Content uploaded by John-Erik Mathisen
Author content
All content in this area was uploaded by John-Erik Mathisen on Aug 30, 2014
Content may be subject to copyright.
http://lea.sagepub.com/
Leadership
http://lea.sagepub.com/content/8/2/169
The online version of this article can be found at:
DOI: 10.1177/1742715011420315
2012 8: 169Leadership
Jan Ketil Arnulf, John Erik Mathisen and Thorvald Hærem
noise-signal ratio in the firing of football managers
Heroic leadership illusions in football teams: Rationality, decision making and
Published by:
http://www.sagepublications.com
can be found at:LeadershipAdditional services and information for
http://lea.sagepub.com/cgi/alertsEmail Alerts:
http://lea.sagepub.com/subscriptionsSubscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
http://lea.sagepub.com/content/8/2/169.refs.htmlCitations:
What is This?
- Apr 11, 2012Version of Record >>
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
Article
Heroic leadership illusions
in football teams: Rationality,
decision making and noise-
signal ratio in the firing of
football managers
Jan Ketil Arnulf, John Erik Mathisen and Thorvald Hærem
BI Norwegian Business School, Norway
Abstract
Similar to practices in top management positions worldwide, there has been an increasing ten-
dency in recent decades to fire football managers when the team does not perform to the
stakeholders’ expectations. Previous research has suggested that improvements after change of
manager are a statistical artefact. Based on 12 years of data from the Norwegian Premier League,
we conduct a natural experiment showing what would have taken place if the manager had not
been fired. In this case, the performance might have improved just as well and even quicker.
Building on theories in expertise and decision making, we explore the data and argue that decision
makers may be fooled by randomness and learn wrong lessons about team leadership. Our
analyses support a post-heroic view of team leadership as an emergent, output variable.
Exaggerated focus on the individual manager may ruin long-term performance. Practical implica-
tions are discussed.
Keywords
post-heroic leadership, decision making, emergent leadership, leader succession, effects of
leadership, soccer
Corresponding author:
Jan Ketil Arnulf, BI Norwegian Business School, Nydalsveien 37, N-0442 Oslo, Norway
Email: jan.k.arnulf@bi.no
Leadership
8(2) 169–185
! The Author(s) 2012
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1742715011420315
lea.sagepub.com
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
Introduction
Does firing the accountable manager or coach help football teams improve their perfor-
mance? Despite the popularity of this option, several decades of research on leadership
succession suggests that this is far from certain and may even be detrimental. The exact
effect of individual leadership on organizational performance has been debated, and the
advocated effects span between the negligible (Lieberson and O’Connor, 1972; Pfeffer,
1977) to the substantial (Mackey, 2008; Thomas, 1988). In recent years, leadership research-
ers have increasingly taken a ‘post-heroic perspective’ (Conger et al., 2000; Ensley et al.,
2006; Huey and Sookdeo, 1994). The essence of this is that ‘heroic leaders do less than we
think they do, but they act as symbols of a cause’ (Stacey et al., 2000: 170). In line with this,
recent theoretical advances in team research (Ilgen et al., 2005) do not see causal relation-
ships in teams as linear, but instead as loops of interactions that each are outputs of previous
processes and inputs of the next (input-moderator-output-input (IMOI)). Leadership may be
treated as an outcome, not an input variable in a team (Day et al., 2004). Nevertheless, the
trend towards dismissing managers as a consequence of disappointing stakeholders by not
performing up to expectations is rising on a global level (Bøhren et al., 2004; Kato and Long,
2006; Lucier et al., 2006). This practice has spread from principal–agent theory in share-
holding organizations (Williamson, 1975) to become almost normative in other areas of
public interest (Ghoshal, 2005) such as sports, where the firing of the accountable managers
or coaches has become commonplace after poor performance (e.g. Audas et al., 1997). In
fact, Richard Audas and co-authors (Audas et al., 1997, 1999, 2002) have shown that a
positive effect on performance after the firing of a manager or a coach may be spurious and
even suggest that the effect may be detrimental to performance. Assuming that a post-heroic
model of leadership is correct and that leadership is just as much an output as an input of
team processes, the frequent termination of football managers’ tenure may be due to a
perception bias that exaggerates the importance of football managers to the team’s perfor-
mance. Based on research on distributed leadership (Day, 2001; Day et al., 2004; Ensley
et al., 2006) it is possible to argue that effective leadership is a kind of social capital within
the teams that may be ruined by the leader as an individual (Khurana, 2002). Our research
question is therefore: Might the perceptible situational characteristics associated with firing a
manager prevent boards from developing expertise on manager performance, and instead
induce an overconfident reliance on the value of individuals that disrupts effective develop-
ment of team performance?
Like most other organizations, sports teams have appointed leaders that practically and
symbolically represent the teams in the public eye, and also are accountable for the success
or failure of the team’s performance. Depending on the type of sports, this person is some-
times referred to as ‘coach’ (e.g. in American football) or sometimes as ‘manager’ (e.g. in
British soccer). In this paper, we choose to use the term ‘manager’, designating the person in
a leadership role in a soccer team. We do this in order to align our analysis with the language
in earlier research that we refer to. The reader may bear in mind that the leader/manager/
coach discussed here is thought to be the main decision maker on all issues that affect team
performance directly, and is thus viewed as accountable by the board and the public. The
case of sports team managers has already been explored by several researchers in this field.
Grusky (1963) identified an inverse relationship between the number of managerial changes
and the average team performance in a Major League Baseball sample. Using data from the
English Football League, Audas, Dobson, and Goddard (e.g. 2002) found that poor
170 Leadership 8(2)
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
performance predicted the mid-term firing of managers, and that teams were performing
worse upon dismissing a manager. They also found that teams performing below average
were more frequently changing managers. Using Dutch data, Koning (2000) claims that
changing a manager has an apparently positive effect on team performance, but that this
effect is a statistical artefact caused by opponent playing order. This proposal has since been
verified by several other researchers (Audas et al., 2002; De Paola and Scoppa, 2008). Based
on this we hypothesize that:
H1: Teams that fire their manager after a series of losses improve their performance after
changing their manager.
H2: Teams that do not fire their manager after a series of losses improve performance more than
teams which in the same situation do fire their manager.
Despite these disconcerting findings on changing managers, the tendency to replace
the head of the team has been strengthened in line with the trend in international
business (De Paola and Scoppa, 2008). Driven by the interests of professional investors,
modern-day corporate governance has seen an increasing global tendency to fire CEOs
in companies that do not perform up to shareholders’ expectations (Kato and Long,
2006; Lucier et al., 2006). In corporate governance, the decision to part with the CEO
may not be due to a thorough evaluation of this person’s capability in itself. Instead,
the boards may want to assure the owners that bad performance will have conse-
quences, so that the CEO in charge holds the confidence of both the board and the
owners. The governance of a sports club may not be entirely the same. The ownership
structure may be simpler, the business more transparent and the capability of the
manager could be easier to assess. The decision to fire a club manager should ideally
be a well-informed decision, in most cases leading to an improved performance by the
team. Why should sports clubs start copying the tendency to fire leaders if this turns
out to have dubious, even detrimental effects on the clubs?
We wonder whether this trend can be explained by assuming a learning effect coupled
with a specific type of decision-making scenario. According to so-called norm theory
(Kahneman and Miller, 1986; Kahneman and Tversky, 1982) negative outcomes are per-
ceived as worse when people can easily imagine that a better outcome could have occurred.
A loss will feel worse if it is due to something one does rather than if it stems from inaction.
Later research (Bar-Eli et al., 2007) has shown that the action/inaction ingredient can be
reversed if the expected norm is to act. In the study of Bar-Eli et al. (2007) football goal-
keepers were found to be expected to jump even if standing still would have been a statis-
tically more promising option. Similarly, a negative trend in performance for a football
team may put pressure on the team’s board, creating a norm to act – i.e. to dismiss the
manager. According to norm theory, further losses in a scenario of not acting would create
stronger negative emotions in the stakeholders than similar results after having ‘done
something’.
The decision to fire or keep a manager should optimally be an informed decision
made by an assessment of the performance of the manager in question. This requires
that the board can compare reliable information about the manager’s performance to a
valid body of knowledge about good and bad managers – the professional expertise of
the board in assessing manager talent and performance. However, research on recruit-
ment suggests that the assessment of performance and the knowledge about necessary
knowledge, skills and abilities are usually imperfect (Arnulf et al., 2010; Binning and
Arnulf et al. 171
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
Barrett, 1989) and subject to overconfidence, particularly in the field of hiring managers
(Kaiser et al., 2008; Khurana, 2001). The development of professional expertise, such as
in assessing manager performance, is dependent on certain types of contextual scenarios
(Kahneman and Klein, 2009). Environments are called ‘high-validity’ if ‘there are stable
relationships between objectively identifiable cues and subsequent events or between cues
and the outcomes of possible actions’ (Kahneman and Klein, 2009: 524). In the absence
of such relationships, the decision makers cannot develop expertise but fall victims
instead to an illusion of validity. For the boards of football clubs to develop expertise,
they will have to pay attention to whatever stable relationships there will be between
performance, the dismissal of a manager and the subsequent development of the team.
If the relationship between individual manager performance and organizational outcome
is weak, then the development of board expertise in hiring and firing managers will not be
driven by an evaluation of valid cues, but instead by possibly irrelevant cues such as ran-
domly distributed previous outcomes. This is similar to what Skinner (1978) calls ‘supersti-
tious behaviour’ or being fooled by randomness (Taleb, 2004). The crucial point is that this
behaviour cannot be modified by a counterfactual hypothesis. A counterfactual hypothesis is
stating what would have happened if the board had decided not to fire the manager. In line
with the original premises of norm theory, the decision to act provides a basis for learning
since it represents a salient point in time and a baseline from which future developments can
be traced. The decision not to fire a manager happens in principle after each match where the
manager stays on, but is probably less salient as a decision to the stakeholders and does not
provide a similar opportunity for learning.
Instead, the most salient cues to evaluate a decision would be the immediate effect
after the termination of the previous manager’s tenure, and thereafter the monitoring of
the development of the team’s performance under the new one. The difference between
a good and a lesser manager should optimally be discernible from the learning curve of
the team. The question is therefore: how long does it take for the influence of a good
manager to take effect on the performance of a team? Adherents of a heroic view on
leadership may adopt an unrealistic expectation of almost instant effects of leadership
(Conger, 2000; Khurana, 2002; Meindl et al., 1985), but a more realistic approach
would take into consideration that leadership needs time to exert influence on the
surroundings, a view supported by empirical research (e.g. Huey and Sookdeo, 1994;
Lieberson and O’Connor, 1972; Mintzberg, 2004; Strang and Macy, 2001). We therefore
hypothesize that:
H3: Team/manager mutual performance improves over time such that a team’s average
performance continues to improve over at least one season after the entry of a new
manager.
Whatever the leadership theory the stakeholders hold (implicitly or explicitly), there is no
rule for establishing when – and how abruptly – changes should come to indicate that the
new manager is on a promising trajectory. If changing a manager is seen as an investment in
better future performance, the monitoring of the new manager may be following similar
patterns to investors monitoring the performance of their assets. Taleb (2004) has shown
how the time horizon and the frequency of observations may distort the observer’s impres-
sion of the returns. Assuming that an investment is earning a return on 15% per year with an
error rate (volatility) of 10%, the probability of a gain on a yearly basis is 93%. But within
172 Leadership 8(2)
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
any one day, the probability of a gain is only 54%, which is experienced as close to random –
the noise almost outweighs the signal. Applied on the football manager scenario, this means
that any actual effect of the manager on the team performance will be distorted by statistical
noise from other sources, such as home/visiting team, opponent strength and rank order as
well as random effects. In order to build expertise, the information about valid causal effects
must be discernible from the noise of random variations (Kahneman and Klein, 2009). Thus,
stakeholders monitoring the development of a new football manager will be susceptible to
different noise/signal ratios depending on the chosen timeframe:
H4: The salience (noise/signal ratio) of contingencies in the relationship between manager and
team performance will depend on the time perspective of the analysis.
The only sure sign that a team is on a winning trajectory is that it is accumulating more
points than it is losing. But since humans are shown to let negative emotions bias decision
making towards short-term outcomes (Gray, 1999), the effects of losses may overshadow the
effects of gains. In football, there are three possible outcomes of a match: three points for
victory, one for a draw, and zero for a loss. The average yield of points in a group of teams
will therefore be 1.38 points. A team which can keep a stable floating average performance of
more than 1.38 points will thus be outperforming others. This criterion is objective and fixed,
but a learning curve is not. A tendency towards improvement can be identical for two
different managers and teams (e.g. expressed as a beta coefficient), but the time it takes to
reach a stable winning position of more than 1.38 points will be different if the baseline is
different. This means that a well-performing team will be easier to bring into a winning
position than a team with a less favourable starting point. A better manager could actually
witness his team lose more matches than a lesser manager, if the team in question needs to be
moved from a lower rank. In organizational terms, this means that the social capital of the
team – the competence, culture, distributed leadership and other non-individual assets – will
be major determinants in the eventual success of the new manager (Bourdieu, 1983; Day,
2001; Day et al., 2004, 2006; Ensley et al., 2006). Thus:
H5: The time for a manager to reach a stable winning performance depends less on the man-
ager’s learning curve than on the baseline of learning, which is a property of initial team per-
formance (ranking) as the new manager enters the team.
It follows from our line of argumentation that a focus on the manager as responsible for a
team’s success may be exaggerated by a combination of heroic expectations and decision
biases. If this is indeed a distortion of the actual causal mechanisms implied in team lead-
ership, as the IMOI model suggests (Day et al., 2004), then this may create types of learning
that deflect attention from the really effective relationships and instead destroy value, as
suggested by previous findings (De Paola and Scoppa, 2008). There are at least four reasons
why this could be destructive: a) A ‘fascination for leadership’ reinforced by brief, but
spurious improvements of performance creates an illusion – also shared by the manager
himself – of a heroic manager (Chen and Meindl, 1991). ‘Heroic’ managers will appear more
attractive in their own eyes and the eyes of possible employers, increasing the likelihood that
they will leave their present positions. b) Such illusions are in turn vulnerable to sudden
‘disenchantment’ (Cha and Edmondson, 2006), and since the teams themselves may be
protected by the exculpating ‘halo effect’ of teams (Naquin and Tynan, 2003), this disen-
chantment may lead to quicker termination of managerial assignments. c) Increased focus on
Arnulf et al. 173
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
heroic leadership can make the teams destroy social and organizational capital by giving new
managers undue leeway to make changes, as shown in the effects of externally recruited
CEOs in corporations (e.g. Lucier et al., 2006). Change initiatives by the new manager may
function as a charismatic marker signifying exceptional characteristics (e.g. Conger, 1992),
but may also disrupt the internal sensemaking processes in the team that belong to the social
capital and secure viable long-term development (Lucier et al., 2006; Weick, 2000). c) Real-
life training in overcoming crises and obstacles are a way of establishing team co-operative
capacity and team leadership (Burke et al., 2006; Halverson et al., 2004; Salas et al., 1999).
Establishing an organizational routine of changing leaders when in crisis is a type of scape-
goating that may prevent learning and the establishment of social capital. This leads us to
hypothesize that:
H6: Firing a manager because of deteriorating team performance is a learnt organizational
routine, such that:
H6a: the more frequently a team changes managers, the quicker the next manager will leave; and
H6b: the more frequently a manager changes teams, the briefer the next assignment will be; and
H6c: the interaction between length of managerial tenure and team manager firing rate leads to a
negative development in performance.
Methods
We decided to test our Hypotheses 1 and 2 in a natural experiment. By identifying the typical
performance curve that leads to the dismissal of a manager, we can also pose the counter-
factual question: what would have happened if the manager had not been dismissed? The
testing of this hypothesis is possible by identifying a similar series of developments that fit
the typical curve towards manager dismissal, but where this did not happen and the manager
stayed on for the necessary number of comparing matches after the point at which he would
have been fired. We will then test Hypotheses 3–6 by applying the whole dataset and explor-
ing the relationships between the several cues visible to the stakeholders and how they
appear to relate to the outcomes. If a change of manager appears to the participants as
an effective treatment, it is likely that this behaviour will be learnt, i.e. there is an increased
tendency to repeat this choice. If this behaviour is superstitious, in the sense that it is not in
accordance with – or even violating – the effective underlying relationships, an increased
tendency to change managers could have a negative effect on performance.
The sample
We have chosen to use the Norwegian Premier League (‘Tippeligaen’) as the basis for our
dataset. The observations are all matches from the 1995 season through to 2006, because this
is when the league was enlarged from 12 to 14 teams; it also represents a turning point in the
league’s professionalism regarding more sponsorship and media coverage. During the 12
observed seasons, 2184 matches were played and 119 managerial spells were covered, 86
managers leaving their position involuntarily and 33 leaving their position voluntarily. All
information regarding outcome variables for matches played was manually obtained from
the NIFS (Norwegian and International Football Statistics). The total of different teams that
went in and out of the top division during the time of study was 25.
174 Leadership 8(2)
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
When creating the sample for the counterfactual question (what would have happened if
the manager had not been fired), we created a more restricted sample to ensure the compa-
rability of the two groups. The first group selected was the ‘treatment group’, where the
‘treatment’ was to replace the manager. This group consists of 21 cases where the manager
left his position involuntarily. Each single case consists of the last 15 matches before the
manager was fired from his position, and the 15 first matches under the new manager.
Accordingly, each case consists of 30 matches.
The control group teams were selected for following the same negative slope for
15 matches in a row that ultimately led to the dismissal of the manager in the ‘treatment
group’. In this control group the same manager was in charge of the team during the
30 matches, since there was no dismissal of the manager anywhere in the series of matches.
The mean point obtained for the teams over the 30 matches was calculated across all the
cases. This control group consisted of 82 cases.
We classified managerial termination as voluntary or involuntary according to the guide-
lines set out by Audas et al. (1999). They suggest investigating the circumstances of each
dismissal in terms of team performance prior to termination, the managers’ subsequent
employment (if any) and the publicized reason for termination if available. We scanned
all Norwegian newspapers from 1995 to 2006 for information and classified cases as invol-
untary if there was no obvious reason other than poor results or other pressures emanating
from the current employment. Terminations were classified as voluntary if the information
revealed that the initiative came from the manager himself. Such information included an
immediate or imminent move to another club of comparable or higher status; a move to a
European or national team manager’s position; or voluntary retirement from management.
Our data shows that 53.7% of involuntary job terminations occur after the end of a seasons.
Accordingly, involuntary job terminations occur approximately at the same frequency
between seasons as within seasons. Therefore the selection procedure for assigning teams
to the two groups was followed regardless of the fact that some matches in some cases
crossed two different seasons.
To increase similarities between the groups which in turn enhances internal validity
(Shadish et al., 2002), the teams assigned to the two groups were matched such that each
group consisted of the same teams. By applying the matching procedure the two groups
contained 19 teams each. The 19 teams represented 11 out of the 25 teams in the sample.
The treatment group consisted of cases from the whole range of seasons from 1995 up to
and including 2006. The same is true for the control group except for the 1995 and 2006
seasons. In the treatment group team’s positions on the table at the time when the decision to
fire the manager was made ranged from 2nd to 12th place, except for positions 5, 9 and 11.
In the control group the position in the table for the teams selected ranged from 2nd to 12th
place.
The matching procedure also left the treatment group with 29 different managers and the
control group with 15 different managers; 10 different managers appear in both groups.
Finally, out of the 86 different managers included in the whole dataset 44 were included
in the two groups.
To test that the control group had the same development as the treatment group in the
period before firing the manager we calculated the average score over the 15 matches for the
control and treatment group. The average scores were 21.76 and 20.0, respectively. The one-
way analysis of variance (ANOVA) shows that the difference between the scores for the two
groups was not significant (N ¼ 19, F ¼ 1.20, p > .10).
Arnulf et al. 175
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
Results
Figure 1 shows the teams’ performance during the last 15 matches before a manager is fired.
The statistically typical scenario for a team that fires its manager is a steep decline in per-
formance from gaining position (> 1.38 points on average) to losing position. An average of
about 1.1 points seems to be the threshold for the decision to make a forced dismissal of the
manager. As shown by Figure 1, the typical development after firing a manager is a return to
winning levels 6–7 matches after the arrival of the new manager. To the decision makers,
their decision to fire the manager will appear justified. H1 is therefore supported.
On the right-hand side of the ‘0 match number’ on the x-axis, it is discernible that the
performance of the teams that did not fire the manager seems to recover more quickly than
the teams that did. The ‘post-decision’ scenario for the two types of teams – manager fired/
not fired – is strangely identical, even to the extent of later fluctuations. We tested
Hypothesis 2 (that teams that do not fire their managers after a series of losses improve
performance more than teams, in the same situation, that fire their managers) with a one-
way ANOVA. The treatment group scored over 15 matches on average 2.87 fewer points
than the control group (F ¼ 3.12, p < .05 in a one-tailed test), indicating that teams firing
their manager achieved significantly fewer points than teams keeping their manager. H2 is
therefore supported.
This difference, however, will not be clearly visible to the stakeholders, who probably will
be most influenced by the fact that a change of manager seems to be an effective treatment
for bad performance. Are there other cues of effective relationships that may appear to the
participants? To check how expertise may or may not be established, we correlated a number
of features of the managers’ and teams’ situations with outcome variables, but varied the
time horizon (Table 1). In the perspective of a single match, the overwhelmingly important
determinant is the arena (home/visitor). The longer the time perspective, the less important
this determinant becomes. The correlation between manager experience within the team and
outcomes is stronger when using average points over 10 matches (a medium-long
Figure 1. Performance before and after firing the manager.
176 Leadership 8(2)
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
Table 1. Correlations of outcome variables by contextual factors and time frame . N ¼ 4368, all correla-
tions significant and not flagged.
Single
match
outcome
Points
average last
5 matches
Points
average
last 10 matches
Current table
position
Number of previous
manager assignments
0.01 0.01 0.02 0.01
Match played by this manager
with this team
0.03 0.10 0.15 0.31
Number of managers previously
had by the team
0.00 0.00 0.02 0.10
Career length of the manager 0.09 0.20 0.26 0.25
Table position (when the new
manager was hired)
0.14 0.29 0.37 0.74
Home / visiting team 0.19 0.05 0.03 0.00
Figure 2. Effects of manager tenure on performance.
Note: reference line ¼ 1.38 points, signifying stable gains.
Arnulf et al. 177
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
perspective) than over only 5 matches (medium-short perspective). In the long run, indicated
by the current table position, it seems as if the initial position within the series is the most
important determinant (note when reading Table 1 that in the column ‘table position’, lower
scores indicate better performance – e.g. 1 ranks better than 12). H4 is thereby supported.
Is it possible to distinguish between career trajectories of successful and not-so-successful
managers? Figure 2 plots the average development of the teams’ performance as a function
of manager tenure.
This table seems to indicate that a good manager is only performing as a stable winner
after the first season. H3 is thereby supported. An obvious objection to this might be that
this average plot is created by attrition of the worst-performing managers, making the good
performers more salient among the longer careers. To investigate this possibility, we broke
down the plot by career lengths.
Figure 3 shows that while all managers are improving their results during their first
season, there seems to be a tendency for some managers to attain higher levels of perfor-
mance quicker. The shortest tenures are systematically operating at performance levels infe-
rior to the longer. Is it possible, however, that this initial performance is due to an
extraneous factor – the initial performance of the team? We ran a general linear model
(GLM) analysis with average performance on five matches as dependent variables, entering
the tenure (match number in this team) of the manager and the initial position of the team as
fixed factors. The effects of manager tenure and initial position as well as their interaction
effects were significant (p < .01). However, the most powerful effect was that of the initial
Figure 3. Effect of manager’s career length on performance.
178 Leadership 8(2)
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
position alone (partial eta squared ¼ 0.11 versus 0.03 for the tenure of the manager).
Figure 4 plots these relationships.
It turns out that the differences in the initial performances of the teams – their social and
structural capital – are stronger determinants of later performance than the learning curves
of the managers themselves. H5 is thereby supported.
To test Hypothesis 6, we ran a GLM analysis, again with average points won during the
last five matches as dependent variable, but this time with number of managers had by the
team and number of assignments held by the managers as fixed factors. The main effects and
the interaction were small, but significant (p < 0.5). It turns out that although there is
considerable variation in effects, the statistically most likely effect of frequent changes in
manager assignments will be a deterioration of performance below the break-even level of
1.38 points. The best results seem to be achieved by managers working for a limited number
of teams – when the teams also employ a limited number of managers, as shown in Figure 5.
This supports Hypothesis 6.
Discussion
The purpose of this study was to use a counterfactual approach to test how the decision to
not fire a manager affects team performance compared to the decision to dismiss the man-
ager. Our findings are in accordance with previous findings in this field (Audas et al., 1997,
1999, 2002) in that average results have a tendency to improve immediately upon changing a
2.5–
2.0–
1.5–
1.0–
0.5–
0.0–
l l l l l l l l l l l l l l l l l l
5 8 11 13 15 17 21 27 31 38 48 52 58 71 77 87 99 121
Total no. of games played by this manager in this team
Note: Non-estimable means are not plotted
Estimated marginal means
Average points last 5 games by initial competitive position of the team
Initial Grouping
4th or better
5th through 8th position
9th or worse
Figure 4. Performance development by initial position and length of assignment.
Arnulf et al. 179
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
manager. When we pose the counterfactual question – what would happen in a similar sce-
nario of negative development if the manager is not fired – it turns out that the improvement in
performance is similar or even markedly better. This finding is in line with previous research
suggesting that the quick improvements seen after manager change may be attributable to the
sequence of opponents within the league (Koning, 2000). We then hypothesized that according
to norm theory (Bar-Eli et al., 2007; Kahneman and Miller, 1986), the effects of what one does
will be more salient in shaping beliefs about negative outcomes than the choice of inaction,
particularly when the norm is to act in the face of adverse events. It is therefore more likely that
teams will learn to fire managers than to keep them in times of trouble. We think that our data
supports that this is what actually happens, as there is a tendency for teams to increase their
frequency of firing managers in times of trouble. This does not only seem to be a temporary
response to worries among supporters and board members, but a heightened readiness to
dismiss managers continues to exist in these teams as if a routine of early dismissal has been
established through organizational learning (Argyris and Scho
¨
n, 1996).
There are of course many other reasons to fire a manager than mere performance mea-
sures. Communicative ability, relationships, loyalties, public popularity and other social
issues may be just as important. But the social reputation of a manager is a narrative con-
struction emanating from the experiences of the surroundings (Berger and Luckman, 1966;
Weick, 1995), and there are probably few pure facts separate from the explanations con-
structed around the people involved. Our data show that lack of results informs this
sensemaking process, in that deteriorating performance often precedes the firing of a man-
ager. The more rational this process is, the better informed the firing of the managers should
be. It is actually surprising that we find no statistical tendency in favour of terminating the
contract of a manager since there surely must be instances where the firing of the manager is
justified or a constructive option. Our interpretation of this is that a benign, positive
1.6–
1.4–
1.2–
1.0–
l l l l
1 2 3 4
Manager assignment no.
Non-estimable means are not plotted
Estimated marginal means
Average points last 5 games by manager career and team manager turnover
Managers per team
<=4 managers
5 - 8 managers
9 managers or more
Figure 5. Team performance as a function of the interaction between the managers’ number of previous
positions and the team’s number of previous managers.
180 Leadership 8(2)
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
interpretation of bad performance is actually indicating that a good leadership process is at
work, whereas the individual attribution involved (Kelley, 1967) in the scapegoating of
dismissed managers may be inseparable from the effective team leadership process itself:
as long as there is hope, the manager is a hero even when facing superior opponents, but
when doubt and despair prevails, the same person is perceived as a loser.
According to more recent theories in teams, leadership and team leadership (Burke et al.,
2006; Day et al., 2004, 2006; Ensley et al., 2006; Lord and Hall, 2005), the establishment of
leadership in teams is just as much an output as an input variable and heroic attributions
of leadership as an individual variable can disturb the emergence of effective distributed
leadership phenomena. Considering the extensive and hands-on experience of football man-
agers and the stakes involved in team leadership and teams, one should expect the stake-
holders to develop sufficient expertise to be able to handle manager succession in a qualified
way, making the best possible use of the available knowledge. However, as shown by decades
of research in expertise and decision making, even professionals are susceptible to decision
biases, and some scenarios do not lend themselves to the establishment of professional
expertise at all (e.g. Kahneman and Klein, 2009; Tversky and Kahneman, 1974). Our find-
ings here suggest that the boards of football teams may be overly biased by information
related to the individual manager, thereby having their attention directed at spurious statis-
tical information. From our analyses, it seems reasonable to infer that successful manager
performance takes time to develop and depends on a more thorough mutual learning process
than acquiring a few quick victories. The prospects of success for a manager may appear
different depending on the time frame used. Statistically, the effect of the individual manager
on team performance is increasing with longer frames of time. However, in our data, the
effect of the manager comes nowhere near the effect of the team’s own social capital, mea-
sured as its ranking within the league when the manager started his assignment.
When comparing early signals of competence, it seems as if all managers need to be going
through a period of ‘apprenticeship’ in the sense that all of them show increased perfor-
mance during the first months in office. Superficially, the managers who later will be fired
early seem to be performing at a lower level than their more successful colleagues.
When controlling for social capital or structural resources – the competitive performance
of the teams prior to the manager’s engagement – there does not seem to be a big difference
between the long-lasting managers and those who leave earlier.
Our analysis concerns data that are cyclical in nature, and objections are possible. Or the
finding that better results are created by managers with longer tenures may be caused by an
attrition of the lesser managers, judged on other criteria than the similarity of their learning
curves. Also, we have assumed that managers are dismissed due to bad performance, and
that these managers are identical to the managers who were not fired in a similar situation,
judged from the accumulated points alone. It may be that the managers who were not fired
were kept because the boards possessed other types of information about the situation, such
as an impression about personality or social skills with the team. To us, this would just prove
the point: team leadership is about creating a social reality and a different interpretation of
events that helps overcome hard times (Berger and Luckman, 1966; Burns, 1978; Drucker,
2002). Despite an encompassing literature on team-building, the knowledge about how to
instrumentally install this type of phenomenon is still not available (Salas et al., 1999).
While mature leadership may foster a collective we-feeling (Lord and Hall, 2005), research
on celebrity leadership (Treadway et al., 2009) and the dark side of leadership (e.g. Kaiser
et al., 2008) shows that boards and leaders alike are susceptible to heroic images that boost
Arnulf et al. 181
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
the leader’s career, compensation and fame, but actually harm the organization and orga-
nizational performance. While our data stem from professional sports, these issues are well
known and relevant to a large range of organizational contexts.
When success or failure is attributed to the manager as an individual, this may actually
damage the possibility of building such organizational capital. This proposition is in accor-
dance with the detrimental development befalling teams and managers who are increasingly
‘promiscuous’ in hiring and firing each other. Teams are reinforced in their firing behaviours
by the apparent gains and managers learn how to appear ‘employable’ by behaving and
communicating in ways that seem promising to the boards that hire them. Khurana (2002)
calls this ‘the curse of the superstar CEO’ since organizations do not seem to be able to
achieve successes by this approach. It indicates a profound difficulty among the decision
makers in identifying success factors that may lead to an effective combination of team needs
and manager characteristics.
The findings in this study are instead compatible with a post-heroic, IMOI-perspective on
leadership (Day et al., 2004) as a mutual social interaction. We do not argue with the fact that
managers sometimes need to leave a team, but we believe our data suggest that this option is
used too often. The causality issue inherent in leadership/followership concepts is far more
circular than often assumed, and the development of a professional expertise in team leader-
ship is possibly still suboptimally existent in the management of football teams.
Limitations
There are several limitations to our study. Our findings are based entirely on the statistical
analysis of archival data. We therefore do not know if the actual processes that led to the
dismissal of these managers were motivated by other, more reliable cues about performance
that were not discernible from our data.
The actual games and points are reliable as entered in the official sports statistics, but
information about the managers, their assignments and careers are taken partly from official
statistics and partly from the sports press. We cannot rule out that this information is
incomplete or distorted.
References
Argyris C and Scho
¨
n DA (1996) Organizational Learning II. Addison-Wesley: Reading, MA.
Arnulf JK, Tegner L and Larssen O
¨
(2010) Impression making by re
´
sume
´
layout: Its impact on the prob-
ability of being shortlisted. European Journal of Work and Organizational Psychology 19(2): 221–230.
Audas, R., Dobson, S., and Goddard, J. (1997) Team performance and managerial change in the
English football league. Economic Affairs (September): 30-36.
Audas R, Dobson S and Goddard J (1999) Organizational performance and managerial turnover.
Managerial and Decision Economics 20: 305–318.
Audas R, Dobson S and Goddard J (2002) The impact of managerial change on team performance in
professional sports. Journal of Economics and Business 54(6): 633–650.
Bar-Eli M, Azar OH, Ritov I, Keidar-Levin Y and Schein G (2007) Action bias among elite soccer
goalkeepers: The case of penalty kicks. Journal of Economic Psychology 28(5): 606–621.
Berger P and Luckman T (1966) The Social Construction of Reality: A Treatise in the Sociology of
Knowledge. Garden City, NY: Doubleday.
Binning JF and Barrett GV (1989) Validity of personnel decisions – a conceptual analysis of the
inferential and evidential bases. Journal of Applied Psychology 74(3): 478–494.
182 Leadership 8(2)
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
Bøhren Ø, Sharma S and Vegarud T (2004) Eierstyring i store norske selskaper: Oppsigelse av
toppleder. In: Thorsby Ø (ed.) Corporate governance i et norsk perspektiv. Oslo: Abstrakt forlag.
Bourdieu P (1983) Soziale Ungleichheiten. In: Kreckel R (ed.) Soziale Welt, Sonderheft, Vol 2.
Goettingen: Otto Schartz and Co, 183–198.
Burke CS, Stagl KC, Klein C, Goodwin GF, Salas E and Halpin SM (2006) What type of leadership
behaviors are functional in teams? A meta-analysis. Leadership Quarterly 17(3): 288–307.
Burns JM (1978) Leadership. New York: Harper and Row.
Cha SE and Edmondson AC (2006) When values backfire: Leadership, attribution, and disenchant-
ment in a values-driven organization. The Leadership Quarterly 17: 57–78.
Chen CC and Meindl JR (1991) The construction of leadership images in the popular press: The case of
Donald Burr and People Express. Administrative Science Quarterly 36: 521–551.
Conger JA (1992) The dark side of leadership. Organizational Dynamics 19(2): 44–55.
Conger JA (2000) Effective change begins at the top. In: Beer M and Nohria N (eds) Breaking the Code
of Change. Boston, MA: Harvard Business School Press, 99–112.
Conger JA, Kanungo RN and Menon ST (2000) Charismatic leadership and follower effects. Journal of
Organizational Behavior 21(7): 747–767.
Day DV (2001) Leadership development: A review in context. Leadership Quarterly 11(4): 581–613.
Day DV, Gronn P and Salas E (2004) Leadership capacity in teams. The Leadership Quarterly 15: 857–880.
Day DV, Gronn P and Salas E (2006) Leadership in team-based organizations: On the threshold of a
new era. Leadership Quarterly 17(3): 211–216.
De Paola M and Scoppa V (2008) The effects of managerial turnover: Evidence from coach dismissals
in Italian soccer teams. Available at: http://mpra.ub.uni-muenchen.de/11030/.
Drucker PF (2002) The Effective Executive Revised. New York: HarperBusiness.
Ensley MD, Hmieleski KM and Pearce CL (2006) The importance of vertical and shared leadership
within new venture top management teams: Implications for the performance of startups.
Leadership Quarterly 17(3): 217–231.
Ghoshal S (2005) Bad management theories are destroying good management practices. Academy of
Management Learning and Education 4(1): 75–91.
Gray JR (1999) A bias toward short-term thinking in threat-related negative emotional states.
Personality and Social Psychology Bulletin 25: 65–75.
Grusky O (1963) Managerial succession. American Journal of Sociology 49: 21–31.
Halverson SK, Murphy SE and Riggio RE (2004) Charismatic leadership in crisis situations: A lab-
oratory investigation of stress and crisis. Small Group Research 35(5): 495–514.
Huey J and Sookdeo R (1994) The leadership industry. Fortune 129(4): 42–50.
Ilgen, D. R., Hollenbeck, J. R., Johnson, M., and Jundt, D. (2005) Teams in organizations: From
Input-Process-Output Models to IMOI Models Annual Review of Psychology (January).
Kahneman D and Klein G (2009) Conditions for intuitive expertise a failure to disagree. American
Psychologist 64(6): 515–526.
Kahneman D and Miller DT (1986) Norm theory: Comparing reality to its alternatives. Psychological
Review 93(2): 136–153.
Kahneman D and Tversky A (1982) The psychology of preferences. Scientific American 246(1): 160–173.
Kaiser RB, Hogan R and Craig SB (2008) Leadership and the fate of organizations. American
Psychologist 63(2): 96–110.
Kato T and Long C (2006) CEO turnover, firm performance, and enterprise reform in China: Evidence
from micro data. Journal of Comparative Economics 34(4): 796–817.
Kelley HH (1967) Attribution theory in social psychology. In: Levine D (ed.) Nebraska Symposium on
Motivation, Vol. 15. Lincoln: University of Nebraska Press.
Khurana R (2001) Finding the right CEO: Why boards often make poor choices. MIT Sloan
Management Review 43(1): 91ff.
Khurana R (2002) The curse of the superstar CEO. Harvard
Business Review 80(9): 60–66.
Arnulf et al. 183
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
Koning RH (2000) An econometric evaluation of the firing of a coach on team performance. Applied
Economics 35(5): 555–564.
Lieberson S and O’Connor JF (1972) Leadership and organizational performance: A study of large
corporations. American Sociological Review 37(2): 117–130.
Lord RG and Hall RJ (2005) Identity, deep structure and the development of leadership skill. The
Leadership Quarterly 16: 591–615.
Lucier C, Kocourek P and Habbel R (2006) CEO Succession 2005: The crest of the wave. Booze
Hamilton Features Report. Available at: http://www.strategy-business.com/.
Mackey A (2008) The effect of CEOs on firm performance. Strategic Management Journal 29(12):
1357–1367.
Meindl JR, Ehrlich SB and Dukerich JM (1985) The romance of leadership. Administrative Science
Quarterly 30: 78–102.
Mintzberg H (2004) Leadership and management development: An afterword. Academy of
Management Executive 18(3): 140–142.
Naquin CE and Tynan RO (2003) The team halo effect: Why teams are not blamed for their failures.
Journal of Applied Psychology 88(2): 332–340.
Pfeffer J (1977) The ambiguity of leadership. Academy of Management Review January. 104–112.
Salas E, Rozell D, Mullen B and Driskell JE (1999) The effect of team building on performance: An
integration. Small Group Research 30(3): 309–329.
Shadish WR, Cook TD and Campbell DT (2002) Experimental and Quasi-experimental Designs for
Generalized Causal Inference. Boston, MA: Houghton Mifflin.
Skinner BF (1978) Reflections on Behaviorism and Society. Englewood Cliffs, NJ: Prentice-Hall.
Stacey RD, Griffin D and Shaw P (2000) Complexity Theory and Management: Fad or Radical
Challenge to Systems Thinking? London: Routledge.
Strang D and Macy MW (2001) In search of excellence: Fads, success stories, and adaptive emulation.
American Journal of Sociology 107(1): 147–182.
Taleb NN (2004) Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets, 2nd
edn. New York: Random House.
Thomas AB (1988) Does leadership make a difference to organizational performance?. Administrative
Science Quarterly 33: 388–400.
Treadway DC, Adams GL, Ranft AL and Ferris GR (2009) A meso-level conceptualization of CEO
celebrity effectiveness. The Leadership Quarterly 40(4): 554–570.
Tversky A and Kahneman D (1974) Judgment under uncertainty: Heuristics and biases. Science
185(4157): 1124–1131.
Weick KE (1995) Sensemaking in Organizations. Thousand Oaks, CA: Sage.
Weick KE (2000) Emergent change as universal in organizations. In: Beer M and Nohria N (eds)
Breaking the Code of Change. Boston, MA: Harvard Business School Press, 223–240.
Williamson OE (1975) Markets and Hierarchies: Analysis and Antitrust Implications. New York: Free
Press.
Jan Ketil Arnulf earned his Ph.D. from the University of Oslo. He is Associate Professor at
BI Norwegian Business School teaching leadership development and organizational behav-
iour, and was recently the associate dean of the school’s MBA programme in Shanghai,
China. His research areas are leadership and cross-cultural leadership development, and he is
also involved in practical consulting on these issues.
John-Erik Mathisen has an MSc degree in organizational psychology from BI Norwegian
Business School, where he is currently a doctoral student. Prior to this he was a practising
manager and entrepreneur in transport and logistics.
184 Leadership 8(2)
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from
Thorvald Hærem earned his Ph.D. at Copenhagen Business School in Denmark and is now
Associate Professor of Organizational Psychology at BI Norwegian Business School. His
research interests include technology in organizations, organizational and individual rou-
tines, behavioural decision making, and expertise. He has published his research in journals
such as Journal of Applied Psychology, Journal of Behavioral Decision Making,
Organizational Studies, and Organization Science.
Arnulf et al. 185
at BI Norwegian Business School on April 11, 2012lea.sagepub.comDownloaded from