TRANSFERS IN SOCCER:
THE IMPACT OF THE ENTRY AND DEPARTURE OF CENTRAL TRANSFERS ON TEAM PERFORMANCE
Master thesis, MscBA, specialization Human Resource Management University of Groningen Faculty of Economics and Business
January, 17th, 2010
W. NATHALIE FEITSMA Msc Student number: 1531913
Spoorlaan 3 9753 HV Haren (GN) Tel.: +31 6 16 954 606 E-mail: w.n.feitsma@student.rug.nl
Supervisor/university Prof.dr. G. S. van der Vegt
Acknowledgement:
The author is indebted to Prof.dr. van der Vegt, who initiated this research project and provided me with advice and encouragement. I would also like to thank drs. van der Kam who made a great contribution to the retrieval of football data. Finally, I am grateful for the support I received during any stage of this research project and during my studies from my parents, friends and co-students.
TRANSFERS IN SOCCER:
THE IMPACT OF THE ENTRY AND DEPARTURE OF CENTAL TRANSFERS ON TEAM PERFORMANCE
ABSTRACT
In this study the impact of transfers, i.e. team members who switch teams and who are acquainted with the team task, on team performance in soccer is assessed. It is proposed that the extent to which team performance is affected by transfers, is dependent on the centrality of those transfers within the team. Both the centrality of transfers who enter and who depart a team are considered. The impact of player transfers is examined at two moments; the legally regulated transfer periods in a soccer season. Hypotheses were tested for eight European top soccer leagues and 143 teams for the 2009-2010 season.
The results found in this study were inconclusive. No differential effects were found for the centrality of transfers and performance, independent of time and whether it concerned an entering or departing transfer. More research is warranted to explain these findings and to assess other measures for centrality.
Keywords:centrality, transfers, soccer, teams, team development.
INTRODUCTION
Team membership changes are fairly common in many organizations (Shaw, Duffy, Johnson, & Lockhart, 2005a). Membership change can be defined as the departure and/or entry of a formally selected team member (cf. Arrow & McGrath, 1995; Levine & Choi, 2004). Various types of members can depart or enter a team, including visitors, who expect to remain in the team for a short time; replacements, who take the place of former team members; and transfers, who recently belonged to a similar group (Arrow &
McGrath, 1995). Each type of membership change, in turn, may elicit a differential impact on team performance (Levine, Moreland & Choi, 2001). Research on these specific types of membership change and its consequences for team performance is, however, scarce (Levine et al., 2001).
This paper will focus on the impact of one specific type of membership change, i.e. transfers, on team performance. Transfers are members of a team who recently belonged to a similar team and who have expertise on the team task (Arrow & McGrath, 1995; Levine et al., 2001). In line with the definition of membership change, transfers may either enter or depart a team. As it is expected that he departure of transfers elicits other effects for team performance than the entry of transfers into a team, both aspects of change are taken into account.
In this study it is hypothesized that the extent to which entering or departing transfers affect performance is dependent on certain individual and team characteristics.
An individual characteristic on which transfers are likely to differ, is their centrality within a team (Arrow & McGrath, 1995; Humphrey, Morgeson & Mannor, 2009). A team characteristic which may affect performance is the timing of the transfer(s) in relation to team development, i.e. the stage of development of a team such as forming, established or adjourning teams.
The aim of this study is to explore the impact of the entrance and departure of transfers on performance for, a) more and less central transfers, and b) contingent on the timing of change in relation to team development. Hypotheses are tested for transfers in 143 soccer teams in eight European top leagues in the season of 2009-2010.
THEORETICAL BACKGROUND
Membership changes within teams have been studied for over half a century (Steers &
Mowday, 1980). Many studies have tried to figure out which antecedents cause membership change, such as demographic composition and internal conflict (Ilgen, Hollenbeck, Johnson & Jundt, 2005). Stated differently, these studies have focused on membership change as an output variable. Other studies have focused on the effects of membership change on subsequent team processes and performance. Here, membership change has been taken as an input variable while its subsequent outcomes, i.e. team process and performance, are explored (Ilgen et al., 2005). In this study this last strand of research is followed.
The input variable in this study concerns a specific type of membership change, i.e. transfers. It is proposed that transfers - in contrast to other types of membership changes– elicit (partly) differential effects on team performance (Levine et al., 2001).
There are a couple of reasons why this differential effect is expected. First of all, a significant difference between other types of membership changes and transfers concerns their expertise on the team task (Arrow & McGrath, 1995; Levine et al., 2001). Transfers are already acquainted with the team task and are therefore expected to be more easily incorporated within the team. Transfers require few energy from other team members, as socialization and internalization of the team task are only marginally necessary (Levine et al., 2001). Subsequent team performance may consequently be less negatively affected by changes in team composition when it concerns a transfer rather than other types of
membership change - all other things being equal. Second, this study focuses on top teams and (sub)top transfers. Transfers are therefore expected to be indeed talented and well-educated (Feess & Mühlheußer, 2002). In addition, transfers and teams are already knowledgeable about each others tactics and playing style, due to exacerbated media attention (Franck & Nüesch, 2007). Taken together, both aspects (talent/education and knowledge) will make integration and socialization of transfers into a team more easy (Levine et al., 2001). Finally, for transfers the regularity and predictability of change is relatively high due to specific guidelines on transfer periods (FIFA, 2010); there are two legally regulated periods when transfers are allowed. Thus, teams can to a large extent
anticipate change and therefore prevent the degree of negative effects of transfers on performance while enhancing the positive effects (Arrow & McGrath, 1993; 1995).
To explore the impact of transfers on team performance, we turn to sports. Sports has proven to be an ideal context within which to perform research on organizational phenomena (Wolfe et al., 2005). Especially, as in sports things like internal cooperation, external competitiveness, war on talent, good management and strategic human resource management are as much of relevance as it is in organizations (Brady, Bolchover, &
Sturgess, 2008: 60; Keidel, 1984). The main advantages of sport studies are : 1) availability of data due to the frequency and regularity of athletic events (Wolfe et al., 2005); 2) accuracy and objectivity of data on a broad scope of variables, including objective measures for individual and team performance; 3) research in sports resemble laboratory studies, as hypotheses can be tested in relatively controlled fields (Wolfe et al., 2005; Brady et al., 2008). Moreover, especially in (top) sports teams, player transfers are fairly common (see FIFA, 2010 for soccer transfers). This study, then, will test
hypotheses using sports data from eight top soccer leagues in Europe.
Membership change and performance
Previous research on membership change has in general found a negative or an attenuated negative relationship between membership change and performance (Argote, 1995; Shaw, Gupta & Delery, 2005b; Trow, 1960; Van der Vegt & Bunderson, 2009). As has been outlined in the previous paragraph, transfers are considered to be a specific type of membership change; transfers are well-educated, acquainted with the team task and their entrance or departure can be anticipated due to regulation. Therefore changes within teams caused by transfers, either entering or departing, are expected to be less disruptive and probably under some conditions even non-disruptive for team performance. In specific, the extent of the (non)-disruptive effects is expected to be dependent on certain individual and team characteristics (cf. Humphrey et al. 2009). In the following
paragraphs these two characteristics will be discussed.
Entering transfers
In general, the entrance of a transfer into a team is expected to be disruptive. A transfer can be integrated and socialized within the team quickly due to his knowledge of the
team task, his training history and experience (Levine et al., 2001). Also the team can anticipate the transfer’s entrance into the team due to the legally regulated transfer periods and media attention (FIFA, 2010; Feess & Mühlheußer, 2002). However, still a team has to integrate and socialize a new team member – which requires energy of the other team members and distracts attention from the common team goal (Levine et al., 2001). So, the integration and socialization period may take a shorter time for transfers, but can be expected to (still) affect performance for a while.
H1: The entry of transfers in a team is negatively related to overall team performance.
Departing transfers
Also the departure of a transfer out of a team is expected to be slightly disruptive. A player who has been in the team for a while has build up shared experience with the other team members and may possess tacit team knowledge (i.e. playing tactics; Berman, Down & Hill, 2002). Part of this experience and knowledge will be lost by the departure of the transfer. Therefore team performance is expected to be negatively affected by the departure of a transfer.
H2: The departure of transfers from a team is negatively related to overall team performance.
Transfer characteristics: centrality
An individual characteristic of transfers which is considered to be of importance for performance concerns centrality (Arrow & McGrath, 1995). According to Arrow and McGratch (1993), centrality is one dimension along which members in teams differ.
Some team members are more important or central to the team than other members. This implies that changes involving players who occupy central roles in the team should have stronger effects on group performance, than changes involving players who occupy peripheral roles (Arrow & McGrath, 1993).
In teams there are often various roles defined. Each role in the team requires a certain set of skills. For soccer this may include skills such as acceleration, dribbling, shot power, agility and passing (Carmichael, Thomas, & Ward, 2000). However, as
soccer is a team sport in which all roles are moderately and sequentially interconnected (Katz, 2001: 64) and for which the outcome of a play depends on defensive skills as well as on attacking skills (Carmichael, Thomas, & Ward, 2000; Franck & Nüesch, 2008: 9;
2009) there is not one role which in essence is more central than another role. Therefore, in this paper the focus will be on central players in general, independent of the specific role differentiation and the specific role assigned to a player in the team. Centrality, then, is determined by the extent to which a player is valuable to a team.
Transfer characteristics: timing of change
In addition to the individual characteristics of transfers also team characteristics are of relevance for final team performance. Here the timing of a transfer in relation to team development is studied. Transfers within just-forming teams are proposed to have a differential impact on subsequent performance than changes in well-established or disbanding teams (Arrow & McGrath, 1993; 1995).
At the start of the soccer season the composition of many teams has changed due to transfers. The team members of the newly formed teams at the start of the soccer season have to get acquainted to each other, set themselves a common goal and start to build up shared field experience. Team transfers at the start of the season are not expected to be fairly disruptive, because teams are still deliberating the common team goal, targets and are trying to get to know each other (cf. Arrow & McGrath, 1993; Levine et al., 2001).
After half a year, i.e. at the temporal midpoint, a second possibility to transfer players is offered. At this moment it can be expected that the team members are
acquainted to each other, know the teams playing tactic and have set a common team goal, i.e. the team is established. Transfers during the central midpoint are therefore expected to be fairly disruptive for team process and performance (Gersick, 1988).
Entering transfers and timing of change
For entering transfers, in general, a disruptive effect is expected for performance. This is caused by the distraction of energy from the common team goal towards the socialization of the transfer (Levine et al., 2001). At the start of the season this disruptive effect on performance will only be marginal, as all players in the team are still searching for the
common team goal and are part of the socialization process. The entrance of a transfer will therefore not worsen team performance.
In contrast, at the central midpoint the entry of a transfer is expected to be fairly disruptive. At that moment teams have already been established; a common team goal is set and team members are acquainted to each other. The entrance of a new player may at that moment distract energy from the common team goal, whereas also energy is required for socialization.
H1a: The entry of transfers in a team at the start of the season will be less negatively related to overall team performance than will the entry of transfers during the season.
Departing transfers and timing of change
The departure of a transfer is disruptive due to a loss of shared experience and tacit knowledge (Berman et al., 2002). It is expected that at the start of the season, when a partly new team has been formed, shared field experience and tacit knowledge between team members is limited. However at the central midpoint, when team members have been playing together in the team for a while, shared experience and tacit knowledge has indeed been build up. The departure of a transfer at the start of the season therefore is expected to be less disruptive than the departure of a transfer at the central midpoint.
H2a: The departure of transfers from a team at the start of the season will be less negatively related to overall team performance than will the departure of transfers during the season.
Centrality and timing combined
The propositions on centrality and timing of change can be combined to predict the impact of transfers on team performance. In the first paragraph on centrality we proposed in line with Arrow & McGrath (1993) that transfers will have less consequences for performance when peripheral rather than central members are involved (Arrow &
McGrath, 1993). In the second paragraph we proposed that the timing of change in relation to team development mattered. It was hypothesized that changes at the start of the soccer season, i.e. in newly formed teams, were less disruptive than changes at the
temporal midpoint, i.e. in established teams. Combining these propositions for entering and departing transfers at the two legally regulated transfer periods (t1 and t2) leads to the following hypotheses.
Entering transfers: centrality and timing [t1]
In general, the entrance of a transfer is expected to elicit a negative effect on performance.
However, at the start of the season when teams are just forming, the entrance of new players is expected to be only marginally disruptive. The more central an entering transfer is at the start of the season, i.e. the better the playing quality of a player is
expected to be, the less disruptive the entrance of this transfer will be. In other words, the more central an entering player was in his previous team the more this player is expected to be able to (also) contribute to his new team. Therefore, for entering players, the more central the player was the less disruptive the entrance of this new player will be for team performance.
H3: At the start of the season, the entry of central transfers in a team will be less negatively related to overall team performance than will the entry of non-central transfers in a team
Departing transfers: centrality and timing [t1]
In general, also the departure of a transfer is expected to elicit a negative effect on performance. This negative effect is expected to be less strong at the start of the season when teams are just forming than at the central midpoint. However, especially the departure of central transfers, i.e. those players who posses the most shared field experience and tacit knowledge, is expected to be disruptive. The more central a departing player is in the team the more disruptive his departure will be for subsequent performance of that team – even at the start of the season.
H4: At the start of the season, the departure of central transfers in a team will be slightly more negatively related to overall team performance than will the departure of non- central transfers in a team.
Entering transfers: centrality and timing [t2]
After half a year, i.e. the temporal midpoint, teams can rearrange. According to Katz (2001: 61) the temporal midpoint is “a time of special opportunity in teams”, as it is a moment to revise strategy, reformulate the common team goal and to implement change.
At the temporal midpoint teams can look back at what they have achieved and estimate what will be needed to attain the common team goal. In specific, the challenge during temporal midpoint is to transfer players so as to adjust the match between the team goal and the resources (abilities of players) available (Gersick, 1988).
Transfers who enter the team can be expected to be valuable (extra) resources.
These transfers might be needed especially if there is a shortage of certain skills in the team. The entry of players during the season therefore is expected to be only slightly negatively related to performance. Moreover, the entrance of in specific central players will be less negatively or even positively related to performance – especially as the new player is a real contribution to the team.
H5: At the temporal midpoint, the entry of central transfers in a team will be less
negatively (or even positively) related to overall team performance than will the entry of non-central transfers in a team.
Departing transfers: centrality and timing [t2]
The departure of a central player is expected to be disruptive. Especially, as it concerns a player who has build up shared experience and tacit knowledge with the other (central) players; with the departure of this central transfer also part of this knowledge and experience is lost. Therefore a negative relationship between the departure of central transfers and performance is expected.
H6a: At the temporal midpoint, the departure of central transfers in a team will be negatively related to overall team performance
In contrast to the departure of central players, those players who have been non-central in the team in the previous half year, i.e. those players who have barely played due to the fact that their resources and skills were not needed or not enough developed, have not considerably contributed to the overall team performance. Still, non-central players may have been valuable to the team, especially if they possess some skills and insights which can be used and applied during the training (cf. Franck & Nüesch, 2008). However, the non-central players did not, or only marginally, contribute to the overall team
performance on the field in the previous half year (i.e. field performance). Therefore, it is expected that the departure of non-central players will only have a marginal negative impact on performance. Moreover, if those players who are transferred have not been fielded at all this relationship between turnover and performance may even be positive
H6b: At the temporal midpoint, the departure of non-central transfers in a team will be slightly negatively (or even positively) related to overall team performance.
METHODS
The impact of transfers is hypothesized to be dependent on the centrality of the players involved and contingent on the timing of change in relation to team development.
Hypotheses are tested using data from teams and players in eight European top soccer leagues during the 2009-2010 soccer season.
Setting
Soccer is a team sports, in which two teams of eleven players including a goalkeeper, play against each other, using a spherical ball. During a match of 90 minutes a maximum of three players may be substituted. Matches are played in a country specific league system, often comprising several divisions. Most commonly, a double round-robin tournament system is used, whereby each team plays all other teams in its league once at home and once away.
During a soccer season there are, in line with the Federation Internationale de Soccer Association (FIFA) guidelines, two periods for which it is possible to transfer
players to another team (FIFA, 2010). The first period begins after the completion of the season and normally ends before the start of the new season. This period may not exceed twelve weeks. The second period normally occurs in the middle of the season and may not exceed four weeks. Only those players who are on loan at another club are eligible to sign a contract at any other moment than the specified periods above (FIFA, 2010).
Sample and Data Collection
To test hypotheses, data from eight top division soccer leagues in Europe for the season 2009-2010 were used, including FA Premier League (England), La Liga (Spain), Serie A (Italy), Bundesliga (Germany), Ligue 1 (France), Eredivisie (The Netherlands), Scottish Premier League (Scotland) and Jupiler Pro League (Belgium). The time period was restricted to one seasons, as full data for all eight leagues were only available for this period. Data were retrieved from www.elfvoetbal.nl and www.transfermarkt.de , which are comprehensive websites providing data on league, team, game and individual level.
In total the sample consists of 143 different teams.
Measures
Dependent variable
Winning percentage 2009-2010. Overall team performance is measured as the percentage of total matches won divided by the total number of matches played by a team. Winning percentage varies between 81% (most matches won) and 13% (least matches won).
Independent variables
Entering/departing transfers. Two measures for transfers were derived from the soccer data: transfers of members who departed and of members who entered the team. The variable is measured as the rate of the total number of team members who respectively left or entered the team divided by the total group size at the start of the season. This definition of entering and departing transfers is equal to the definition of team turnover in previous studies on membership change (McElroy, Morrow & Rude, 2001; Shaw et al., 2005b; Van der Vegt & Bunderson, 2009). In soccer, a transfer is considered to be voluntary as players are free to negotiate their entrance into another team as soon as their
contract has expired (see Feess & Mühlheußer, 2002). The transfer rates are measured at two time points: at the start of the season and at the temporal midpoint.
Central and non-central transfers. The minutes of playtime a transfer has been in the field is used as a proxy for the extent to which a transfer was central in the team, i.e. the more minutes of playtime a transfer received, the more likely it is that the transfer was important to the team . The average of minutes of playtime per transfer per match is used as cut off point; transfers who received more minutes than the average minutes of
playtime are classified as central, whereas those transfers who received less minutes than average are classified as non-central. The rate of central and non-central transfers in a team is measured by the formula: total number of (non)-central transfers who leave or enter a team divided by the total group size at the start of the season.
At the start of the season (t1), the cut off point to classify central and non-central transfers was calculated over the amount of play minutes received by all players in the previous soccer season, i.e. 2008-2009. At the central mid point of the soccer season (t2) the cut off point to classify the central and non-central transfers was calculated over the play minutes of those team members during the first half of the soccer season who are transferred at the central mid point1.
Control variables
Many other factors may affect the final league standing of a team. In order to eliminate alternative explanations for the observed findings due to these other factors, several variables were controlled for:
Team size at start. The variable team size controls for potential group size effects and is based on the roster size of the team at the start of the soccer season.
Direction of change. Dummy variables were created to identify whether a team has, declined, grown or remained stable over the year. Team size at the start of the season and at the end of the season were compared to identify the direction of change.
1As a consequence of missing values at the individual player level for the minutes of playtime received, the aggregated team measure which indicated the number of core and non-core players transferred differs from the total number of transfers.
Coach turnover. A dummy variable was created to identify whether or not a coach was replaced during the soccer season, coded as zero (no replacement) and one (coach replacement).
Coach experience with team. The variable coaching experience with the team measures how long the coach has been with the team at the end of the season in years, i.e. shared experience of the coach with the team. If a coach was replaced during the season the mean of coaching years with the team was calculated.
Winning percentage 2008-2009. Overall team performance in the previous season is used as a control variable for predicting the performance within the current season. In line with the construction of the dependent variable, the percentage of matches won divided by all matches played in the league in the 2008-2009 soccer season was calculated.
Analysis
The data was hierarchically structured, i.e. sets of central and non-central tranfers in teams within leagues, which introduced the concern that there might be dependence in the data. Testing for differences in variance between teams and leagues did not confirm that there was dependence in the data. Therefore it was sufficient to apply a regression analyses instead of multilevel analyses.
Descriptive statistics, correlation analyses as well as hierarchical multiple regression analyses were performed to identify the predictors of final league standing, using SPSS 17.0 (Chicago, Il., USA). For all tests a p-value of five percent or less was considered significant.
RESULTS
Descriptives
An overview of the descriptive statistics of all study variables is shown in table 1 and the outcomes of correlation analyses are shown in table 2. The mean transfer rate of entering transfers at the start of the season (t1) was around 19 per cent, with considerable variation in magnitude across teams. The transfer rate of entering transfers at the mid point of the season (t2) was about 7 per cent with less variability across teams. For departing transfers these percentages were for the start and the mid point of the soccer season respectively
about 8 and 4 per cent. Again the variability in magnitude of team turnover was larger at t1 than for t2. A comparison of means provided evidence that the mean transfer rate was significantly larger at the start of the season than at the central mid point, i.e. indicating that more transfers entered or departed the team at the start of the season than at the central midpoint (mean differences entering transfers=.12; s.d.=.01; p<.01: mean differences departing transfers=.05; s.d.=.01; p<.01).
In the 2009-2010 soccer season, the majority of teams, 81.8% (n=117), grew due to an inflow of transfers which was higher than the outflow of transfers. Further, 13.3%
(n=19) of the teams declined as more transfers left the team than transfers who entered the team and 4.9% (n=7) of the teams remained stable as a consequence of replacement.
A comparison of means showed that direction of change in the size of the team significantly affected performance (F(2, 140)=11.31; p<.01). Bonferroni post hoc tests provided evidence that growing teams performed worse than declining teams (mean difference 17.34; s.d.=3.66; p<.01). Stable teams did not perform significantly worse or better than declining or growing teams.
An examination of table 2 shows that there are no high correlations between the independent variables, so mulitcollinearity is not a problem. Negative correlations were found between transfer rate and performance, indicating that, by and large, the transfer rate of entering transfers negatively affects final team performance. The entrance of non- central transfers was slightly more negatively related to performance than the entrance of central transfers. These findings are in line with the anticipated directions based on the hypotheses above (H3 & H5). For the rate of departing transfers most of the correlations are in the opposite direction of what was hypothesized, though these relationships were not significant. Only the rate for departing transfers at the central midpoint and winning percentage a significant and positive correlation was found. This finding provides preliminary evidence for hypothesis H6b, indicating that the departure of non-central transfers increased team performance.
Table 1 Descriptive statistics
Measures Min. Max. Mean
(%)
s.d.
1. Entering transfers (t1) 0 85.71 18.88 1.12
2. Central entering transfera(t1) 0 42.86 6.00 .48
3. Non central entering transfera(t1) 0 36.84 5.84 .60
4. Entering transfers (t2) 0 25.00 6.53 .51
5. Central entering transfera(t2) 0 4.00 .05 .04
6. Non central entering transfera(t2) 0 9.09 .77 .15
7. Departing transfers (t1) 0 34.29 8.25 .62
8. Central departing transfera(t1) 0 8.33 .68 .13
9. Non central departing transfera(t1) 0 20.00 1.89 .25
10. Departing transfers (t2) 0 21.74 3.52 .35
11. Central departing transfera(t2) 0 7.69 .24 .09
12. Non central departing transfera(t2) 0 16.67 1.11 .21
13. Coach turnover (52.4)
14. Coach experience with team 1 36 13.7 .65
15. Team size at start 14 39 27.43 .39
16. Direction of change
Growth (81.8)
Decline (13.3)
Stability (4.9)
17. Winning percentage 2008-2009 20.59 73.68 40.53 1.25
18. Winning percentage 2009-2010 13.16 81.58 37.35 1.32
aThe percentage of (non) central transfers in a team is based on minutes of playtime N = 143; t1 = start of season; t2 = at the mid point of the soccer season.
Multivariate regression analyses
Regression analyses were performed to test hypotheses concerning the effect of central and non-central entering and departing transfers at two points in time on overall
performance. The outcomes of the models are shown in table 3 for the start of the season (t1) and table 4 for the central midpoint (t2), with the coefficients and standard errors in parentheses. In order to make interpretation easier and to be able to assess the importance of a predictor, also the standardized beta coefficients are shown. The beta coefficients are measured in standard deviation units and therefore directly comparable.
Table 2 Zero-order correlation matrix
Measures 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1. Entering transfers (t1) -
2. Central entering transfera(t1) .58** -
3. Non central entering transfera(t1) .81** .23** -
4. Entering transfers (t2) .26** .05 .26** -
5. Central entering transfera(t2) -.07 -.05 -.01 -.02 -
6. Non central entering transfera(t2) .05 .02 .06 .44** -.05 -
7. Departing transfers (t1) .25** .25** .08 .03 .03 -.11 -
8. Central departing transfera(t1) .04 .09 -.00 -.18* .08 -.10 .38** - 9. Non central departing transfer a
(t1)
.03 .19* -.08 .05 .06 -.00 .60** .17* -
10. Departing transfer (t2) .04 .11 -.04 .16 -.05 -.03 .15 -.003 .13 -
11. Central departing transfera(t2) -.04 -.01 -.01 -.05 -.03 -.06 .002 -.05 .09 .20* - 12. Non central departing transfera
(t2)
.01 .05 .02 .14 -.05 -.07 .01 -.08 .07 .67** .12 -
13. Coach turnover .12 .02 .07 .23** -.13 .12 .05 -.20* -.02 .04 -.13 -.07 -
14. Coach experience with team .01 -.04 .04 -.20* .03 -.08 -.09 .09 .03 .13 .001 .02 -.13 -
15. Team size at start -.56** -.32** -.51** -.30** -.03 -.11 .23** .13 .19* -.01 -.04 -.04 -.05 .09 -
16. Winning percentage 2008-2009 -.16 .00 -.22** -.25** -.09 -.15 .05 .10 -.03 .37** .00 .26** -.10 .20* .34** - 17. Winning percentage 2009-2010 -.11 .10 -.22** -.34** -.19* -.19* .15 .20* .11 .31** -.03 .23** -.23** .29** .29** .69** -
aThe percentage of central and non/central players in a team is based on minutes of playtime
*p<.05 **p<.01; N=143
t1 = start of season; t2 = at the mid point of the soccer season
Table 3 Hierarchical regression analyses with team turnover at the start of the season and winning percentage 2009-2010
aReference category is growth of a team
†p<.10; *p<.05 **p<.01; N = 143
Model 1 Model 2: Model 3: Model 4:
B (S.E.)
Beta B
S.E.)
Beta B
(S.E.)
Beta B
(S.E.)
Beta Winning percentage 2008-2009 .74
(.09) .64** .71
(.09) .61** .74
(.09) .64** .71
(.09) .61**
Team size at start -.16
(.29)
-.04 -.31 (.32)
-.08 -.13
(.34)
-.03 -.14
(.32)
-.04
Coach turnover -5.51
(2.15) -.17* -5.25
(2.16) -.16* -5.62
(2.18) -.17* -5.34
(2.23) -.16* Coach experience with team .23
(.15)
.11 .24 (.15)
.11 .27 (.15)
.13† .27 (.15)
.12† Direction of changea
Decrease 3.42
(3.61)
.08 -1.94
(4.40)
-.04 2.26
(4.13)
.05 3.51
(3.70)
.08
Stable -.53
(5.93)
-.01 -5.59 (6.32)
-.06 -2.05 (6.29)
-.02 -1.94 (6.05)
-.02
Entering transfers (overall) -.25
(.16)
-.18
Departing transfers (overall) .44
(.20) .23*
Entering transfers (t1) .04
(.18)
.03
Departing transfers (t1) .18
(.23)
.08
Central entering transfers (t1) .30
(.27)
.09
Non central entering transfers (t1) -.14
(.21)
-.05
Central departing transfers (t1) .34
(.67)
.04
Non central departing transfers (t1) .07
(.46)
.01
Constant 12.59
(8.24
19.00 (10.76)
8.71 (10.44)
11.04 (9.67)
∆ R2 compared to model 1
.50 117
.02† .52
117
.01 .50
117
.01 .50 Adjusted R2 117
N
Table 4 Hierarchical regression analyses with team turnover at the central midpoint and winning percentage 2009-2010
Model 5 Model 6
B (S.E.)
Beta B
(S.E.)
Beta Winning percentage 2008-2009 .67
(.09)
.57** .70 (.09)
.61**
Team size at start -.06
(.29)
-.02 -.10
(.29)
-.03
Coach turnover -5.27
(2.16)
-.06* -5.49 (2.19)
-.17*
Coach experience with team .15
(.15)
.07 .24 (.15)
.11
Direction of changea
Decrease .98
(3.71)
.02 2.94 (3.65)
.07
Stable -2.37
(5.89)
-.03 -2.41 (6.89)
-.03
Entering transfer (t2) -.50
(.23)
-.16*
Departing transfer (t2) .47
(.29)
.13
Central entering transfer (t2) 1.33
(3.80)
.03
Non central entering transfer (t2) -.55
(.72)
-.05
Central departing transfer (t2) -1.20
(1.23)
-.07
Non central departing transfer (t2) .47
(.43)
.08
Constant 15.20
(8.52)
11.95 (8.39)
∆ R2compared to model 1 .02†
.52 117
.01 .50
Adjusted R2 117
N
aReference category is growth of a team
†p<.10; *p<.05 **p<.01; N = 143
In the first model, as shown in table 3, only the control variables were taken into account.
Together the control variables explained about 50 per cent of the total variance in winning percentage. The addition of the study variables on turnover of central and non- central players in the subsequent models did not significantly change the explained variance in winning percentage, except for model 2 and 5. These findings indicate that model 1 is not significantly improved by the addition of the specific study variables, i.e.
these variables did not explain additional variability in winning percentage over and above the control variables.
In light of the explained variance a description of the control variables is warranted. First the years of coach experience with the team was positively related to winning percentage, and tended to significance (p<.10) in two of the six models. This finding suggested that the longer the coach is with a team the better the performance of the team would be. Second, the measure of coach team turnover was negatively related to performance. Coach team turnover is a dummy variable, indicating whether the head coach of a team had been replaced during the 2009-2010 soccer season. The negative relationship found in all models, indicates that turnover of coaches is disruptive for a team and has a strong negative effect on overall performance. Third, performance in the previous soccer season was highly predictive of performance in the current season.
Winning percentage in the previous year was accountable for a large part of the explained variance in winning percentage of the soccer season 2009-2010. The control variables measuring, team size at the start of the season and direction of change, i.e. growth, decrease or stability of the team, were not significant in either model.
Now we turn to a discussion of the results in the other models in relation to our hypotheses. The first set of hypotheses suggested that the entrance and departure of transfers would be disruptive, i.e. negatively related to performance (H1, H2). As shown in model 2, the entrance of transfers was indeed negatively related to performance, though the parameter was not significant. The departure of transfers was positively and significantly related to performance. This finding suggests, in contrast to hypotheses, that the departure of players over the season 2009-2010 resulted in increased team
performance.
The second set of hypotheses focused on the effect of team turnover at the legally regulated transfer periods in a soccer season. It was hypothesized that at the start of the season the entrance and departure of transfers would be less negatively related to
performance than the entrance and departure of transfers at the central midpoint (table 3, model 3; table 4, model 5). Only for entering transfers at the central midpoint a
significant beta coefficient was found (b=-.5; p<.05), indicating that the inflow of transfers was related to a lower winning percentage. This finding supports hypothesis (H1a) that the entry of transfers at the start of the season is less disruptive than the entry of transfers at the central midpoint.
Then, it was tested whether there were differential effects of central and non- central entering and departing transfers on performance at the two legally regulated transfer moments. At the start of the season, the inflow of central transfers was
hypothesized to be more negatively related to performance than the inflow of non-central transfers. For departing transfers it was hypothesized that the outflow of central transfers would be more negatively related to performance than the outflow of non-central players (H3 – H6). As shown in model 3, neither the entry of central and non-central transfers nor the departure of central and non-central transfers was significantly related to performance.
However, the sign of the parameters for entering transfers were in the correct direction;
non-central transfers had a negative impact on performance, while central transfers had a less negative, i.e. even a slightly positive effect, on performance.
DISCUSSION
This paper aimed to investigate the effects of the entry and departure of central and non- central transfers on overall team performance, measured for two legally regulated transfer periods during the soccer season. It was argued that the entry and departure of transfers would be negatively related to performance and that this effect would be stronger at the central midpoint. The departure of central transfers would be more negatively for team performance than the departure of non-central transfers (loss of shared experience and tacit knowledge), whereas the entrance of central transfers would be slightly less negatively related to performance than the entrance of non-central transfers (valuable extra resources). Results of the present study do not confirm that the characteristic of centrality mattered for team performance, independent of the timing of change. However, there is some preliminary evidence that the timing of change in relation to team
development affects performance; well-established teams are less able to accommodate entering transfers than teams which are just formed (cft Arrow & McGrath, 1995).
First of all the relationship between overall transfer rate and performance was assessed. In contrast to hypothesis, the entrance of transfers was not (negatively) related to performance. A possible explanation may be that transfers, i.e. players who are already acquainted to the team task, are indeed easily accommodated within the team and
therefore have no or less impact on team functioning and subsequent performance.
Moreover, it may be that the decision to transfer a player into the team is a well thought- out decision; i.e. players are only transferred if there is a need of the skills and knowledge of that specific player within the team (Katz, 2001). Entering transfers in this sense are as a consequence more easily accommodated within the team and may have a more positive effect on performance or at least do not harm performance. Also the hypothesis
concerning the overall rate of departing transfers was not confirmed. Instead of a negative relationship with performance, the departure of transfers turned out to be positively related to final team performance. Such a relationship between departure of team members and performance has, to the best of my knowledge, not been found before in literature. Though, previous research has found preliminary evidence for an inverted u- relationships, implying that changes in teams are beneficial up to a certain point (Abelson
& Baysinger, 1984; Glebbeek & Bax, 2004)2. Also within soccer the departure of a subset of players may be beneficial for the performance of the entire team, -getting rid of
dysfunctional team members and providing momentum to remaining team members-.
However, more research is needed to find out if the positive relationship found in this study for transfers indeed exists or to find out that it concerns a spurious relationship.
Next, the focus in this study was on the timing of change in relation to team development. There is some preliminary evidence that the timing of change in relation to the development stage of change matters for performance; teams which are first forming are found to be better able to accommodate new players than well-established teams (cf.
Arrow & McGrath, 1995). In this study no effects on performance were found at the start of the season independent of those who departed or entered the team, whereas at the temporal midpoint in specific the entrance of new transfers significantly and negatively affected team performance. In line with Moreland and Levine (1988), the findings in this study suggest that during early stages of team or group development, teams are more open to entering transfers and are better able to accommodate them. At the start of the season, differences in playing tactics between entering transfers and the other players in the team have less or no impact on overall performance as there are not yet any
prescribed norms and regulations on playing tactics. In later stages of team development
2Also in this study we tested for an inverted u-relationship for departing team members. However, there was no evidence for such a relationship in this study.
teams have formed such specific norms and guidelines and are therefore more likely to ask for conformity of entering transfers. Even if those transfers try to confirm to the new playing tactic as requested by the team and the coach, the entrance of transfers at this stage is disruptive (Moreland & Levine, 1988). Especially, for the specific type of membership change of transfers this (disruptive) effect might even be stronger as they have internalized a certain playing tactic and style due to training and education at another soccer club. Transfers have to adjust to other players in the squad who have already been trained at the club for at least half a year. This may lead to difficulties in adjustment of new as well as established players in the team and therefore cause negative team outcomes.
The difficulties in adjustment to team members have not been found for departing transfers. Apparently the timing of change in relation to transfers who depart does not require as much adjustment of the entire team as new players do. In the squad there are members available who can be substituted in for the departing transfers. These substitutes are already trained within the team and have already been adjusted to the competition team’s playing style and tactic. As a consequence losses in outcome are prevented (compare to Franck and Nüesch, 2008, who emphasize the importance of talent disparity in the team).
In addition to the timing of change, the focus was also on the extent to which more ore less central transfers on the two points in time affects performance. The entrance and departure of central and non-central transfers in a team was not
differentially related to performance in all tested models, independent of the timing of change. All hypotheses on central and non central transfers are therefore not confirmed.
These findings are in contrast to those of Humphrey et al. (2009) who studied the differential impact of characteristics of a set of role holders in major league baseball teams. Humphrey et al. (2009) found that role holder characteristics were more strongly related to performance than were the non-central role holder characteristics. Even though Humphrey et al. did not study the entry and departure rate of transfers, they did find differences in the importance of sets of players in a team. A potential explanation for the disconfirmation of hypotheses in this study is that the measure used to identify groups of central and non-central transfers was inadequate. It might be that minutes of playtime
received refers to the extent to which a player is better than other available players in the squad at the same position and therefore not per se distinguishes between more and less important players within the competition team. In addition it may also be that the overall quality of the competition team matters more for team performance than the centrality of a group of leaving or entering transfers in the team. Indeed, Fank and Nüesch (2009) found that player homogeneity in terms of talent (i.e. as based on expert evaluations and an objective performance index) within the competition team positively affects
performance (i.e. the team which is playing during the competition stage), whereas heterogeneity in the entire squad (all players in the team, training and competition team) also contributed positively to final league standing. In this light it is warranted to assess other measures of central and less central transfers, while taking into account the quality or talent of the individual transfer in relation to the whole team (cf. Trow, 1960) as well as the level at which transfers takes place (i.e. in the competition team or in the squad).
Finally a discussion of the outcomes of control variables is justified. For the control variable coach shared team experience a positive relationship was found with performance; the more years a coach has been with the team the better for the
performance of that team. This finding can be explained in two ways. On the one hand it implies that it takes some time for players to become familiar to the coach and vice versa (cf. organizational learning; Berman, Down & Hill, 2002). On the other hand, the longer the coach is with a team the better the coach has been able to (re)arrange a squad of players which fit his style, i.e. fit between human resource staffing and strategy. The longer the coach is with the team the more time the coach has had to make a fit and therefore the better the performance (Wright, Smart & McMahan, 1995)3.
The control variable coach turnover was negatively related to performance. This finding partly confirms previous literature on coach dismissal. From one point of view, part of the literature provides evidence that coach turnover is disruptive and negatively affects performance (e.g. Audas, 2007; Dios Tena & Forrest, 2007). Coach turnover is disruptive due to the fact that the squad has to get acquainted to the style of the new coach as well as that the coach has to rearrange the team in order to make the fit between
3Analyses confirmed that the longer a coach is with a team the better team performance. In contrast to Berman et al. (2002) we did not find an attenuating relationship, i.e. lower performance levels for teams with higher levels of shared coach experience (see table 1A in Appendix A).
desired staffing levels and strategy (cf. Berman et al., 2002; Wright et al., 1995). Also Audas (2007) found a strong negative effect on performance while controlling for the effects of regression to the mean – an aspect according to Audas which has been neglected in previous research. From another point of view, in literature on coach dismissal positive effects for team performance are found. Balduck, Prinzie and Buelens (2010) point out that team performance is positively affected by coach turnover only to the extent that the new coach is able to improve the team quality. Dios Tena and Forrest (2007) found that performance was positively affected by mid-season coach turnover for home-matches only - pointing out the importance of crowd support (Dios Tena and Forrest (2007). Other studies found that coach turnover led to short-lived improvements in performance where after the performance of the team returned to its previous level (McTeer & White, 1995). In sum, there is evidence that coach turnover can be positively related to performance under some conditions, i.e. depending on the ability of the coach to adjust team quality and the level of analyses (per match or over all). Here it is found that, when controlling for the shared experience (i.e. the time a coach has been with the team and should have been able to adjust the team), coach dismissal is negatively related to final seasonal team performance.
Strengths
One of the main strengths of this study on transfers is that it is performed in a non- experimentally controlled research setting (cf. McGrath 1991). As a consequence the processes and structures observed in soccer teams have great resemblance with processes an structures in other real-life (organizational) teams. Moreover, this study gained from the opportunity to use accurate and objective performance data on a great variety of variables ranging from individual to team and league level (Wolfe et al., 2005).
Limitations
There are also some methodological limitations of this study that merit consideration.
First of all, in contrast to previous studies, a positive relationship between the rate of departing transfers and performance has been found. One possible explanation for this finding is reversed causality, i.e. that overall team performance affects the departure of transfers. Teams and the talented players within those teams which have performed well
in the previous year are magnified within the media (based on press citations; Frank &
Nüesch, 2009) and as a consequence may become more visible to recruiters and coaches within the soccer scene. Accordingly, popular players of those well performing teams are more likely to be qualified for transfer. However, only the availability of longitudinal data would provide for the possibility to test for causality. If the reversed causality theory holds, turnover in 2009-2010 should be more strongly related to performance in the year before, than the other way around.
Second, this study is based on a cross-sectional dataset, gathering data without regard to small differences in time of measurement. However, to confirm analyses and to provide for more robust findings, longitudinal data as well as data on teams in other sports has to be examined. Sports data are in general freely available and can for instance be retrieved from websites of sports unions and teams. Then, what is needed is a more concerted and systematic effort in order to further theory and to find empirical evidence in the field of sports studies (Wolfe et al., 2005).
Moreover, in this study the focus was on the number of central and non-central transfers calculated on the basis of minutes of playtime received by a transfer – as this was considered to be the most objective measure. However, other measures to calculate the centrality of transfers in a team are also conceivable, such as the number of matches started in base or the total number of matches played. As in soccer also a division of roles consists, i.e. attackers, midfielders and defense, the centrality of transfers could as well be determined on the basis of their role in the field (cf. Humphrey et al., 2009). Also, other dependent measures of performance could have been used such as final league standing or goal difference (Franck & Nüesch, 2008; 2009). Future research should apply other conceptualizations to determine the number of central and non-central transfers as well as testing these new conceptualizations in relation to a variety of outcome measures.
Finally the generalizability of the findings in this research to other contexts can be questioned. However, when taking into account the differences and similarities between sport and non-sport teams the extent to which the obtained knowledge about sports teams can effectively be applied to other teams is increased (Wolfe et al., 2005: 200). Important differences which have to be taken into account when determining comparability of teams are: team task, training and development, structure, time and boundaries. Soccer
teams are, in this context, relatively permanent entities with regulated episodes of change.
Just like organizational teams, soccer teams have relatively long life spans. The structure of the team, i.e. the role of each player in the team, is based on prescribed role
descriptions. Moreover, there is a sharp division between training episodes and performance episodes. In addition, the performance and development of players is tracked, comparable to the tracking of employees in organizations, and players are rewarded for their own and team performance (cf. the determination of transfer fees;
Feess & Muehlheußer, 2003). This is in line with organizational settings where
employees can be promoted, demoted or lose their job if their talents and qualities do not suffice anymore.
Theoretical implications
The theoretical contribution of this paper is twofold. On the one hand, this study advances literature on membership change by focusing on the timing of change in relation to team development (Arrow & McGrath, 1995). On the other hand this study contributes to team theory by defining roles within teams on the basis of the centrality of transfers (Arrow & McGrath, 1993; Humphrey et al., 2009).
Theory on membership change is advanced by focusing on the impact of transfers at two pre-determined moments in a soccer season on final team performance. Transfers may heighten commitment of the team and may be able to transform the team because of their expertise (i.e. in this case are transfers equal to newcomers only: Levine et al., 2001).
However, the extent to which transfers elicit commitment depends on the timing of change in relation to team development (Arrow & McGrath, 1995). In this paper it was found that entering transfers at the start of the season, when a team still has to establish itself, did not affect performance. Changes at the central mid point, when a team has already formed, had a negative effect on performance. For theory this implies that future research should take into account the timing of change, also when it concerns transfers.
Second this research contributes to literature on team theory by differentiating between more and less central transfers. Here the definition of centrality was based on a measure of relevancy of a player in a team, based on past decisions of the coach. The measure used here is not directly related to team performance, but rather implies
structural aspects of transfers within the team. Even though in this research no differential
