Leader Dismissal or Continuity, President Longevity, Geographic Orientation of Owners and Team Performance: Insights from French Men’s Football, 1994–2016

Abstract

We investigated the impacts of president longevity and the geographic orientation of owners on team performance and on the effectiveness of dismissing the leader. In addition, we considered their impacts on the effectiveness of not dismissing the leader while the same organisation fires them at another time for a similar performance. We also tested the impact of dismissing the leader or not on performance. We explored the aforementioned risk-taking relationships in the first tier of French men’s football over the 1994–2016 period (n = 4918 observations). To do so, we used a counterfactual based on the evolution of the team position over the last three games leading to the leader change and estimate linear regression models with fixed team effects. Our findings show that performance improves either after a leader dismissal or not in the same situation, and both president longevity and the geographic orientation of owners impact the effectiveness of dismissing the leader or not. In particular, global- and local-oriented ownerships have a positive impact on the effectiveness of the decision to dismiss the leader or not compared to national-oriented ownership. Practical implications stem from the research, e.g., how organisations with national-oriented ownership can overcome their competitive disadvantage.

1. Introduction

Performance is a key subject tackled from many perspectives in management in general (Chau 2019), as well as in sport management in particular (Sotiriadou and De Bosscher 2018). Chau (2019) notes that it is popular to consider performance as a measured outcome or output rather than an intermediate variable. Consistent with this, the present paper looks at performance as a measured outcome, as defined later.

One of the many perspectives relevant to the study of performance relates to the effectiveness of leader dismissal or continuity. Indeed, this topic is a critical risk-taking issue for organisations, with important implications for financial, market and organisational performance (Adams and Jiang 2017; Bragaw and Misangyi 2017; Desai et al. 2016, 2018; Zeitoun and Pamini 2017). Some authors note that while the performance and actions of leaders (usually Chief Executive Officers, CEO) are often scrutinised, boards and owners may also be subject to criticism regarding the effectiveness of their decisions about leader dismissal or continuity (Desai et al. 2016; Elsaid et al. 2011; Hamori and Koyuncu 2015). Such effectiveness may depend on the experience of the board members, in particular the president (chairperson), and the geographic origin and/or orientation of the ownership. The impact on firm performance of director experience is academically well informed (Chen et al. 2017; Field and Mkrtchyan 2017; Kor and Sundaramurthy 2009; Kroll et al. 2008). The same applies to the impact of foreign ownership (Bürker et al. 2013; Carney et al. 2019) and the geographic orientation of owners (Asmussen and Goerzen 2007; Banalieva and Santoro 2009; Goerzen and Asmussen 2007). Nevertheless, there remains limited empirical research into the impacts of director experience and ownership origin or its geographic orientation on the effectiveness of decisions about leader dismissal or continuity.

In this article, we intend to fill the gap in the literature by examining such impacts. More specifically, we investigated the impacts of president longevity and the geographic orientation of owners on organisational performance and on the effectiveness of dismissing the leader. In addition, we considered their impacts on the effectiveness of not dismissing the leader while the same organisation fires them at another time for a similar performance. We also tested the impact of dismissing the leader or not on performance.

We explored the aforementioned relationships in the football industry, a context well established for examining risk-taking in leader dismissal or continuity and organisational performance (Desai et al. 2016, 2018). Its advantages for the present research are threefold. First, it enables the identification of a clear leader in a club—the manager or head coach, who is then considered as the leader in this paper. Secondly, it provides a clear indicator of organisational performance—the team performance on the pitch or sporting performance, which is the type of performance measured in this paper. Third and finally, it makes it possible to measure such team performance on a regular basis—the outcome at the end of each game played by the club (i.e., more or less every week), which is the way team performance is measured in this paper. We focused on the first tier of French men’s football over the 1994–2016 period. We suggest that the French football context is conducive for examining our ideas for two reasons.

First, French football is characterised by a variety of presidents in terms of longevity. Indeed, during the last season covered by the period studied (2015–2016), the late Louis Nicollin was in his 42nd season as president of Montpellier and Jean-Michel Aulas was in his 29th season as president of Lyon, whereas Vincent Labrune and Pierre-Marie Geronimi were only in their 5th season as presidents of Marseille and Bastia, respectively. These differing longevities may impact their knowledge and understanding of professional football, and more specifically their club, e.g., their culture (Ogbonna and Harris 2014). This may affect the effectiveness of their succession decisions.

Second, following the globalisation of the sources of finance in European football towards the end of the 20th century (Andreff and Staudohar 2000), some French clubs have started to incorporate global companies and/or foreign investors in their ownership. This is in contrast to their traditional model based on local ownership (Andreff and Staudohar 2000) and presidents (Schotté 2016), still present in most clubs, whereas some others have national (rather than local or global) owners and/or presidents.

The remainder of this study is organised into several sections. In the following section, we review the literature and offer our hypotheses. Subsequently, we describe the method and data. We then report the results of the study. Finally, we present our discussion, practical implications, limitations and ideas for future research.

2. Literature Review and Development of Hypotheses

2.1. Leader Dismissal or Continuity and Performance

The impact of leader dismissal on firm performance is well researched (Elsaid et al. 2011; Hamori and Koyuncu 2015; Shen and Cannella 2002; Zhang 2008). Some of the aspects investigated in the mainstream literature are relevant to the present research, such as the impact of a departing CEO tenure (Shen and Cannella 2002) and newly appointed CEO experience (Elsaid et al. 2011; Hamori and Koyuncu 2015; Zhang 2008). Shen and Cannella (2002) start their article by stating that in the decades preceding their publication, research on the performance consequences of CEO succession has been extensive but characterised by inconsistent findings and debates about causes and effects. In their own study, the authors find that successor type interacts with post-succession senior executive turnover to influence firm return on assets (ROA). They also show that there is an inverted U-shaped relationship between departing CEO tenure and post-succession firm ROA, with an inflection point reached when the departing CEO tenure is about 14 years.

Looking at the reasons why some newly appointed CEOs (i.e., those with tenure of three years and less) are dismissed while others are not, Zhang (2008) also addresses the benefits of prior CEOs for hiring organisations. She shows a negative correlation between prior CEO experience and post-succession firm financial performance. Addressing a similar topic, Elsaid et al. (2011) distinguish outside successors who have previous CEO experience (exCEO) and those who have not (non-exCEO). They find that the stock market reacts positively to the hiring of an outsider who is an exCEO. They also show that, compared with firms that hire non-exCEOs, those with exCEOs had higher debt ratios and greater bankruptcy chances pre-succession. However, post-succession, these firms still have worse financial performance. Extending Zhang (2008) and Elsaid et al. (2011); Hamori and Koyuncu (2015) look at the relationship between experience in the CEO position of a different firm and the post-succession financial performance of the firm that they currently lead. They find that experience in the CEO position is negatively related to firm performance.

In the football industry alone, we have identified 35 studies looking at the impact of an in-season leader (i.e., head coach) change on team (i.e., sporting) performance in 14 countries. A total of 32 previous studies in 12 countries were identified in a preliminary paper about France (Scelles and Llorca 2020). The latter study is one of the three additional studies, along with Rocaboy and Pavlik (2020), also about France, and Galdino et al. (2021) wrote about Brazil. A naive approach (NA) consists of comparing team performance before and after a leader change. An alternative approach consists of using a counterfactual or control group (CG) in addition to the treatment group, i.e., no leader change for the same team in another season despite a quite similar performance. A total of 20 studies did not incorporate such a control group, whereas 15 did. Interestingly, the use of a control group enables considering the impact of leader continuity on team performance, compared to a leader change.

Contrasted results were found in the 35 studies identified, including across studies about the same country. Twelve studies (six NA and six CG) found no significant impact/improvement of a leader change on team performance (ritual scapegoat theory). Eleven studies (six NA and five CG) found a significant negative impact (vicious circle theory). Twelve studies (eight NA and four CG) found a significant positive impact (common sense theory). Most of the 15 studies with a control group found a positive impact on team performance of both the treatment group and the control group. This may indicate that team performance improves after a bad run, independently of the change in the leader or not. However, this may also indicate that the decision maker formulates the right choice when dismissing the leader or not. Indeed, it may be the case that, if the leader would not have been dismissed when they had actually been dismissed, team performance would not have improved. Similarly, it may be the case that, if the leader would have been dismissed when they had actually not been dismissed, team performance would not have improved. These aspects are explored further later in the paper. For now, the hypotheses focus only on the impact of a leader dismissal or not on team performance. Hence:

Hypothesis 1a (H1a).

Team performance improves after a leader dismissal.

Hypothesis 1b (H1b).

Team performance improves when a leader is not dismissed while the same organisation has dismissed the leader in a similar situation at another time.

We focused on leader dismissal (or not in the same situation) rather than any leader change; therefore, we also had to control for leader quit, i.e., voluntary departure (or not in the same situation). In contrast to leader dismissal that can be planned by the organisation or at least decided when performance is poor, leader quit may be unpredictable and, as such, disturbing for the organisation. At the same time, a leader may quit because performance is poor and they anticipate their future dismissal. In other words, the disturbing effect of the leader quitting may be counterbalanced by the fact that performance was poor under their leadership. If the leader does not quit in a situation similar (in terms of performance) to when the leader has quitted the same organisation at another time, the disturbing effect of the leader quitting is avoided but performance may remain the same because leadership remains the same. Hence:

Hypothesis 1c (H1c).

Team performance does not change after a leader quit.

Hypothesis 1d (H1d).

Team performance does not change when a leader does not quit while a leader in the same organisation has quitted in a similar situation at another time.

2.2. Director Experience, Learning and Performance

The impact of director experience on firm performance is academically well informed (Chen et al. 2017; Field and Mkrtchyan 2017; Kor and Sundaramurthy 2009; Kroll et al. 2008). Interested in the context of market entry, Chen et al. (2017) note that a key variance in board director expertise involves how the directors acquired their expertise. They stress that the appointment of board directors constitutes an inflow of new market know-how into the focal board through two mechanisms: learning-by-doing and learning from others. Although the present research does not look at market entry, these two mechanisms remain relevant1.

Drawing from the organisational learning literature (Helfat and Peteraf 2003; Levitt and March 1988); Chen et al. (2017) mention that learning-by-doing of the board involves directors building knowledge, skills, and relational capital which are specialised to the focal firm and board context (Castanias and Helfat 2001). According to Chen et al. (2017), this firm-specific board experience enables directors to accumulate tacit knowledge about the firm’s strategy, its unique competencies and vulnerabilities, and the specific challenges that the firm faces in its environment. The authors add that, through this experience, directors also gain familiarity about the board members and key executives. Therefore, this experience captures the directorial knowledge of the unique interactions between each firm and its business domain. Based on these elements, it is expected that directors with more firm-specific board experience make more informed choices, with a positive impact on performance.

Nevertheless, learning-by-doing may be counterbalanced by learning from others. Chen et al. (2017) note that board-level learning from others occurs through appointing outside directors who transfer external market know-how and know-who embedded in other firms and industries. However, it can be argued that board-level learning from others also occurs for the outside directors who learn from board members and key executives already present in the firm prior to their appointment. In this case, learning-by-doing and learning from others are mixed together. This specific learning from others echoes the idea of outside directors’ firm-specific founding experience introduced by Kor and Sundaramurthy (2009). These authors describe it as a source of firm-specific human capital for outside board members through founders who remain on the board after they cease to be involved in the day-to-day operations of the company. This firm-specific experience developed prior to the appointment of the outside directors is actually not limited to founders, but characterises any board members and key executives already present in the firm prior to their appointment. If outside board members are able to learn quickly from those actors, this may counterbalance their lack of firm-specific board experience.

In the football industry in Europe, the key actor when it comes to decide whether to dismiss the leader (i.e., head coach) or not is the president (Kelly and Harris 2010). Based on the elements developed above, it is expected that presidents with more club-specific board experience make more informed choices, with a positive impact on team performance. Nevertheless, the role of football club president induces a strong media presence (Schotté 2016) and direct contact with a range of stakeholders (e.g., shareholders and fans). This is likely to generate much more pressure than simply being a board member. In other words, experience as football club president may be as relevant, if not more, as club-specific board experience. Based on this, it is expected that more experienced presidents make more informed choices and are more likely to enable their organisations to create the optimal environment for their leaders (Arnold et al. 2012). This is expected to have a positive impact on team performance. However, less experienced presidents may be able to overcome their lack of experience by listening to the advice provided by actors knowledgeable about the club. They may even take advantage of a greater openness to external viewpoints than more experienced presidents.

The elements developed here suggest that the impact of president longevity on team performance and, in particular, the effectiveness of the decision to dismiss the leader or not may depend on a trade-off between learning-by-doing and learning from others. Our hypotheses are derived from the idea of around 14 years being an inflexion point (Shen and Cannella 2002), as mentioned in the previous subsection. Although these data applied to CEOs rather than presidents in previous research, the reasons provided for CEOs may apply to presidents. Hence:

Hypothesis 2a (H2a).

President longevity has a positive impact on team performance until reaching an inflexion point around 14 years.

Hypothesis 2b (H2b).

President longevity has a positive impact on the effectiveness of the decision to dismiss the leader or not until reaching an inflexion point around 14 years.

2.3. Ownership Origin, Geographic Orientation and Performance

The impact of ownership origin and, in particular, foreign ownership on firm performance is well researched (Bürker et al. 2013; Carney et al. 2019). Referring to Dunning (1981) and Caves (2007), Bürker et al. (2013) note the advantages related to foreign ownership stressed by the internationalisation literature. These are linked to the fact that multinational firms possess sophisticated assets which domestic firms lack, including managerial expertise, process and production technologies or brand names. However, the authors also emphasise the existence of a well-developed management literature (Buckley and Strange 2011; Filatotchev and Wright 2011; Tomassen and Benito 2009), stressing the role of assimilation and governance costs incurred by multinational enterprises when operating in a foreign country. Consistent with the contrasting forces influencing the operation of foreign affiliates, their literature review highlights mixed evidence on the productivity consequences of foreign ownership.

In the football industry in Europe, Andreff and Staudohar (2000) note that throughout most of the 20th century, revenues came from local spectators, subsidisers (governments and/or industrial patrons) and sponsors. To characterise this, the authors talk about the Spectators–Subsidies–Sponsors–Local (SSSL) model of finance. Nevertheless, they identify a shift for most top-level European professional clubs during the 1980s and even more so in the 1990s. This shift led to a new model of finance based on four pillars (Media–Corporations–Merchandising–Markets) and globalised (MCMMG model of finance). This access to global sources of revenue provides top-level European professional clubs with a competitive advantage compared to other clubs (Scelles et al. 2017). This allows them to sustain better team (i.e., sporting) performance as evidenced by their regular presence at the top of their domestic leagues and in the main European club competition, i.e., the UEFA (Union of European Football Associations) Champions League (Scelles et al. 2020a).

Although we agree with a trend towards a move from the SSSL to the MCMMG model (e.g., in French football, see Scelles and Andreff 2017), we argue that there was/is rather the coexistence of both models in European men’s football (Andreff 2017). This is true even for top-level European professional clubs, as evidenced by the most recent economic information available (Deloitte 2021). As such, top-level European professional clubs rely both on local and global processes, consistent with the idea of a ‘glocalization’ of football (Giulianotti and Robertson 2004). Moreover, revenues come not only from the local or global level, but also from the national level, e.g., domestic broadcasters such as Sky UK for the English Premier League or Canal+ for the French Ligue 1 (Feuillet et al. 2019; Scelles et al. 2020b). In addition, a domestic owner can also be global if its activities are not confined to the domestic territory (e.g., the French film production and distribution company Pathé owning shares in Olympique Lyonnais/OL Groupe since 1999). Thus, the distinction made in the internationalisation literature between domestic and foreign ownership has to be qualified in the context of the football industry in Europe. Indeed, this is the geographic orientation of the organisations and their owners—rather than simply the geographic origin of the latter—and its implications for performance that matters (Asmussen and Goerzen 2007; Banalieva and Santoro 2009; Goerzen and Asmussen 2007). For example, focusing on emerging market multinational enterprises, Banalieva and Santoro (2009) find that a combination of local (country) and global orientations enhances their relative financial performance. In contrast, regional (proximate confines of the country) orientation reduces their relative financial performance.

For the development of our hypotheses, we follow Banalieva and Santoro (2009), although the definition of local is different (territory rather than country) and regional is replaced by national. We suggest that global-oriented ownership may lead to additional managerial expertise. This may translate in a positive impact on team (i.e., sporting) performance. At the same time, the importance of the local dimension in sectors where organisations are historically tied to a territory, as is the case in the European football industry, may lead to local-oriented ownership having a better knowledge and understanding of the organisational culture. This may translate into a positive impact on team performance. Eventually, if national-oriented ownership induces a more limited managerial expertise than global-oriented ownership and a more limited knowledge and understanding of the organisational culture than local-oriented ownership, this may translate into a negative impact on team performance. Hence:

Hypothesis 3a (H3a).

Global- and local-oriented ownership have a positive impact on team performance compared to national-oriented ownership.

Hypothesis 3b (H3b).

Global- and local-oriented ownership have a positive impact on the effectiveness of the decision to dismiss or not the leader compared to national-oriented ownership.

3. Method and Data

We tested our ideas with team performance data from games played in the French men’s football Ligue 1 over the 1994–2016 period (n = 4918 observations, out of the 7990 games played over the period). We focused on in-season leader dismissals and controlled for leader quits (i.e., voluntary departure). Our models required the use of a counterfactual for each leader change. Thus, we only used the games played by teams with a leader change in the seasons when the change occurred, as well as the games played by the same teams in the counterfactual seasons (i.e., without leader changes despite a similar level of performance as for the seasons with a leader change).

3.1. Counterfactual to Leader Dismissal and Quit

A dummy variable was used for the counterfactual to leader dismissal and quit. It was based on the evolution of the team position in the league table over the last three games leading to the leader change. We calculated the cumulative difference (CD) in the team positions over the same three matchdays between two seasons, one with a leader change after the third matchday and another without a leader change. For example, considering that the team was 15, 16 and 17 before the leader change (occurring after the 26th matchday) and 16, 17, 16 after the 24th 25th and 26th matchdays of a season without leader change. CD is the absolute value of the sum of the differences between both seasons after each of the three matchdays, i.e., CD = |(15 − 16) + (16 − 17) + (17 − 16)| = |−1| = 1. We made this calculation between the season with a leader change and several seasons without a leader change to identify the season without a leader change the closest to the season with a leader change. The criteria for decision were as follows:

-

Ee retained only the season(s) with the lowest CD(s), i.e., that/those equalled no more than 9 in absolute value. Among these seasons, we retained only that/those with a difference of no more than 6, 4 and 3 in absolute value for the third, second and last matchdays before the leader change, respectively2;

-

If we had more than one season after the first step, we then retained only the season(s) with a CD of no more than 3 in absolute value. If there were only CDs of more than 3 in absolute value, we retained only the season(s) with the lowest CDs;

-

If we had more than one season after the second step, we then observed whether these seasons had a similar evolution as the season with a leader change in terms of positions over the last three matchdays. We then chose the season without a leader change with the closest positions to the season with a leader change. The idea is that a quite similar CD between a season with a leader change and two different seasons without leader change may be associated with two types of evolution in positions for those two seasons without leader change. For example, such evolution in positions for the season without leader change can be quite similar to the season with a leader change (e.g., CD = 0 with 1, 0 and −1 for the third, second and last matchdays, respectively). However, the evolution in positions for the season without leader change can also be not as similar as the season with a leader change (e.g., CD = 0 with 3, 0 and −3);

-

If several seasons without a leader change met the expectations above, we chose the season without a leader change the nearest in time to the season with a leader change, ideally with the same leader. Another criterion was the same number of matchdays as the season with a leader change (34 matchdays over 1997–2002 vs. 38 matchdays over 1994–1997 and 2002–2016).

Sometimes none of the seasons of the same club had a CD sufficiently close to the season with a leader change (i.e., no more than 9 in absolute value). Sometimes the same team-season was used several times as counterfactual. In this case, only those matches for which the counterfactual dummy alternatively took the values of 1 (from the matchday just after a leader change in a season with a leader change) and 0 (before the matchday with a leader change in a season with a leader change) were retained several times. Sometimes it was not possible to find a counterfactual leader change that did not happen because the club with a leader change was present in Ligue 1 for just one season. This is consistent with the idea that younger firms are systematically exposed to higher risks of market exit in professional football (liability of newness in organisational ecology; Oberhofer et al. 2015)3.

It is worth noting that we could have relied on average points per game and their evolution instead of position. However, average points per game are impacted by the numbers of wins, losses and draws during a given season in football. This is due to a game with a win/loss allocating more points than a draw (3 vs. 2 points). Relying on average points per game would have made comparisons between seasons complicated. In addition, the same average points per game can correspond to a very different position, depending on the seasons and their competitive balance (e.g., when comparing a season with all teams being balanced vs. another season with a few teams far better than the others).

Our approach differs from Van Ours and Van Tuijl (2016); Besters et al. (2016); Scelles and Llorca (2020), who used cumulative surprise (CS). CS is based on “match surprise”, i.e., the difference between the actual and expected number of points for a match, based on the odds of the bookmakers (Stadtmann 2006). CS is the sum of all “match surprises” since the start of the season. Van Ours and Van Tuijl (2016, p. 596) note that “If this cumulative surprise sinks below a certain threshold, then continuation of the cooperation between club and head-coach might become doubtful.” We favoured a different approach in this paper due to betting odds being not available prior to the 2000–2001 season and our willingness to include seasons prior to 2000–2001. This is due to the Buffet law introducing a new ownership form in 1999, namely, the professional sport limited company (‘‘sociétés anonymes sportives professionnelles’’, SASP) form. This allowed a private partner to own the entire capital of a club and receive dividends (Scelles et al. 2018). All French football clubs gradually transitioned from voluntary associations to companies. This evolution impacted their ownership and management. These dimensions are paramount to the present research; therefore, we decided to include seasons prior to 2000–2001 in order to be able to discuss the impact of the change4.

A limitation related to our approach could be that expectations from one year to another are not automatically the same for a team. As such, it might be the case that two quite similar situations in terms of team position and its evolution are associated with two different levels of CS. For example, it could be associated with a negative level when the leader was dismissed (with the hope that the situation would improve following the dismissal). Conversely, it could be associated with a level close to 0 when a leader was not dismissed (with a better acceptance that the team position was representative of its sporting level that specific season). If the limitation underlined here was true, it would be expected that the counterfactuals with our approach would not be significant. Indeed, no improvement should have occurred after a leader was not dismissed. However, the counterfactuals with our approach had a significant positive impact on team performance, meaning that the latter improved after a leader was not dismissed.

3.2. Models and Variables

We estimated linear regression models explaining team performance at game level:

y_ijk = x_ijkβ + η_i + ε_ijk

(1)

where y_ijk represents the performance indicator of team I in game j of season k; x_ijk are potential determinants of the performance; β is the vector of parameter estimates; η_i is the team fixed effects used to account for the (unobserved) quality of a team; and ε_ijk is the error term.

Contrary to Van Ours and Van Tuijl (2016); Besters et al. (2016); Scelles and Llorca (2020), we did not use team-season fixed effects. They capture the impact of both leaders when a change occurred. As such, they set a benchmark for performance based on results achieved under both regimes. The weights varied when an in-season dismissal occurred. For example, a very early dismissal will mean that the new leader is essentially judged on the benchmark of their own performance. In the present research, the benchmark is set through the position of the team in the league table at the end of the previous season, as mentioned below.

Team performance was measured as the goal difference between both teams at the end of the game. Goal difference has been theoretically shown as the mathematically optimal measure of a football team performance (Heuer and Rubner 2009; Heuer et al. 2011). More specifically, Heuer and Rubner (2009) find that goal difference displays a minimum sensitivity on statistical effects compared to the number of points. In particular, the authors show that the random component in the final ranking is somewhat smaller when using goal difference rather than the number of points.

Leader Dismissal and Leader Quit took the value of 1 for games played with the new leader after a leader dismissal and quit, respectively.

Interim took the value of 1 for games played with an interim leader, i.e., after the leader dismissal or quit but before the arrival of the new leader (when the latter was not in charge right after the previous leader dismissal or quit).

Counterfactual for Leader Dismissal (i.e., leader not dismissed for a similar level of performance for a team having dismissed its leader during another season) and Counterfactual for Leader Quit (i.e., leader not quitting for a similar level of performance for a team where the leader has quitted during another season) took the value of 1 right after the last game when the leader should have been in charge.

President Longevity was measured through the number of years as president prior to the current season and its square (President Longevity²). To assess the impact of President Longevity on the effectiveness of the decision to dismiss the leader or not, we also tested a number of variables based on the interaction between President Longevity or President Longevity² with Leader Dismissal or Counterfactual for Leader Dismissal.

Ownership Orientation was measured through three dummy variables with local, national and global orientation each taking the value of 1. To assess the impact of Ownership Orientation on the effectiveness of the decision to dismiss the leader or not, we also tested a number of dummy variables based on the interaction between the three dummies for Ownership Orientation and Leader Dismissal as well as Counterfactual for Leader Dismissal.

We also used three control variables specific to the football industry. The first control variable was a dummy variable for Home Advantage, taking the value of 1 when a team plays at home. The second control variable, Competitor Position, represents the position of the opponent in the league table before the game. The third control variable, Position Last Year, represents the position of the team in the league table at the end of the previous season.

Overall, 15 variables were dummy variables, However, because the dependent variable, six model variables and two control variables were not dummy variables, we estimated linear regression models rather than qualitative variables econometrics (Gourieroux 1989).

3.3. Data

We collected our data from various internet sources accessed for the last time on 26 July 2021: lfp.fr, racingstub.com/games, transfermarkt.com, oddsportal.com, football-data.co.uk and the French version of Wikipedia. Over the 1994–2016 period, 103 team-seasons changed their leader (i.e., 4.68 teams changing their leaders per season in average). Figure 1 provides information about the number of teams with a leader change per season. The two seasons 2004–2005 and 2015–2016 had the highest number of teams with a leader change (10). Interestingly, these two seasons preceded important increases in TV rights in 2005–2006 and 2016–2017 (Feuillet et al. 2019; Scelles et al. 2020b). It might mean that there was more pressure on teams to remain in Ligue 1, with the consequence that leaders were more likely to be dismissed.

Table 1. Variables used in the analysis: means and standard deviations (n = 4918).

	Mean	Standard Deviation
Goal Difference	−0.22	1.58
Home Advantage	0.51	0.58
Competitor Position	10.19	5.67
Position Last Year	11.47	6.06
Dismissal	0.22	0.42
Quit	0.05	0.22
Interim	0.002	0.05
Counterfactual—Dismissal	0.20	0.40
Counterfactual—Quit	0.05	0.21
Longevity	7.56	9.46
Longevity2	146.72	300.87
Dismissal * Longevity	1.73	5.54
Dismissal * Longevity2	33.64	166.65
Counterfactual—Dismissal * Longevity	1.46	5.06
Counterfactual—Dismissal * Longevity2	27.76	125.52
Local	0.70	0.46
National	0.11	0.32
Global	0.19	0.39
Dismissal * Local	0.16	0.37
Dismissal * National	0.02	0.15
Dismissal * Global	0.04	0.20
Counterfactual—Dismissal * Local	0.14	0.35
Counterfactual—Dismissal * National	0.01	0.12
Counterfactual—Dismissal * Global	0.04	0.21

provides information about the mean and standard deviation for each of the variables used in the analysis.

4. Results

Table 2. Results of the team fixed effect regressions (dependent variable: goal difference).

	Model 1	Model 2	Model 2′	Model 3	Model 3′
Home Advantage	0.663 ***	0.663 ***	0.663 ***	0.663 ***	0.663 ***
	(0.037)	(0.037)	(0.037)	(0.037)	(0.037)
Competitor Position	0.040 ***	0.040 ***	0.040 ***	0.040 ***	0.040 ***
	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)
Position Last Year	−0.019 ***	−0.019 ***	−0.016 ***	−0.018 ***	−0.018 ***
	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)
Dismissal	0.231 ***	0.232 ***		0.235 ***
	(0.056)	(0.056)		(0.056)
Quit	0.129	0.119	0.063	0.14	0.142
	(0.105)	(0.105)	(0.105)	(0.105)	(0.106)
Interim	0.041	0.046	−0.040	0.046	0.046
	(0.454)	(0.454)	(0.453)	(0.454)	(0.454)
Counterfactual—Dismissal	0.248 ***	0.251 ***		0.245 ***
	(0.059)	(0.059)		(0.059)
Counterfactual—Quit	0.504 ***	0.515 ***	0.458 ***	0.491 ***	0.500 ***
	(0.108)	(0.110)	(0.109)	(0.108)	(0.109)
Longevity		−0.001	−0.010
		(0.010)	(0.011)
Longevity2		0.0002	0.0003
		(0.0003)	(0.0003)
Dismissal * Longevity			0.032 ***
			(0.013)
Dismissal * Longevity2			−0.001 *
			(0.0004)
Counterfactual—Dismissal * Longevity			0.069 ***
			(0.016)
Counterfactual—Dismissal * Longevity2			−0.003 ***
			(0.001)
Local				ref.	ref.
National				−0.146	−0.133
				(0.197)	(0.209)
Global				0.116	0.065
				(0.098)	(0.109)
Dismissal * Local					0.192 ***
					(0.067)
Dismissal * National					0.182
					(0.170)
Dismissal * Global					0.424 ***
					(0.126)
Counterfactual—Dismissal * Local					0.241 ***
					(0.069)
Counterfactual—Dismissal * National					0.276
					(0.203)
Counterfactual—Dismissal * Global					0.259 **
					(0.126)
Constant	−0.588 ***	−0.585 ***	−0.501 ***	−0.710 ***	−0.700 ***
	(0.100)	(0.102)	(0.099)	(0.143)	(0.143)
Observations (Teams)	4918 (29)	4918 (29)	4918 (29)	4918 (29)	4918 (29)
R2	0.122	0.122	0.122	0.122	0.123

Notes: * p < 0.1; ** p < 0.05; *** p < 0.01. Standard errors are displayed in brackets.

provides the results of the team fixed effect regressions conducted to test our hypotheses. No heteroscedasticity was identified in our models; therefore, the results were not corrected. The three control variables had a significant impact on performance, with the expected sign, whereas interim had no significant impact.

H1a and H1b predict that performance improves after a leader dismissal or not in the same situation, whereas H1c and H1d predict that performance does not change after a leader quit or not in the same situation. Model 1 confirms H1a, H1b and H1c, but not H1d. This may indicate that performance improves after having been poor, except with an unplanned quit of the leader, which disturbs the organisation.

H2a predicts that president longevity has a positive impact on performance until reaching an inflexion point. Model 2 does not confirm H2a (no significant impact of Longevity and its square), which may indicate that the trade-off between learning-by-doing and learning from others described in the literature review has a counterbalanced effect.

H2b predicts that president longevity has a positive impact on the effectiveness of the decision to dismiss the leader or not until reaching an inflexion point. Model 2′’ confirms a significant positive impact of Longevity when interacting with Dismissal as well as with Counterfactual—Dismissal. It also confirms a significant negative impact of Longevity² when interacting with Dismissal as well as with Counterfactual—Dismissal, consistent with the idea of an inflexion point. It is reached after 22 years with Dismissal and 12 years with Counterfactual—Dismissal. In the latter case, this is not far from the 14 years identified by Shen and Cannella (2002) in another context. Overall, these results confirm H2b.

H3a predicts that global- and local-oriented ownership have a positive impact on performance compared to national-oriented ownership. Model 3 does not confirm H3a since both local- and national-oriented ownership have no significant impact compared to global-oriented ownership.

H3b predicts that global- and local-oriented ownership have a positive impact on the effectiveness of the decision to dismiss the leader or not compared to national-oriented ownership. Model 3′ confirms H3b because both global- and local-oriented ownership have a significant positive impact when interacting with Dismissal as well as with Counterfactual—Dismissal, whereas national-oriented ownership has no significant impact when interacting with Dismissal as well as with Counterfactual—Dismissal.

5. Discussion and Conclusions

In this paper, we examined the impacts of leader dismissal (or not in a similar situation), as well as president longevity and geographic orientation of owners on team performance. We tested our ideas in the context of French football organisations. Our empirical findings contribute to the literature on the effectiveness of risk-taking in leader succession. They inform not only the impact of a leader dismissal on team performance, but also some variables likely to affect the effectiveness of the decision to dismiss the leader or not.

5.1. Leader Dismissal or Continuity and Performance

Our findings support the hypothesis that performance improves after a leader dismissal. At the same time, they also support the hypothesis that performance improves after a leader dismissal that does not happen (control group), i.e., the leader is not dismissed in a situation similar to when the leader has been dismissed at another time. These findings are consistent with some other studies which have tested the impact of a leader dismissal on team performance in football with a control group in the English, German and Dutch contexts (Besters et al. 2016; Heuer et al. 2011; Van Ours and Van Tuijl 2016), as well as in the French context (Scelles and Llorca 2020). Nevertheless, further tests by the authors in the latter context showed that the control group was not significant anymore when controlling for expected performance.

The significant positive impact of the control group on performance questions the assumption that a leader dismissal has a positive impact compared to situations when the leader is not dismissed. However, the findings discussed here do not inform whether there are some factors affecting the impact of a leader dismissal or not on performance. Given that the president and the owners are the two stakeholders likely to have the most weight in the decision to dismiss the leader, it was worth investigating whether different types of presidents and owners affect the impact of a leader dismissal or not on performance.

5.2. President Longevity, Leader Dismissal or Continuity and Performance

To assess the impact of different types of presidents on the impact of a leader dismissal or not on performance, we looked at president longevity. The hypothesis was that a longer longevity generates a better knowledge of the organisation and its environment, leading to better decisions. At the same time, when the longevity reaches a certain point, the president may stop to question the evolutions encountered by the organisation and its environment, negatively impacting decisions. These elements may not only affect decisions, but also the working environment for the leader and the employees, with a direct impact on team performance.

Our findings do not support the hypothesis that president longevity has a positive impact on performance until reaching an inflexion point. In the Results section, we suggest that the trade-off between learning-by-doing and learning from others described in the literature review (Chen et al. 2017; Kor and Sundaramurthy 2009) may have a counterbalanced effect.

In contrast, our findings support the hypothesis that president longevity has a positive impact on the effectiveness of the decision to dismiss the leader or not until reaching an inflexion point. For the interaction between president longevity and leader dismissal, the inflexion point is reached after 22 years. For the interaction between president longevity and the control group for leader dismissal, the inflexion point is reached after 12 years. This is quite consistent with the 14 years found as an inflexion point by Shen and Cannella (2002) in another context. Although organisational cultural perpetuation is important in football organisations, it can hinder organisational change (Ogbonna and Harris 2014). This can ultimately have a negative impact on organisational performance.

5.3. Geographic Orientation of Owners, Leader Dismissal or Continuity and Performance

To assess the impact of different types of ownership on performance, we looked at the geographic orientation of owners, with a distinction between local, national and global orientation. The hypothesis was that local-oriented ownership has a better knowledge of the local environment. This was anticipated as a key factor for organisations historically tied to a territory such as football organisations. It was also expected that global-oriented ownership leads to additional managerial expertise. As such, local- and global-oriented ownerships were supposed to have a positive impact on performance, compared to national-oriented ownership. This hypothesis was an adaptation of Banalieva and Santoro (2009)’s findings on emerging market multinational enterprises to football organisations.

Our findings do not support the hypothesis that global- and local-oriented ownership have a positive impact on performance compared to national-oriented ownership (no significant impact), contrary to Banalieva and Santoro (2009). This suggests that even football organisations with access to global funding do not perform better than the others, despite their additional income. However, it must be noted that our sample only relied on the team-seasons with a leader change, or without a leader change but a performance more or less similar to the team-seasons with a leader change. These seasons are among the poorest for global-oriented organisations in terms of performance. As such, the conclusion that teams with global-oriented ownership do not perform better than those with local or national ownership has to be qualified. When it comes to explaining why local-oriented ownership has no significant impact on performance compared to national-oriented ownership, a reason may be that the additional income coming from a national-oriented ownership compensates the lower knowledge of the local environment.

Our findings support the hypothesis that global- and local-oriented ownerships have a positive impact on the effectiveness of the decision to dismiss the leader or not compared to national-oriented ownership. This is more consistent with Banalieva and Santoro (2009). This may confirm that the quality of the decisions made by football organisations are based on a trade-off between knowledge of the local environment (favouring local-oriented ownership) and managerial expertise (favouring global-oriented ownership).

5.4. Practical Implications

We suggest that there are four main practical implications that stem from our work.

First, both dismissing the leader (football head coach in our paper) or not dismissing them have a significant positive impact on performance. This highlights that there is not a single decision that can be made about the leader when performance is lower than expected. Instead, it is necessary to understand the context and history of the organisation, as well as the relationships between the leader and their employees, in order to inform such a risky decision.

Second, presidents having been in charge for a certain period of time are more likely to dismiss the leader or not without a positive impact on performance. They may consider that they have been in charge over a sufficiently long period to be able to make an informed decision alone, or do it to show to others that they are the ones in charge. However, the findings do not support the relevance of this approach. The implication is that presidents should listen to others when being faced with the decision to dismiss the leader or not, whatever their longevity.

The third implication is also related to presidents having been in charge for a certain period of time being more likely to dismiss the leader without positive impacts on performance. Nevertheless, it applies to owners instead of presidents. This implication is that owners should consider the longevity of the president and how it impacts their choices when deciding whether to dismiss them or not.

Additionally, global- and local-oriented ownerships have a positive impact on the effectiveness of the decision to dismiss the leader or not compared to national-oriented ownership. The implication is that organisations with national-oriented ownership should ensure that they are open to managers with global experience and expertise, as well as local actors understanding the local environment.

5.5. Limitations and Future Research

We acknowledge that this paper has a number of limitations, opening the door to future research.

First, we assumed a trade-off between learning-by-doing and learning from others when interpreting our findings for president longevity. A more qualitative approach based on case studies in different organisations would help confirm such a trade-off and better interpret our findings.

Second, although the president is usually the main decision maker when it comes to dismiss the leader or not in the context studied (football), board members are also important actors (Mangena et al. 2012). Unfortunately, it was not possible to gather information related to board members that could be used in this study. As for the first limitation, a more qualitative approach may help understand the relationships between the president and the board members, and their impact on the decision to dismiss the leader or not in different organisations.

Third, the distinction between local-, national- and global-oriented ownerships may miss the existence of intermediate types based on two or three geographic levels of orientation, consistent with the idea of ‘glocalization’ (Giulianotti and Robertson 2004). A finer analysis may reveal more types of ownership.

Fourth, some of the answers provided in the paper could be deemed ambiguous, e.g., when the results are not similar in the main model with “cumulative difference” (Table 2) and the alternative model with “cumulative surprise” (Table S1). A different model and other statistical tests might give more unambiguous answers.

Finally, the French football context is very specific, which questions the generalisability of the findings. However, the ideas to look at president longevity and geographic orientation of owners—with a distinction between local/national/global rather than local understood as national/regional (proximate confines of the country)/global—may provide relevant findings in other contexts.