
Statement of Originality
This document is written by George Reynolds who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
This research analysed the effect of managerial change on performance, focusing on the different effects between British and Non-British managers, hereby referred to as ‘Foreign’. Data from English Premier League football matches between 2009 and 2019 was
implemented for this study, instead of data from corporate firms. The frequent occurrence of manager turnover combined with highly visible performance results made this industry particularly suitable for manager turnover analysis. Comparisons between samples of British and Foreign managers were made using panel data containing match information of 10 games before the manager change, and 10 games after. The most important result of this research was that hiring Foreign managers had a significantly greater positive effect on points gained than British managers, in the short term, over the first 5 games of a new managers’ tenure. One reason for this trend may be due to greater fan support, as Foreign managers might come across as more exciting, flamboyant or even have an element of the unknown if they have not managed in England before. More fan support can reinforce the advantages of playing at home. This finding could be used to the advantage of firms who are seeking an immediate upturn in results. As for the overall effect of manager change on performance, a positive effect was found in this research. However, since only clubs who made manager changes were included in the sample, the absolute effect of manager change on performance may be overstated.
4 Table of Contents 1. Introduction ………... 5 2. Literature Review ……….. 6 3. Methodology ……….. 10 4. Results ………. 16 5. Discussion ………... 21 6. Conclusion ………. 23 7. References ………... 24
5
1. Introduction
A manager faces the difficult task of keeping a motivated, happy workforce, while also fostering a healthy competitive environment to get the best out of their subordinates. When a manager underperforms, pressure mounts from the board, who may make the decision to replace the manager. The occurrence of manager turnover is more common in some industries than others. For example, football clubs are renowned for frequently replacing their managers. Combining frequent manager turnover with easily observable performance indicators - match results - makes this a suitable environment to test the effects of manager change.
Previous research using football data has shown that following a managerial change, there is a small increase in the level of performance (Huson, Malatesta and Parrino (2004), Tena and Forrest (2007), Van Dalen (1994)). However, there is limited research into the impact that the nationality of a new manager has on performance. Besters, Ours and Tuijl (2016) investigated the effects of nationality on performance but found no significant effect. It will be interesting to research further whether following a managerial change, the nationality of new managers has an influence on the performance of the team. This leads on to the research question for this paper:
After the process of managerial turnover, do British football managers have a greater positive impact on performance than Foreign managers, using data from the English Premier League?
Using OLS regression analysis combined with various subsamples, the following outcomes were drawn using match data and manager characteristics from the English Premier League between 2009 and 2019. Firstly, it was found with this sample and method that changing manager leads to an improvement in performance. Secondly, there is presence of nationality difference between British and Foreign managers, in the short term.
The next chapter will proceed with an overview of the current research climate and previous findings. Chapter 3 will then explain in more detail the data and methodological approach of this research. Following that will be a thorough explanation of the results, which will then be discussed in more detail in Chapter 5, before concluding the paper.
6
2. Literature Review
Managerial turnover can occur for a variety of reasons. First of all, a manager may choose to retire at a time suitable to them, or that choice may be taken out of their hands in the
unfortunate case of death of the manager. Alternatively, managers may be fired due to their performance not being adequate enough, leading to a forced change from the board level in a bid to improve performance. In some situations, a manager may be fired in order to appease shareholders or stakeholders. This is known as ‘‘offering a scapegoat’’ (Tena & Forrest, 2007, p. 364). This is more likely to occur if the support of stakeholders has an influence on
performance results.
Previous research has been carried out to determine to what extent managerial change impacts performance, however there is a noticeable distinction between the types of data used. The first data type focuses on financial performance measures, such as accounting figures and stock prices. The second type of data is based around sporting data, which in this research will be focused on match results from association football.
This literature review will start by explaining the findings which use data available from conventional companies. Next there will be an explanation of the reasoning behind the use of football data, and how managers in ‘normal’ businesses can be compared to football managers. Following this will be a summary of the researches using football data, before discussing theories that are found through previous research, notably the impact of home advantage, and the important role played by fans.
Huson et al. (2004) found a positive impact of managerial change on performance using accounting data as performance measures. They explain that manager turnover is non-random, since poor performance is a predictor for a change in manager. Additionally, it is found that the announcement of a managerial turnover - through the study of abnormal stock returns - sends a positive signal to stakeholders. Denis and Denis (1995) used operating income as a performance measure, with a sample of 908 corporate managerial turnovers between 1985 and 1988, to determine the impact of managerial turnover on performance. Their findings show that managerial change prompted improvements in firm performance, so long as a forced resignation occurred.
Contrary to the findings of Denis and Denis (1995) and Huson et al. (2004), a negative effect was found from manager succession by Shen and Cannella (2002). This was found from a sample of 228 manager successions in large US public corporations, using firm ROA
7
as a means of representing operational performance. A negative effect was unexpected, since it was anticipated managerial succession would rejuvenate a business by means of strategic replacement of inefficient incumbents.
Using stock returns, Cools and Van Praag (2003) find a 5% positive abnormal return when a managerial dismissal is announced at the same time as an appointment of a new manager. This suggests that a change of manager is viewed positively by investors, in a similar way to Huson et al. (2004), although this research does not shed light on post announcement performance improvements.
Managers of football teams and managers in corporate business have similarities: “He undertakes a number of strategic and operative decisions which affect team performance”, De Paola and Scoppa, 2012, p.153. Weel (2011) and Koning (2003) state that football
performance data is accessible on a frequent basis, through weekly match results. With conventional firms, performance data is available quarterly or yearly in the form of financial reports or accounting figures. Weel (2011) also claims that football data provides clearer quantitative measures for results. A team can either win, draw or lose a match. Additionally, he believes the format of a football league allows fair comparison between the clubs, or firms, in the competition.
Koning (2003) uses match data from the Dutch first division from the 1993/4 season until the 1997/8 season, and focuses only on mid-season managerial changes, ensuring the quality of the teams remain as constant as possible. This enables the bulkhead of performance change to be explained by the change in manager. In order to avoid sample selection, he controls for the quality of opposition with a ranking model, and also controls for home advantage. However, the findings do not suggest that managerial change leads to improved performance.
Also using data from the Dutch first division, Van Dalen (1994) found an
improvement in goal difference for teams who changed their manager, by incorporating a
dummy variable for in-season manager change into the analysis.However, just one season of
match results were analysed, impacting the reliability of the findings.Weel (2011) used data
over a period of 18 years between 1986 and 2004 of the first division in The Netherlands. However, there was no significant change in performance found post managerial turnover, after implementing difference-in-difference regressions and 2SLS estimates.
8
De Paola and Scoppa (2012) use data from Serie A, the top league in Italy, over the period 1997/8 until 2008/9. The results find no significant effect on performance after managerial turnover. In the same way as Koning (2003), they only look at mid-season
managerial changes. In order to correct for selection bias based on the variability in quality of opposition teams, and the fact that a new manager will not play the exact same matches as the previous manager, Probit and Poisson estimators are used by De Paola and Scoppa (2012). Additionally, a matching estimator compared teams who have made a managerial change with those who did not. The Probit model resulted in no significant effect of manager change on performance, whereas the Poisson estimator found only a slight improvement in the number of goals scored after a manager change. As for the matching estimator, no significant effect was found.
Tena and Forrest (2007) find improvements in results only for home matches using data from the first division of the Spanish Football League (La Liga) for three seasons, 2002/3 to 2004/5 inclusive. They focus on the impact of whether the match is played at home or away as well as travelling distance for away games, by measuring the distance between the two stadia of the teams involved.
Using a sample similar to the one used in this paper, Besters et al. (2016) analysed data from the English Premier League between 2000/1 and 2014/15, finding inconclusive results about the effect of managerial change on performance. After using control groups to analyse the possible effects when managerial turnover did not occur, Besters et al., 2016 claim that there is no significant effect on the team after managerial change. Instead, they propose that the situations are specific and vary from team to team, presenting this through the use of case studies.
Tena and Forrest (2007) find improvements of results exclusively in home matches and Koning (2003) found that home advantage increased after manager change in the 1993/4 and 1996/7 seasons. Furthermore, Tena and Forrest (2007) discuss the effect of home crowds on referee decisions and find evidence that supports home bias of referees from studies by Nevill, Balmer and Williams (2002) and Dohen (2005). The presence of a teams’ home fans in the stadium can have a significant impact on the match outcome. At the time of writing, following the first few rounds of fixtures since the restart of the Bundesliga (German first division), it has been found that the number of wins by home teams is far below average (Da Silva, 2020). Of the 33 games played, the win percentage for home teams is just 21%, whereas the average for the season before the Corona-related break was 40%. All games are
9
now Geister-Spiele (ghost-games), without any fans, for obvious safety reasons. Hence, although a team may be playing at home, without fans, home advantage is not so prevalent.
Tena and Forrest (2007) argue that fans can have an influence on whether a manager gets fired. If fans have a disliking for the current manager, then the directors may make the decision to fire him in order to please their fans, rather than necessarily attempting to improve performance. The process of offering a ‘scapegoat’ in the form of a managerial sacking may “rekindle enthusiasm of a crowd” (Tena & Forrest, 2007, p. 364). Furthermore, a combination of media and fan pressure may push directors into firing a manager, without considering the good work and knowledge they have developed with the squad (Höffler & Sliwka, 2003). From a non-football perspective, Huson et al. (2004) find that turnover announcements are viewed positively by investors following stock-price analysis. It seems that on the whole, people are encouraged by managerial change.
Looking at other potential influences, an analysis of the effect of the type of manager turnover is studied by Denis and Denis (1995). They find that involuntary manager turnover is followed by much greater performance improvement than with voluntary turnover cases. Furthermore, the majority of previous researches do not analyse the impact of characteristics of new managers on firm performance. Besters et al. (2016) briefly compared subsamples of managers with varying characteristics: whether they were British, over or under 50 years old, or had represented their country as a player (capped). They found no significant differences between any of the subsamples.
Although no significant difference between British and Foreign managers was found by Besters et al. (2016), there remain profound differences between British and Foreign managers. It is assumed that foreign managers have lower English proficiency than British managers. When investigating the effects of language barriers on workers within
multinational teams, Tenzer and Pudelko (2015) found that anxiety can arise for non-native speakers of the common spoken language in a workplace. However, it was also found that test subjects feel more comfortable in situations where colleagues have similar levels of
“linguistic difficulties” (p. 613). Hence, coaching a squad with a greater majority of British players may lead to a foreign manager feeling more anxious than if he were to coach a squad with more foreign players, where linguistic difficulties are more prevalent. Moreover, by applying the similarity attraction paradigm discussed by Thomas and Ravlin (1995), working with a higher proportion of players who are from a similar cultural background will make the process of gaining trust and respect faster and more straightforward.
10
Since there is not a large pool of research studying the effects of managerial change for different nationalities, it is expected that there will be a similar result in this research to that of Besters et al. (2016). Hence, it is predicted that there will be no significant difference in performance between Foreign and British managers following managerial change. The following research methodology will attempt to disprove this. As for an overall effect, not contingent on nationality differences, a small increase in performance following managerial change is expected, in line with the combination of business and football data-based research.
3. Methodology
Panel data containing match information for goals scored, goals conceded, and match venue is available for all fixtures of the English Premier League from 2009 until 2019, from the online database Football Data (17 April 2020). For these years there are sufficient examples (63) of mid-season manager turnover to provide an adequate sample size. Data about the managers’ characteristics regarding nationality, age, starting date and playing history was gathered via the Premier League website (visited: 10 April 2020) as well as the following websites: worldfootball.net (visited 10 April) and transfermarkt.com (visited 10 April). Data regarding the nationality composition of the squads is also obtained from these sources, which will be a valuable tool in assessing the effectiveness of British managers on team performance in further analyses.
In this paper, I will follow the method of Koning (2003) and De Paola and Scoppa (2012), by only considering mid-season managerial changes. Managerial contracts most commonly expire at the end of the season. Hence, mid-season changes will only occur if the board decides to take drastic measures by sacking the current manager in a bid to improve results, appease fans or shareholders, or if the manager is given an opportunity to move to a
better job.Furthermore, although there is potential for teams to strengthen in the January
transfer window (a period where teams are allowed to add new players to their squad), squad turnover is much less significant than in the busier summer transfer window. Hence, it is assumed that the composition of squads, and subsequently team quality, does not significantly change during the season.
For similar reasons to Audas (1997), the inclusion of differentiation between voluntary and in-voluntary turnover will be foregone. Establishing the circumstances of why managerial change occurred in the first place involves scorning news sources and providing in some cases a subjective opinion on the sacking type, since the exact reasons may be unclear. In this paper,
11
an analysis can be made of all managerial changes occurring during the season, in order to avoid sample selection between voluntary or involuntary managerial dismissals.
In order to analyse the effect of managerial change on firm performance, information from 10 games before and 10 games after the managerial change was assembled. Weel (2011) uses 4 games before and 4 games after a managerial change for his analysis. However, for the purposes of this research, including additional matches allowed for more comprehensive and potentially significant conclusions to be drawn. Each match was given a “MatchDay”, starting from 1 and finishing at 20, in order to unify the panel data. A dummy variable for manager change was then introduced, with the change in manager occurring on MatchDay 11 for all cases in the sample.
I aim to delve deeper into the impact of manager characteristics on performance following manager turnover, using a similar method to the one used by Besters et al. (2016). Subsamples of British and Foreign managers will be analysed to see if there are any
differences between nationalities. The key independent variable for my analysis will be the dummy variable ManagerChange, which shows what impact changing manager has on performance.
Previous research has included variables to consider the previous playing experience of the new manager. Besters et al. (2016) incorporated a dummy variable for whether the manager has at least one international cap (appearance) representing his country. Weel (2011) added a variable to consider what type of player the manager was in his playing days, in particular whether they are an attacking player or not. Managers who were formerly attackers may adopt a more attacking style of play, leading to more goals scored. It is assumed that managers with playing careers will find it easier to gain the respect of the players than non-ex-pros. In this paper, dummy variables will be included for: Ex-player, Capped and Attacker. Other variables include: ManagerAge (the age of the manager at the start of the new
managerial position) and, importantly for further regression analysis, the percentage of British players in the squad (PropBritishinSquad).
The issue with comparing teams’ performance before and after managerial change is that the two managers face very different circumstances. For starters, each team in a league plays the other teams once at home, and once away. Therefore, a new manager won’t play the same opposition at the same time or at the same venue as his predecessor. Managers are not sacked randomly, they are fired due to a combination of poor performance, or perhaps an
12
accumulation of bad luck (Koning, 2003). The order of the fixtures may well have been a contributing factor as to why the previous manager has been replaced. De Paola and Scoppa (2012, p.154) explain the issue of “Ashenfelter dip”, which is the theory that over time, a series of bad results is equally likely to be followed by a series of good results. Another term for the Ashenfelter dip is regression to the mean, used by Weel (2011). If one were to
compare results without controlling for the different circumstances of teams, then selection bias would ensue. As can be seen in Figure 1, there is a distinct difference in performance before and after manager change occurs, at MatchDay 11.
Figure 1: An Illustration of the Ashenfelter Dip Theory
Using the average number of points gained in the previous 5 matches as a performance indicator, Figure 1 shows that team performance decreases leading up to a manager change at
MatchDay 11. Following the change, performance increases sharply in the first 5 games and
then levels off to a seemingly normal level, supporting the Ashenfelter dip theory.
In order to account for regression towards the mean, and to combat potential selection bias, two thorough control variables were created. The first control variable is inspired by Weel (2011). He incorporated a control variable consisting of a continuous point average of the last 4 games, to avoid potential fixture, team quality and form biases. In the same vein, a 5-game instead of 4-game average will be used in this paper (GPPG). This provides a less variable measure and is a more commonly used barometer by football statisticians to denote current form.
13
Table 1: A Summary of the Data Set
The next control variable was created based on the availability of betting prices. For each match played, data is available on the price for a home win, a draw, or an away win from the top 10 largest betting companies in the UK. A new probability value is calculated, to find the proportion of the total implied probability for each match outcome. This provides useable probability statistics to denote the likelihood of a home win, draw, or away win for each match. This additional control variable controls for various sources of noise, mainly the quality of opposition, but also home advantage and team form. Also, the venue of the match (Home as a dummy variable where home=1 and away=0) will be used as an additional control variable. Due to the high correlation with GPPG, the variables L5GS (goals scored in last 5 matches) and L5GC (goals conceded in last 5 matches) have been removed from the
following regressions to avoid multicollinearity in the results. See table 2 for the full coefficient correlation table.
Some data points were missing from the dataset, both at the start and end of the implemented panel. In order to determine the control variable of point per game in the
previous 5 games (GPPG), a manager needed to take charge 15 games into a current season to have a complete data entry. Additionally, he must have managed for at least 10 matches. Of the 63 manager changes included, 37 of them had complete data for all 20 match days.
Variable Obs Mean Std.Dev. Min Max
MatchDay 1260 10.5 5.769 1 20 ManagerCha~e 1260 .5 .5 0 1 Points 1165 .995 1.218 0 3 GPPG 1165 .996 .582 0 3 GoalsScored 1165 1.114 1.105 0 8 L5GS 1165 5.567 2.588 0 17 GoalsConce~d 1165 1.652 1.321 0 7 L5GC 1165 8.23 2.957 1 19 Home 1165 .483 .5 0 1 WProb 1165 .302 .159 .038 .786 DProb 1165 .257 .046 .088 .321 LProb 1165 .441 .182 .065 .87 Ageatstart 1196 50.956 8.073 33.726 69.162 British 1196 .589 .492 0 1 ExPlayer 1196 .917 .276 0 1 Capped 1196 .378 .485 0 1 Attacker 1196 .227 .419 0 1 Ageofsquad 1196 25.455 1.103 22.853 28.136 PropBritis~d 1196 .521 .136 .267 .788 ID 1260 32 18.191 1 63
14
Table 2: Pairwise Correlations Between Key Variables
Variables GoalsSc
ored GoalsConceded Points GPPG L5GS L5GC Home WProb ManagerChange GoalsScored 1.000 GoalsConceded -0.028 1.000 Points 0.608* -0.596* 1.000 GPPG 0.081* -0.068* 0.109* 1.000 L5GS 0.091* -0.087* 0.116* 0.677* 1.000 L5GC -0.016 0.059* -0.035 -0.594* -0.126* 1.000 Home 0.149* -0.151* 0.211* -0.065* -0.016 0.078* 1.000 WProb 0.288* -0.330* 0.386* 0.257* 0.264* -0.134* 0.413* 1.000 ManagerChange 0.068* -0.088* 0.120* 0.033 0.052 0.043 -0.015 -0.027 1.000 * shows significance at the .05 level
The dependent variable for this research is Performance, which is represented by three separate dependent variables: GoalsScored, GoalsConceded and Points, in a similar approach to De Paola and Scoppa (2012). Even though there is correlation between GoalsScored and
GoalsConceded with Points respectively (seen in Table 2), including Points as an additional
separate dependent variable enables more thorough analysis. For example, scoring more goals does not necessarily imply winning, so only utilising GoalsScored and GoalsConceded will lead to misleading results of performance.
Stage 1 of the regression analysis tests to what extent manager change influences performance. The statistical model for Stage 1 is:
(1): 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 = 𝛽0+ 𝛽1𝑀𝑎𝑛𝑎𝑔𝑒𝑟𝐶ℎ𝑎𝑛𝑔𝑒 + 𝛽2𝐺𝑃𝑃𝐺 + 𝛽3𝐻𝑜𝑚𝑒 + 𝛽4𝑊𝑃𝑟𝑜𝑏 + ∈ For Stage 2 of the regression analysis, the nationality of the managers is controlled for. Two samples are selected, British and Foreign. The coefficients between the two samples are then compared. For further time-specific analysis, there will be a comparison between the first
5 games in charge, and the next 5 games (the new manager’s 6th-10th games). This allows
comparison between short- and medium-term effects on team performance.
(2a):
𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 = 𝛽0𝐵+ 𝛽1𝐵𝑀𝑎𝑛𝑎𝑔𝑒𝑟𝐶ℎ𝑎𝑛𝑔𝑒 + 𝛽2𝐵𝐺𝑃𝑃𝐺 + 𝛽3𝐵𝐻𝑜𝑚𝑒 + 𝛽4𝐵𝑊𝑃𝑟𝑜𝑏 + ∈
(2b):
15
As a means of explaining any potential differences between the coefficients for British and Foreign managers, a third stage of analysis is included. Stage 3 will test to what extent the proportion of British players in a squad affects the performance of British and Foreign
managers respectively. Time-specific analyses will also be implemented. The regression model for Stage 3 is:
(3): 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 = 𝛽0+ 𝛽1𝑀𝑎𝑛𝑎𝑔𝑒𝑟𝐶ℎ𝑎𝑛𝑔𝑒 + 𝛽2𝐺𝑃𝑃𝐺 + 𝛽3𝐻𝑜𝑚𝑒 + 𝛽4𝑊𝑃𝑟𝑜𝑏 +
𝛽5 𝑃𝑟𝑜𝑝𝑜𝑟𝑡𝑖𝑜𝑛𝐵𝑟𝑖𝑡𝑖𝑠ℎ𝑖𝑛𝑆𝑞𝑢𝑎𝑑 + ∈
The hypotheses for this research are split in accordance to the three stages of analysis, each adding more value. For each stage there are three hypotheses, one for each of the three performance measures. Stage 1 analysis looks at the simple effect that manager change has on performance. Based on the review of previous literature, a small positive effect on
performance is expected. Hence, the statistical hypotheses for Stage 1 are:
Hypothesis 1.1 Hypothesis 1.2 Hypothesis 1.3
H0: β1GS = 0 H0: β1GC = 0 H0: β1P = 0
H1: β1GS > 0 H1: β1GC < 0 H1: β1P > 0
The Stage 1 hypotheses predict there to be a positive effect of manager change on all three performance measures. GoalsScored and Points are expected to be greater after manager change, and GoalsConceded is expected to be lower.
In Stage 2, nationality differences are introduced. In order to analyse whether there is a significant difference between the two nationality samples, coefficient equality tests are carried out. The following Z-test of Clogg, Petkova and Haritou (1995) will be used:
𝑍 = 𝛽1− 𝛽2
√(𝑆𝐸𝛽1)2+ (𝑆𝐸𝛽2)2
It should be reminded at this stage that there is no significant evidence in previous research of any nationality effects on firm performance following manager change. Hence, 2-sided tests will be utilised. The subsequent statistical hypotheses for Stage 2 are:
Hypothesis 2.1 Hypothesis 2.2 Hypothesis 2.3
H0: β1GSB = β1GSF H0: β1GCB = β1GCF H0: β1PB = β1PF
16
For the final stage of analysis, looking at the effect of squad nationality composition, the same Z-test will be utilised as in Stage 2. Based on theory of the effect of language barriers by Tenzer and Pudelko (2015) and the similarity attraction paradigm of Thomas and Ravlin (1995), the hypotheses for Stage 3 are:
Hypothesis 3.1 Hypothesis 3.2 Hypothesis 3.3
H0: β5GSB = β5GSF H0: β5GCB = β5GCF H0: β5PB = β5PF
H1: β5GSB > β5GSF H1: β5GCB < β5GCF H1: β5PB > β5PF
It is predicted that an increase in the proportion of British players in a squad will lead to greater performance benefits for British managers than for Foreign managers. In other words, British managers would perform better with more British players in their squad, and Foreign managers would perform better with more Foreign players in theirs.
4. Results
The Stage 1 regressions show that the variable ManagerChange has a positive effect on performance. The effect on the dependent variable GoalsScored when the independent variable ManagerChange equals 1 is positive (Z=2.69, p<0.01). Furthermore, following manager change, there is a negative effect on GoalsConceded (Z=-3.56, p<0.01) and a
positive effect on Points (Z=4.85, p<0.01). Hence, the null hypotheses for Stage 1 analysis are all rejected. These results support the previous research, indicating that there is a small
improvement in performance following managerial change. From this simple analysis, the absolute coefficient value of the independent variable ManagerChange is greater for the dependent variable GoalsConceded than GoalsScored. This indicates that new managers are place a focus on improving defensive aspects of performance, perhaps due to risk aversion tactics. The exact regression values can be seen in Table 3.
The aim of this research was to assess whether or not there are significant differences between British and Foreign managers in their ability to get the best performance out of a team, following managerial change. Stage 2 analysis tested (two-sided) for equality between independent ManagerChange coefficients, finding a difference of weak significance between British and Foreign managers with regards to the dependent variable Points (Z=-1.341, p=0.18). This weak significance means that the null hypothesis of H2.3 cannot be rejected.
17
From the sample studied, significant positive effects of the independent ManagerChange on dependent variable Points are found for both new British managers (Z=0.253, p<0.01) and new Foreign managers (Z=0.402, p<0.01), with Foreign managers having a larger effect than British managers. This difference could be explained by different management approaches between nationalities. Foreign managers may be likely to bring in more radical and
progressive changes, which can provide an impetus for the players to perform at a higher level. With regards to the two other performance measures GoalsScored and GoalsConceded, no significant differences between the two manager samples were found (Z=-0.597, p=0.55 and Z=0.551, p=0.58 respectively), meaning null hypotheses H2.1 and H2.2 are not rejected. Table 4 contains the complete Stage 2 regression results.
Table 3: Regression Results of Stage 1 Analysis
Dependent variable Goals Scored Goals Conceded Points
(1) (2) (3) ManagerChange 0.167** -0.259*** 0.317*** (0.007) (0.000) (0.000) GPPG 0.0222 0.0433 0.0351 (0.692) (0.513) (0.553) Home 0.0861 -0.0389 0.163* (0.214) (0.633) (0.026) Win Probability 1.880*** -2.749*** 2.737*** (0.000) (0.000) (0.000) Intercept 0.396*** 2.593*** -0.110 (0.000) (0.000) (0.217) No. Observations, n 1165 1165 1165 p-values in parentheses ="* p<0.05 ** p<0.01 *** p<0.001"
18
In order to analyse time sensitivity, the Stage 2 regressions were split into the first 5 games, and the subsequent 5 games of the new manager’s tenure. For the first 5 games, there was a significant difference in the independent variable ManagerChange between the two
nationality samples, with Points as the dependent variable (Z=-1.712, p=0.087). This enables
the null hypothesis for H2.3 to be rejected in the short term. However, when looking at the following 5 games in a new managers rein, there is no longer a significant difference between British and Foreign managers’ effects on Points (Z=-0.29, p=0.77). Foreign managers may be more likely to treat all squad players as equals at the start of their tenure, since they may not have seen them perform before. This may rejuvenate the squad in the short term, since there is now more fair competition to make the starting team. British managers on the other hand may have misconceived preconceptions about certain players based on previous encounters, limiting potential rejuvenation.
For Stage 3 of regression analysis, a new variable was added to assess whether the proportion of British players within a squad has a greater effect on performance for British managers than Foreign managers. Over the whole 10 game period, a difference was found between the two nationality samples, with GoalsConceded as the dependent variable.
However, this effect was weakly significant (Z=1.265, p=0.1029), meaning that H3.2 cannot be rejected. The direction of this difference, however, was not as expected. Foreign managers were found to concede less goals than British managers when the proportion of British players increased. Although this cannot confidently suggest that Foreign managers perform better with more British players in their squad, it suggests that Foreign managers may adopt more conservative tactics if they inherit a predominately British squad, potentially due to risk aversion.
A supplementary analysis was made to observe the differences between home and away fixtures, using home and away samples instead of home as a dummy control variable. These findings showed presence of home advantage effects for games which fell under the first 5 of a manager’s rein, for Points as a dependent variable. In the subsequent 5 games, the home advantage diminished on all performance measures. These results can be seen in Tables 5a and 5b. Adding nationality samples to this analysis, it was found that British managers outperform Foreign managers in away matches on dependent variables GoalsScored (Z=1.65, p<0.05) and Points (Z=2.42, p<0.01). This could be explained by British managers having more league-specific experience, making them better equipped to tactically negotiate difficult away matches.
20
In order to develop a deeper understanding of the influences of manager
characteristics other than nationality, further analyses were made. These included additional variables for the following manager and squad characteristics: manager age, ex-player, capped, attacker and the average age of the squad. All of these additional variables were found to have insignificant effects on the three dependent performance measures. This extra research was carried out to further categorise managers, to paint a picture as to what type of manager would be suitable under different circumstances.
Table 5a: The effect of ManagerChange on perfromance in home and away matches, for the first 5 games of a new manager.
Dependent variable Coefficient Equality Coefficient Equality Coefficient Equality
Home Away Home Away Home Away
(1) (2) (3) (4) (5) (6) ManagerChange 0.30** 0.147 1.13 -0.36** -0.151 -1.29 0.51*** 0.189 2.30 (0.0954) (0.0918) (0.113) (0.113) (0.102) (0.0968) GPPG -0.0286 0.0444 0.0439 0.0440 0.0246 0.0715 (0.0741) (0.0723) (0.0866) (0.0903) (0.0784) (0.0763) Win Probability 1.92*** 1.94*** -2.67*** -2.92*** 2.78*** 2.91*** (0.246) (0.255) (0.288) (0.316) (0.261) (0.268) Intercept 0.48*** 0.40*** 2.55*** 2.63*** -0.0341 -0.117 (0.101) (0.101) (0.117) (0.128) (0.106) (0.107) No. Observations, n 713 718 713 718 713 718
Standard errors in parentheses
="* p<0.05 ** p<0.01 *** p<0.001"
Points
Goals Scored Goals Conceded
Table 5b: The effect of ManagerChange on perfromance in home and away matches, for the subsequent 5 games of a new manager.
Dependent variable Coefficient Equality Coefficient Equality Coefficient Equality
Home Away Home Away Home Away
(1) (2) (3) (4) (5) (6) ManagerChange 0.115 0.111 0.03 -0.320 ** -0.29* -0.18 0.35** 0.25* 0.64 (0.100) (0.0956) (0.120) (0.117) (0.107) (0.101) GPPG 0.00794 -0.0300 0.107 0.0865 -0.0257 0.0120 (0.0719) (0.0726) (0.0867) (0.0898) (0.0764) (0.0741) Win Probability 1.67*** 1.73*** -2.656 *** -3.06*** 2.74*** 2.81*** (0.255) (0.263) (0.306) (0.324) (0.270) (0.271) Intercept 0.517*** 0.53*** 2.483 *** 2.63*** 0.0276 -0.0309 (0.101) (0.108) (0.123) (0.134) (0.107) (0.107) No. Observations, n 692 716 692 716 692 716
Standard errors in parentheses
="* p<0.05 ** p<0.01 *** p<0.001"
21
5. Discussion
The most important finding of this research is the difference found between new British and Foreign football managers on their team’s performance. This is the first research to find this effect using football match statistics. Besters et al. (2016) previously controlled for British subsamples but found no significant differences. Foreign managers had a greater positive effect on Points than British managers over the first 5 matches. This could be explained by greater home crowd support for more ‘exciting’ Foreign managers. However, over the subsequent 5 matches, British managers appear to catch up with regards to on-field
performance. This finding is similar in part to that of Weel (2011), who also found presence of a short term “shock effect” (p. 288), however this was found to only last for one game.
Similar to the results of Huson et al. (2004) and contrary to De Paola and Scoppa (2012), there was a significant positive effect of managerial change on all 3 performance measures, before the two nationality samples were introduced. After applying a
supplementary home vs away analysis, to replicate the findings of Tena and Forrest (2007), it was found that home advantage was more profound in matches that occurred during the first 5 games of a manager’s tenure.
Only the dependent variable Points was significantly different between the nationality samples, which may seem inconclusive. However, Points is arguably the most important of the 3 performance measures used. For example, scoring more goals could be attributed to merely a more reckless attacking style of play, which may lead to more goals conceded, and hence less points gained at the final whistle.
The presence of a new Foreign manager appointment signals a more immediate upturn in performance than British managers. There is potential for this finding to be used
strategically by a board when selecting a new manager, especially if there is a pressing need for immediate performance improvements. For example, in a battle against relegation, or when a company is nearing bankruptcy. Furthermore, boards should consider the
demographic composition of their squads (or employees) before deciding what sort of manager to appoint. Although the difference is weakly significant, it was found that as the proportion of British players in a squad increases, less goals are conceded for Foreign managers than British managers. This result was surprising, as it was expected British managers would work more effectively with squads containing more British players, according to the similarity attraction paradigm explained by Thomas and Ravlin (1995).
22
An important factor for deciding whether or not to sack a manager are the opinions of stakeholders, which in the situation of this research scenario is predominately fans. The significant difference in home game performance under the first 5 games may be partly attributed to increased crowd support. There pertains a feeling of excitement when a team hires a new manager, as it signals a fresh start and new ideas coming into the club, especially so with Foreign managers. Foreign managers may be seen as more flamboyant or mysterious, especially if they have not managed in England before, which breeds excitement. This
excitement may lead to more raucous home crowd support and hence higher performance levels of the players. In this research, the act of offering a “scapegoat” (Tena and Forrest, 2007, p. 364) does appear to have a positive effect. However, this wave of excitement, and associated home advantage, seems to quickly wear off, as in the subsequent 5 games there is no observable difference in the effects of home support. When nationality differences were introduced to deepen this finding, it was found that British managers prove to get better away results. One reason for this may be higher levels of experience in England, enabling British managers to navigate tricky away fixtures more effectively. Weel (2011) incorporates a variable for years’ experience in his methodology. Integrating this into this research with a league-specific experience variable may provide more insight to this finding.
This paper faces a similar limitation to that of Balduck et al. (2010), since only teams that changed their manager were included in the sample for this paper. Comparing these conclusions to other researches making use of all matches over their time period may be misleading, as significant results are more likely to be found. Furthermore, there is potential for a proportion of the significance in the findings of the basic Stage 1 regression to be attributed to regression towards the mean (Weel, 2011). Teams who changed their manager were likely to have experienced a series of bad results, perhaps due to an accumulation of bad luck (Koning, 2003). Hence, according to Weel’s regression towards the mean (2011), a series of poor results are likely to be followed by series of positive results.
By making the focal point of this research the assessment of differences between 2 nationality samples, instead of assessing the absolute effect of manager change on firm
performance, the extent to which regression towards the mean affects the conclusions of Stage 2 and 3 hypotheses is limited. Hence, the key finding of this research holds its significance, and adds a valuable new dimension to previous research. When making a managerial change, there should be careful consideration into the demographic of a candidate, and the potential ramifications that may ensue.
23
6. Conclusion
Using match results from 10 years of the English Premier League, Foreign managers were found to outperform British managers in the short term. If a company is looking for a short-term stimulus on results, hiring a Foreign manager can provide the required impetus.
Furthermore, taking the demographic of a workforce into account is also important, based on the trend found of Foreign managers performing better with squads containing more British players. Hiring a Foreign manager can shake up the status quo of a workforce, potentially leading to better performance. As for the overall effect of managerial change on performance, from the research methods utilised in this paper, managerial change has a positive effect on performance. Therefore, if things are going badly, changing a manager can be a viable strategy to improve the situation.
Although there are many similarities, applying the conclusions of this research to corporate management turnover should be done so with a pinch of salt. Managers of football teams should be used as a proxy for corporate managers. In order to keep the results as valid as possible, by only considering matches that remained in the current season for a new manager, a limited time frame of performance was analysable. In corporate situations, managerial turnover and performance data are much less frequent, so the timeline of analysis is usually over a longer time period. The long run effects of managerial change are difficult to assess using football match data. Each year, the league composition changes because of relegation and promotion. Also, turnover of players in the squads during the summer transfer window affects team quality.
In order to further develop the findings of this research, data to control for the situation of teams at the point of manager change could be implemented. For example, the effect of change on league position as used by Balduck et al. (2010), relegation, number of games remaining in the season, or time of year of the sacking could be utilised. Furthermore, assessing a more detailed nationality comparison, or perhaps considering regions or mother tongues may lead to more detailed understanding of manager differences. Splitting Britain up into its constituting countries and including more detailed squad nationality data could also provide further insight. Analysing a larger sample size by comparing similar sized leagues would be a welcomed future direction for further research, and potentially provide significant findings for the additional manager variables included in the supplementary analysis of this research.
24
Withstanding its limitations, the findings of this paper have implications for the selection criteria of managers. All managers from any nationality or race should be given an equal opportunity to be hired in a management position. However, looking at the correlations found in this research, there are potential performance advantages in choosing a Foreign manager if an immediate short-term effect is desired. Using these trends in a managerial decision-making process enables for a more informed managerial choice.
References
Audas, R., Dobson, S., & Goddard, J. (1997). Team performance and managerial change in the English Football League. Economic Affairs, 17(3), 30–36.
Besters, L., Ours, J., & Tuijl, M. (2016). Effectiveness of In-Season Manager Changes in English Premier League Football. De Economist, 164(3), 335–356.
Clogg, C. C., Petkova, E., & Haritou, A. (1995). Statistical methods for comparing regression coefficients between models. American Journal of Sociology, 100(5), 1261-1293 Cools, K., & Mirjam van Praag, C. (2007). The value relevance of top executive departures_
Evidence from the Netherlands. Journal of Corporate Finance, 13(5), 721–742. Da Silva, M. (2020, May 30). Bundesliga: 'Ghost games' have killed home advantage. DW.
Retrieved from https://www.dw.com/en/bundesliga-ghost-games-have-killed-home-advantage/a-53634123
Denis, D., Denis, D., 1995. Performance changes following top management dismissals. Journal of Finance 50 (4), 1029–1057.
De Dios Tena, J., & Forrest, D. (2007). Within-season dismissal of football coaches: Statistical analysis of causes and consequences. European Journal of Operational Research, 181(1), 362–373.
Dohnen, T., 2005. Social pressure influences the decisions of individuals: Evidence from the behavior of football referees. Discussion paper 1595, IZA, Bonn.
Football-Data. (2020). England (Season 2009/10-Season 2018/19) [Data file]. Retrieved from https://www.football-data.co.uk/englandm.php
25
Höffler, F., & Sliwka, D. (2003). Do new brooms sweep clean? When and why dismissing a manager increases the subordinates’ performance. European Economic Review, 47(5), 877–890.
Huson, M. R., Malatesta, P. H., & Parrino, R. (2004). Managerial succession and firm performance. Journal of Financial Economics, 74(1), 237–275.
Koning, R. (2003). An econometric evaluation of the effect of firing a coach on team performance. Applied Economics, 35(5), 555–564.
Nevill, A.M., Balmer, N.J., Williams, A.M., 2002. The influence of crowd noise and
experience upon refereeing decisions in association football. Psychology of Sport and Exercise 3, 261–272
Paola, M., & Scoppa, V. (2012). The Effects of Managerial Turnover: Evidence from Coach Dismissals in Italian Soccer Teams. Journal of Sports Economics, 13(2), 152–168. Shen, W., & Cannella, A. (2002). Revisiting the Performance Consequences of CEO
Succession: The Impacts of Successor Type, Postsuccession Senior Executive Turnover, and Departing CEO Tenure. The Academy of Management Journal, 45(4), 717–733. Tenzer, H., & Pudelko, M. (2015). Leading across language barriers: Managing
language-induced emotions in multinational teams. The Leadership Quarterly, 26(4). The Premier League. (2020). Managers (Season 2009/10-Season 2018/19) [Data file].
Retrieved from https://www.premierleague.com/managers
Thomas, D. C., & Ravlin, E. C. (1995). Responses of employees to cultural adaptation by a foreign manager. Journal of Applied Psychology, 80(1), 133–146.
Transfermarkt. (2020). Manager Information [Data file]. Retrieved from https://www.transfermarkt.com/
Van Dalen, H. P. (1994). Loont Het om een Voetbaltrainer te Ontslaan. Economisch Statistische Berichten, 79(3987), 1089–1092.
Weel, B. (2011). Does Manager Turnover Improve Firm Performance? Evidence from Dutch Soccer, 1986–2004. De Economist, 159(3), 279–303.
Worldfootball. (2020). History (Eng-premier-league) [Data file]. Retrieved from https://www.worldfootball.net/history/eng-premier-league/
