Managing behavior as a determinant of self-managing team performance

Managing behavior as a determinant

of self-managing team performance

Lisa van den Bor

Master thesis MSc Business Economics Specialization Organization Economics 15 ECTS

First university supervisor: Dr. Jeroen van de Ven

Second university supervisor: Prof. Dr. Randolph Sloof First company supervisor:

Anonymous Student number: 5933471

Second company supervisor: Tel.: +31 (0)6 11 05 78 40

Abstract

Most articles studying the relation between managing behavior and the performance of self-managing teams focus on subjective performance measures. Managers can hold on to traditional behavior, which is active and intervenes in the teams work or can engage in new supportive and encouraging behaviors. In a financial services company 74 teams were studied and in a complete model of self-managing work group performance the managing behaviors were evaluated against both subjective (team and manager rated performance and satisfaction) and objective performance measures (effectiveness, the % of customer requests which are handled according to the target). Results indicated that a one standard deviation increase of active behavior of the manager reduces team effectiveness by 5%. An increase of one standard deviation in encouraging self-managing behavior, one of the supportive behaviors of the manager, increases team effectiveness by 12%. A significant relationship between the subjective performance measures and the managing behaviors was not found.

Key words: Teams, autonomy, self-managing teams, managing behavior, performance

Acknowledgements

I would like to thank my thesis supervisor, Jeroen van de Ven, for thinking along about how to combine my internship with my Master thesis, his support and useful feedback. Furthermore I thank XXXXXXXX, XXXXXXXX and XXXXXXXX. for helping me performing my research and when I was struggling with regression models or variables. Lastly I would like to thank all of the rest of my colleagues for their interest in the results of my research, introducing me to the organization and people I needed to get in touch with for my study and making the time writing my thesis as informative and pleasant as it was.

Table of contents

1. Introduction ... 1

2. Self-managing teams ... 3

2.1 (De)centralization and delegation ... 3

2.2 Self-managing teams ... 4

2.3 Managing a self-managing team ... 6

3. Methodology ... 11

3.1 Hypotheses... 11

3.2 Setting ... 12

3.3 Data and sample ... 12

3.4 Research procedures ... 13 3.5 Measures ... 14 4. Results ... 20 4.1 Descriptive statistics ... 20 4.2 Preliminary results ... 21 4.3 Regression procedures ... 23 4.4 Regression results ... 26 5. Discussion ... 32 6. Conclusion ... 35 References ... 37 Appendix ... 41

A. Cronbach’s alphas on question items within teams ... 41

B. ANOVA ... 43

C. Employee’s survey questions ... 44

D. Manager’s survey questions ... 47

E. Cronbach’s alphas of variables ... 48

F. Mean comparison test of manager and team rated variables ... 49

G. Mean comparison test of subjective and objective sample ... 50

H. Descriptive statistics total sample (cross-sectional data) ... 51

I. Descriptive statistics reduced sample (panel data)... 52

J. Extended regression results ... 53

K. Correlation coefficients ... 57

1

1. Introduction

Self-managing teams, a form of work design in which no longer the manager manages the team, but the team has, or should have, the freedom to manage itself. Where does that leave the manager? Relatively little empirical research has been performed to properly address this question. Most of the existing literature describes that the manager must get rid of his traditional, directive and active role and exchange it for a more supportive one with the focus on enabling the team to manage itself.

An active manager manages the self-managing team himself, where the team should have this responsibility. He does so by for example monitoring team actions, assigning a team member a particular responsibility, dealing directly with a team’s customer without involving the team or overriding a team decision even if it seemed to be a poor one (Wageman, 1997). The role for the manager of a self-managing team has however become more like that of a supportive coach (Fisher, 1993; Wageman, 1995; Spreitzer et al, 1999; Harris, 2011). By directly interacting with the team, encouraging them to manage themselves and learn new tasks, intended to shape team processes to produce good performance.

Managers behaving in this last described way should according to literature be rated as more effective leaders (Manz and Sims, 1987) and have teams that have higher team rated and manager rated performance (Cohen et al., 1997). Beekun (1989) performed a meta-analysis and concluded that self-managing teams with managers performed worse than self-managing teams without managers. His performed study would imply that any managing behavior, no matter how well intended, has a negative effect on subjectively measured performance which seems like a very strong assumption. Manz and Sims found ambiguity and confusion about the role of the appointed external leaders to be the single most troublesome issue of implementing self-managing teams (1987). Research has also shown that the lack of legitimate control over team actions and decisions and the large number of teams for which an external leader is responsible, makes the role more complex and demanding (Druskat and Wheeler, 2003).

Based on the literature discussed above, to clarify the role of the manager of a self-managing team the following research question is formulated:

Do, and if so in what way, active and supportive managing behaviors influence the performance of self-managing teams?

2

This thesis will contribute to existing literature by analyzing the relationship between on the one hand active and supportive managing behavior and on the other hand both subjective as objective performance measures making use of quantitative data from a company in financial services and a complete model of self-managing work group performance. The obtained results indicated that on a 7-point Likert scale a one standard deviation increase of active behavior of the manager reduces team effectiveness by 5%. An increase of one standard deviation in encouraging self-managing behavior, one of the used supportive behaviors of the manager, increases team effectiveness by 12%. A significant relationship between the subjective performance measures and the managing behaviors was not found

This research starts with a literature review elaborating more on self-managing teams and the changed role of its managers. The third chapter starts with formulating the hypotheses, after which the obtained data is described, the research procedures are given and the variable measures will be discussed. After that in chapter four the empirical part of this study the hypotheses will be tested using regression analysis. Chapter five contains a discussion of the results with possible limitations of the study. Finally chapter six concludes with a summary of the research and the results and presents a direction for future research.

3

2. Self-managing teams

In this chapter the existing literature on the subject of delegation to teams will be discussed. First the difference between centralization and decentralization, the definition of delegation and the latter’s effects will be given. Sequentially there will be discussed what a self-managing team is and what the reason for their existence is. In the third paragraph there will be reviewed what is known about, what seems to be contradicting, the role of the manager of a self-managing team.

2.1 (De)centralization and delegation

Centralization and decentralization are contrary organizational structures. Under centralization, authority and responsibility are concentrated at one person or one group of persons. In centralized organizations corporate management is the party owning all of the decision- making power. Under decentralization authority and responsibility are spread over several persons or several groups of persons in the organization (Jensen and Meckling, 1998). Delegation refers to the process of assigning tasks, responsibility and authority to lower level staff within the firm and when the latter is done on a large enough scale, this will result in decentralization. Interval delegation is a phenomenon that is often observed in organizations, the employee is given the authority to make the final decision, but subject to constraints set by his manager (Holmström, 1984).

Holmström was, in 1977, the first to address the advantages and disadvantages that go hand in hand with delegation. He argued that in a simple manager-employee model, optimal delegation, its benefits and costs depends on the relationship between the manager’s and the employee’s preferences and the degree of the employee’s expertise. First, the closer the preferences are, the more the manager can increase the employee’s freedom in choosing his own actions, because the manager’s preferred action equals the employee’s preferred action. When under delegation the preferences are not aligned the problem referred to as the loss of control plays his parts. The manager is no more in control, cannot take any actions himself and may incur costs owing to inconsistent objectives (Jensen and Meckling, 1998). Second, according to amongst others Holmström (1977) and Lazear (1998), under delegation local information is effectively used in the decision making process. Employees possess job specific knowledge which is relevant in making job related decisions. Specific knowledge, opposite to general knowledge which is common information, is costly to transfer.

4

Information must not only be communicated by the employee to the manager which takes time, but the manager must also comprehend the message sufficiently well to act on it. A last benefit from delegation, independent of any conditions, is that by the empowerment of lower level employees, their incentives to do their job are strengthened. By the increase in their responsibility, they become more committed to their job and therefore motivated to show more initiative (Aghion and Tirole 1997; Baker et al. 1999).

2.2 Self-managing teams

When delegation occurs, the delegated tasks, responsibility and authority are most often reassigned to teams of employees instead of individual employees. Cohen and Bailey define teams as a collection of individuals who are interdependent in their tasks, who share responsibility for outcomes, who see themselves and who are seen by others as an intact social entity embedded in one or more larger social systems and who manage their relationships across organizational boundaries (1997). What this results in is an autonomous team, an empowered team or a self-managing team, all terms referring to the same matter and used interchangeably in this thesis. A self-managing team is defined as a group of interdependent individuals, that have responsibility for a relatively whole task and that share this responsibility by self-regulating, monitoring and controlling the contributions of its members (Cohen and Ledford, 1994). Around 1960 self-managing teams were almost never heard of, but now, more than fifty years later, in manufacturing plants using self-managing teams is a familiar way of working (Druskat and Wheeler, 2003).

In contradiction to how traditional teams function, as the name already suggests, self-managing teams are not managed by a manager, but by themselves. The teams may or may not have a direct supervisor, but they are responsible themselves for decisions that traditionally lay in the domain of the manager (Cohen and Bailey, 1997). They are held accountable for decisions about what is done and how it is done, by selecting the work they will take on and the work methods that they will use. Furthermore they are responsible for when the work is done and by whom, by scheduling the work and dividing their tasks (Cordery, 2010; Harris, 2011). Self-managing teams do have output quotas to meet imposed on them from outside. When a team sets its own targets it can be named a self-directive team (Sundstrom et al., 1990). Within the team leadership is shared, thus at any time any member can take the lead (Karriker, 2005). Usually an autonomous team performs a relatively whole

5

task (Caroli, 2001). This holds that they produce a complete product or provide a full service, so the tasks of the team members are highly interrelated (Cohen and Baily, 1997).

Of course it is not popular for companies to implement self-managing teams for no reason. It is argued that organizations or departments with these teams achieve cost reductions, higher productivity and improvements in quality (Cohen and Baily, 1997). These benefits are realized, by employees putting in more effort. Aside from the motivation which arises from working as a team together towards common goals, members of self-managing teams tend to put in even more effort than members of traditional teams. Autonomy gives employees control and initiative over their jobs, since they are the one deciding on how to do their jobs. One the one hand this enables employees to use their creative potential, knowledge, skills and abilities (Batt and Appelbaum, 1995; Gallie et al., 2012; Janz et al., 1997). On the other hand when employees can affect decisions that regard themselves, they are more likely to attach value to the outcomes (Bourgault et al., 2008; Cohen and Ledford, 1994; Delarue, 2007; DeVaro, 2006; Spreitzer et al., 1999). Together this increases commitment, satisfaction and internal work motivation and it thus reduces shirking.

Traditional teams can work more efficient than individuals through the gain in flexibility, but self-managing teams can work even more efficient than traditional teams. Autonomous teams have the authority to actually make decisions themselves, which results in higher decision making quality and a quicker decision making process. The theory behind this is that first of all self-managing team members create greater knowledge of jobs through learning from one another (Lawler, 1986). Second, because they are close to the action they frequently have information that higher management does not have and are now able to quickly solve problems on the job (Campion and Medsker, 1993; Johnson et al., 2013). Finally, since they have the responsibility for their own whole task, they get more motivated to keep continually improving their own work process (Cohen and Ledford, 1994; Glassop, 2002; Rolfsen, 2013).

Apart from employees exerting more effort and working more efficient, self-managing teams establish a gain in efficiency by the reduced need for coordination (Delarue 2007; Glassop, 2002; Ichniowski et al., 1996; Wall et al., 1986). With responsibility delegated to lower levels in the firm, the hierarchy gets flatter and external coordination of the teams diminishes.

It has often been empirically studied if implementing autonomous teams results in the benefits named above. Cotton found that adoption of self-managed teams significantly

6

increased productivity (1993). Cohen and Ledford (1994) and Cohen et al. (1997) analyzed a large telephone company and found that the effectiveness of the autonomous teams was significantly higher for employee and manager ratings of performance and this difference was reasonable in size. Amongst manufacturing workers Seers et al. (1995) reported that autonomous team members reported higher levels of cohesiveness and were more satisfied with their colleagues and jobs than members of traditionally managed teams. Also Wageman (1997) who performed a study at Xerox Corporation’s Customer Service, found that teams who had the authority to make decisions themselves, without interference from their leader, strongly outperformed those that did not. Batt (1999) assesses the difference in self-reported service quality and individual average monthly sales productivity between traditionally and self-managed teams in a company in telecommunications. Her results showed that employees who participated in a self-managing team reported significantly higher levels of service quality and also had significantly higher levels of average individual sales than employees who joined a traditionally managed team. Last of all, Glassop (2002) studied Australian, both private and public, firms throughout seventeen different industries she concluded that self-managing work teams have higher self-reported levels of productivity, compared to firms without these team structures.

2.3 Managing a self-managing team

Although the name does not quite suggest it, but self-managing teams do still have managers. However when the team takes on the activities that were formerly in the province of the manager, what is then the role of the manager? Now it is the team that controls their own work, monitor their own performance and transform their strategies as needed to solve problems (Wageman, 1997). According to Manz and Sims (1987) the ambiguity and confusion about the role of the manager is the single most troublesome issue of implementation of autonomous teams. During this implementation managers often feel threatened and are afraid that they have become redundant (Wall et al., 1986). Yet literature states that managers do not have become redundant, but there is a whole new role for them to play different from their traditional one.

In traditional teams the manager took on a directive role being the one owning the decision rights, the information and the expertise. The role for the manager of a self-managing team has become more of a supportive coach (Fisher, 1993; Wageman, 1995; Spreitzer et al, 1999; Harris, 2011). Coaching refers to direct interaction with the team that is intended to

7

shape team processes to produce good performance (Morgeson, 2005). This can be done in an active way by managing the team or in a supportive way by letting the team manage itself. In the latter situation the manager should still demand high performance of his team, but his primary tasks lay in supporting the team and thereby enabling and encouraging them to manage themselves. The term manager does therefore not quite suite the function anymore, external leader (Harris, 2011), coordinator or facilitator (Manz and Sims, 1987) are nowadays more often used.

When self-managing teams are implemented it is hard for managers to not conduct business as usual, but take on the uncertainties that the autonomous teams bring, especially if the manager has held a leadership position in a traditional work environment (Druskat and Wheeler, 2003). Although the decision-power officially belongs to the autonomous team, ignorant or unwilling managers might still exhibit active behavior. For example by monitoring team actions, assigning a team member a particular responsibility, dealing directly with a team’s customer without involving the team or overriding a team decision even if it seemed to be a poor one (Wageman, 1997). Lawler (1986) stated that reluctance of external leaders to engage in new needed behaviors is one of the major causes of failure in unsuccessful self-managing teams. Since intervening in the work of autonomous teams undermines the team’s authority it thereby weakens the promised benefits, increased effort and efficiency, discussed in the previous paragraph.

In a small parts manufacturing plant Manz and Sims (1987) researched the link between leader behaviors that encouraged the team’s self-management and the overall leader effectiveness evaluated by the team members. The authors found proof for the relationship by significant positive correlations between the encouraging behavior and ratings. They conclude that there does exist a role for leaders of self-managing work teams, but that it is different than traditional leadership roles. Not looking into the relationship with the encouraging behavior and team performance Manz and Sims (1987) left some spots open for future research to validate their findings. Cohen et al. (1996) extended their research to service organizations in insurance and telecommunications and looked for the connection between the encouraging behavior and the subjective performance measures quality of work life, team rating of performance, manager rating of performance and absenteeism. In their confirmatory factor analysis they only find a significant negative relationship between the managing behavior and manager rating of performance. In a later study Cohen et al. (1997) do find the

8

relationship between encouraging managing behavior and employee satisfaction and self-rated effectiveness to be positive and significant, but fail to take into account other factors influencing these outcome variables, which might cause their results to be biased. Other predictors of performance are for example variables concerning task design, managing design or group characteristics. Cohen et al. (1996) illustrate the importance of using a complete model of self-managing team performance. They find encouraging behavior of the manager to be significantly related to several outcome variables. When the authors also take into account other explanatory variables, as stated above, they only find a negative relation between encouraging behavior and manager rated performance. When not taking into account these other independent variables, since they are expected to be correlated with the dependent variable and some might also be correlated with one or more of the independent variables, this biases the estimates of the coefficients in the regression model where omitted variable bias is present. Also Elloy (2006) extended the research of Manz and Sims (1987) and found that among autonomous teams in a paper mill, the teams with managers who displayed the encouraging leadership behaviors had higher perceived levels of team effectiveness.

Wageman (1997) tried to answer the question how managers can get teams to take on self-management and ensure that those teams will perform superbly by performing a case study at Xerox Corporation’s Customer Service department. Wageman distinguishes between on the one hand design factors such as for example a clear goal, task interdependence, skill diversity, size and training and on the other hand managing behaviors both positive influences like providing reinforces and other cues that the group is responsible for managing itself and negative influences such as intervening in the tasks and identifying problems of the team. She found that self-managing teams, who could operate without negative interference from their leader, strongly outperformed those that did not. Wageman also addresses the fact that on poorly designed teams negative coaching had a much more unfavorable effect than on well-designed teams.

Druskat and Wheeler (2003) use interviews with managers from average and superior performing teams of a consumer goods manufacturing plant, to determine effective leader behavior. The performance levels of teams were set according to objective team performance based on the percentage of specific team production goals. Their findings indicate that the average performing teams were not so much self-managing, as they were participating. When managers took a step back from making decisions and solving team problems and started to

9

well-supported empower their teams this raised their self-managing skills, enhanced ownership over their tasks and responsibilities and improved team performance

Making use of interviews and surveys in three different organizations, namely a pharmaceutical company, a food processing plant and building and grounds maintenance unit of an university, Morgeson (2005) studied the effect active and supportive coaching, respectively positive negative and positive coaching as in the study of Wageman (1997), has on team member satisfaction with external leadership and perceived leader effectiveness. His data revealed that supportive coaching did not show a significant relationship with satisfaction with leadership, but it does display a positive significant link with perceived leader effectiveness. Negative coaching however, showed a negative significant relationship with both the outcome variables.

Lambe et al. (2009) found that in a large global pharmaceutical company empowerment, which is the degree to which managers allowed the selling team to run itself without interference, and control, which is the degree to which managers employed rewards and coached to promote desired self-managing team skills, positively influence the team’s self-managing behaviors which in their turn positively influence the subjective survey measure of selling team performance. Since control contains both (financial) rewards and supportive coaching behaviors, the part of the effect owed to the positive form of coaching only is not known.

Empirical studies on the effect of self-managing teams on outcome variables exist in abundance, but studies researching the effect of active or encouraging management of self-managing teams are scarcer (Morgeson, 2005). Of the few that do some do not look at performance, but at perceived effectiveness of the manager (Manz and Sims, 1987; Morgeson 2005). When they do look at performance they only look at subjective performance measures (Cohen et al., 1996; Cohen et al., 1997; Elloy, 2006, Lambe et al., 2009). Besides that, although Wageman (1997) has showed the importance of taking into account other factors that determine a team’s performance in examining the relation between manager behavior and team performance, several studies do not (Manz and Sims, 1987; Cohen et al., 1997; Elloy, 2006; Druskat and Wheeler, 2003; Morgeson, 2005; Lambe et al., 2009). Furthermore there are articles which ground their research on qualitative methods, such as case studies, instead of quantitative methods (Wageman, 1997; Druskat and Wheeler, 2003). Lastly most of the studies are performed in a manufacturing environment, with researches in the services sector being underrepresented. To sum up there was no study which in reviewing managing behavior

10

combined all of the factors discussed above. Since there is still much to learn about managing self-managing teams, this thesis will contribute to the existing literature by analyzing the relationship between on the one hand active and supportive managing behavior and on the other hand subjective and objective performance measures making use of quantitative data from a company in financial services and a complete model of self-managing work group performance.

11

3. Methodology

In this chapter the hypotheses used for testing the relationship between managing behavior and team performance will be formulated. After that respectively the organizational setting, data and sample and research procedures will be discussed. Finally in the last paragraph the measures used for the behavior of the managers and all the control variables will be given.

3.1 Hypotheses

The previous chapter discussed studies reviewing the effect of active coaching behavior of the manager. They all found the relationship to be negatively related to either perceived leader effectiveness or subjective performance measures. In this research is hypothesized that these findings persist in an analysis of this link with both subjective and objective performance measures making use of quantitative data from a company in financial services and a complete model of self-managing work group performance. Thus the three hypotheses of this thesis are the following:

 Hypothesis 1: Active coaching behavior by the manager of a self-managing team will negatively influence team performance

Also studies assessing the influence of supportive coaching behavior of the manager were previously highlighted. All of them found a positive link between either satisfaction with the external leader, perceived leader effectiveness or subjective performance measures. In this study is hypothesized that these findings also persist in the current research and that they can be extended when objective performance measures are used. Two supportive coaching behaviors, encouraging self-management behavior and the encouragement of learning, are taken into account. Therefore the last two hypotheses are:

 Hypothesis 2: The supportive coaching behavior encouraging self-management by the manager of a self-managing team will positively influence team performance.

 Hypothesis 3: The supportive coaching behavior encouraging learning by the manager of a self-managing team will positively influence team performance.

12

3.2 Setting

The present study was conducted in two geographic areas of a large company in financial services. Over the last few years the company had introduced self-managing teams at several departments in aspiration of improving throughput time of customer requests and thereby customer satisfaction. At the time of research self-managing teams were implemented at the operational departments of department A with 10 teams, department B with 7 teams, department C with 14 teams, department D with 36 teams and department E with 17 teams. The primary task of the teams was taking care of customer requests from start to finish. Different departments and different teams handled different sorts of requests, but jobs were mainly office and clerical. Activities regarded paper work, customer contact by phone and recording and retrieval of information. Groups existed out of 4 to 16 employees (average = 8.19, standard deviation = 2.49, median = 8). The male female ratio ranged from 0 to 0.77 (average = 0.37, standard deviation = 0.17, median = 0.33). There were groups that were all female, but no teams that were all male. The highest percentage of males in a group was 77%.

3.3 Data and sample

Although in total there were 84 self-managing teams, 76 teams were included in the research since there was permission to include only 6 of the 14 self-managing teams of department C. The main data sources for this study were an anonymous employee questionnaire sent to employees of the participating teams, a manager questionnaire sent to managers of the participating teams and records of incoming and processed customer requests. Questionnaires were distributed per e-mail to the subjects in February and March 2014 and participating in the survey was not obligated, but encouraged.

In total the employee survey was send to 610 employees of which 480 filled in the questionnaire, for an individual response rate of 79%. Three individual responses were removed because of incompleteness, leaving a sample of 477 subjects. Amongst others Cohen et al. (1996 and 1997) dropped teams from their sample when fewer than two people had responded to the survey. To get a better view of the teams, in particular of the teams larger than four people, in this research only teams with a response rate of at least 50% will be included in the sample. This was the case for all 76 teams and therefore the team response rate was 100%. The individual response rate per team ranged from 50% to 100% (average = 80%, standard deviation = 15%, median = 80%). This average response rate per team was quite high in comparison to the 70% of Cohen et al. (1996) and 60% of Cohen et al. (1997). All the

13

23 managers of the 76 self-managing teams received the manager questionnaire and 23 sent them back, this results in a response rate of 100% of the manager questionnaire.

The average age category was XX, which means on average employees were on average between XX and XX years old. Tenure at the department employees were currently working ranged from XX to XX years, with an average of XX years (standard deviation = XXXX, median = XXXX). With XX % of the employees having tenure of XX years or less, XX % having tenure of XX or more years and only XX % of the employees had tenure of XX years or more. Concerning the highest completed education level of the employees: XX % had completed LBO, VMBO or MAVO; XX % completed an MBO education, XX % rounded off their HAVO or VWO; XX % completed HBO and XX % finished a university study. The average age category of team managers was XX, meaning that on average they were between XX and XX years old. Managers had an average department tenure of XX years, ranging from XX to XX (standard deviation = XXXX, median = XXXX). Of the managers, XX % had completed MBO; XX % rounded of HAVO or VWO; XX % finished HBO and XX % had a university degree.

3.4 Research procedures

Managers and employees were asked to respond to questionnaire items making use of a 7-point Likert scale in which higher values represented higher values of agreement. The employee survey was distributed to individual subjects, although group level was the level of interest. Therefore individual responses were aggregated to the group level. For aggregation it is of importance that questionnaire items refer to the level of interest (Van de Ven and Ferry, 1980). Consequently, for this research all survey questions referred to the self-managing team an individual was in. Aggregation must also be justified in two ways. First of all there has to be a strong argument that suggests the individuals are in fact a group (Roberts et al., 1978). Second of all there must exist a certain level of agreement within a group. Can there be said that if we aggregate individual responses, we have a reliable group measure? When a group of n = 3 answers 1; 4 and 7 or 4; 4 and 4, aggregated it will be both 4, but it if a group-level construct would be what is measured, it would be expected that group members provide similar ratings (van Mierlo et al., 2008). For this reason the Cronbach’s alpha, a coefficient of internal consistency is calculated to determine the interrater reliability of teams. The different raters, individual employees, in a self-managing team are treated as the different items. The table including the alpha coefficients for each team can be found in Appendix A. The teams

14

were considered to show an acceptable level of internal consistency with the coefficient being between 0.5 and 0.7 and good if it exceeded 0.7. Values below 0.5 are considered unacceptable and therefore teams with such a value of Cronbach’s alpha are not included in the research (Field, 2009). Due to this restriction, two teams will be excluded from the research, leaving a sample of 74 teams consisting of 472 employees. Lastly, there is verified if before aggregation the within-to-between group variance is statistically significant, because if all groups have similar scores on a group measure, the measure does not differentiate between groups and thus is not a reliable group-level construct (Klein and Kozlowski, 2000). The results of the ANOVA tests, which can be found in Appendix B, show that in all of the cases the proportion of within-to-between groups variance is statistically significant. In 20 of the 21 cases it is significant at 1%, once at 5% and the between groups mean square was always larger than the within groups mean square. Based on the results from both the Cronbach’s alpha and the ANOVA analysis, aggregation from individual to group level is legitimated.

3.5 Measures

As stated before data was retrieved by employee and manager questionnaires which can be found in respectively Appendix C and Appendix D. Next to managing behaviors, the employee survey measured variables concerning group composition, group beliefs, group process, task design, organizational context and team rated performance. In the manager’s questionnaire managers were also asked to rate their teams’ performances and their survey measured their managing behaviors. Based on previous literature, two to three question items were used to measure all variables and obtain adequate internal consistency yet limit length. To determine if all of the items of one variable showed enough internal consistency Cronbach’s alphas are calculated, an overview of the alpha’s can be found in Appendix E. Variables were considered to show an acceptable level of internal consistency when the coefficient was between 0.5 and 0.7 and good if it exceeded 0.7. Values below 0.5 are considered unacceptable and therefore variables with such a value of Cronbach’s alpha are not included in the research (Field, 2009). If the alpha coefficient was sufficiently high individual question items were averaged to form a construct for each variable, after that as discussed in the previous paragraph the variables on individual level were aggregated to group level.

Managing behaviors is the category containing the three variables of interest of this research: Active behavior (three items, αTeams = 0.70, αManagers = 0.86), encouraging self-managing

15

behavior (five items, αTeams = 0.86, αManagers = 0.78) and encouraging learning (two items, αTeams = 0.90, αManagers = 0.94). Both teams and managers are questioned about their manager’s/own behavior. The teams’ perceptions of the behaviors are used, since this is what they act on. A mean sample comparison test using paired data (Appendix F) is performed to see if the perceptions concerning managing behavior differ for managers and their teams. The three items measuring the active behavior are based on Morgeson (2005) and Wageman (1997) and evaluate the unasked interference of the manager in the daily operation of the team. The supportive behavior items are based on the Self-Management Leadership Questionnaire of Manz and Sims (1987). Manz and Sims differentiate between six encouraging behaviors. In this research, in line with the study of Cohen et al (1996), from these six encouraging behaviors two variables are created. Encouraging self-managing behavior refers to coaching of the team by the manager so that the team can and will monitor, manage and have responsibility for their own performance. The second supportive variable is encouraging learning so that team members will learn and practice activities before they perform them for the first time.

Next to that two other variables concerning the behavior of the manager are included: Recognition (five items, αTeams = 0.91, αManagers = 0.91) and visibility (two items, αTeams = 0.62, αManagers = 0.65). Recognition of the manager is present when he or she noticeably recognizes that a team performs well, hence individual contributions to the team and team contributions to the company are fairly acknowledged and valued (Borrelli et al., 1995; Spreitzer et al., 1999). Visibility is the extent to which team managers are present at the shop floor and available for the team when help is needed (Spreitzer et al., 1999).

The category group composition includes the variables age, % of males, ability, team boundaries (two items, α = 0.65), expertise (two items, α = 0.57) and relative size. Age is the average age category of the team and male-female-proportion of the team is the percentage of males in the team. Tenure is the average of the number of years, team members have worked at the department they currently work at. Tenure is expected to influence group performance since it comes with a certain amount of know-how, but on the other end it might come with boredom and therefore reduced motivation (Gladstein, 1984). To control for the relative size of the group a dummy variable existing out of three categories, too small, appropriate and too large, is used since understaffing is associated with lower levels of group performance (Ganster and Dwyer, 1995). Teams and team managers were asked to indicate if their team

16

size either too small, appropriate or too large for the work to be accomplished (Campion and Medsker, 1993). There was a statistically significant difference between team’s and manager’s ratings of relative size, teams rated their size as being smaller than did their managers (see Appendix F). Therefore in further analysis manager rated size was used, since their opinion is assumed to be more objective. Ability, in this thesis proxied by the average education level of the team, refers to the team’s ability to function adequately and fulfill the assigned tasks which positively relates to team performance (Cohen et al., 1996; Rousseau and Aubé, 2010). Expertise is a variable showing the extent to which the team members have different complementary areas of expertise which enhances performance (Campion and Medsker, 1993; Cohen et al, 1996; Gladstein, 1984). The variable boundaries of the team is included because failure to set specific boundaries for a team, e.g. letting them operate in multiple teams, causes reduced performance (Ichniowski et al., 1997; Wheelan, 2010).

Group beliefs consists of the five variables, cohesion (three items, α = 0.70), conflict (two items, α = 0.63), norms (three items, α = 0.61), preference for self-managing teams (two items, α = 0.48) and structural clarity (three items, α = 0.80). The variable cohesion refers to the tendency for a group to stick together and measures the level to which members relate to one another (Gupta et al., 2010). The variable is included since group cohesion has proved to be positively associated with performance (de Jong, 2011; Evans and Dion, 2012). Conflict on the other hand is seen as an impaired behavior and is said to worsen team performance (Gupta et al, 2010; Jehn, 1995). Norms gives the extent to which a team has standards shared by group members which guide their behavior. Some degree of coordination and consensus is required for group success (Cohen et al., 1996; Wheelan, 2010). Preference for self-managing teams is the extent to which team members prefer to work in self-managing teams. When they do, they may be more satisfied with their work and thus be more effective (Campion and Medsker, 1993). However, since the Cronbach’s alpha of this construct is too low the variable preference for self-managing teams will not be included in the research. Structural clarity stands for clear group roles and goals. It gives an understanding of why the group exists and what it is trying to accomplish and is therefore expected to increase effort and thereby group performance (Guzzo and Dickson, 1996; Pearson, 1992; Wageman, 1997)

The category group process contains two variables that are not included in this study due to unacceptable levels of Cronbach’s alphas, namely ability to renew (two items, α = 0.16) and

17

participation (three items, α = 0.42). Four variables that are included are communication (two items, α = 0.72), feedback (three items, α = 0.60), flexibility (two items, α = 0.51), self-managing behavior (six items, α = 0.73) and workload sharing (two items, α = 0.83). Ability to renew referred to the team’s ability to innovate and find and implement new solutions that address changing task demands (Cohen et al., 1996) and participation referred to the degree to which it was possible for all team members to participate in the decision making process (Campion and Medsker, 1993). High performing teams communicate with each other, which is the reason why communication is included in the research. Cooperating that way invites all members to take part in the operational process and share their ideas (Gladstein, 1984; Wheelan, 2010). Superb teams get, give and use feedback to improve the way they are functioning (Hackman and Oldham, 1975; Wheelan, 2010). Flexibility positively influences performance and is the extent to which team members are willing to fill in for each other when needed (Campion and Medsker, 1993). Self-managing behavior addresses the extent to which the team takes responsibility for, monitors and manages its own performance. It has been empirically proven that this factor is positively related to team performance (Lambe, 2008, Millikin et al., 2010; Rousseau and Aubé, 2010). Workload sharing refers to the extent to which the team members do their fair share of the work. It is positively linked to team performance by not having the reduced efficiency caused by social loafing (Campion and Medsker, 1993)

Task design comprises out of three variables: Interdependence (two items, α = 0.27), significance (two items, α = 0.82) and variety (two items, α = 0.35). Interdependence is considered, since well performing teams demand team members to work together as a unity which increases motivation by a sense of responsibility for the work of others, however excluded from further research due to an unacceptable level of Cronbach’s alpha (Campion and Medsker, 1993; Wheelan, 2010). Significance refers to the degree to which team members feel their work is important to the organization and their customers, to again induce a feeling of responsibility and thereby it strengthens motivation (Campion and Medsker, 1993; Hackman and Oldham, 1975; Spreitzer et al., 1999). Variety, excluded from this study as a result of a too low internal consistency, refers to the extent to which a job requests a variety of tasks in performing the work. Variation motivates employees since it demands them to make use of different skills (Campion and Medsker, 1993; Cohen et al., 1996; Hackman and Oldham, 1975).

18

The category organizational context includes the variables resources and information (two items, α = 0.57), rewards (two items, α = 0.26) and training (two items, α = 0.82). Resources and information refers to the extent to which information, equipment, space and materials teams need to execute their job is always available for them (Cohen et al., 1996; Wageman 1997; Yang, 2008). According to Lawler (1986) these organizational design factors should be included for self-managing teams to be effective. Concerning rewards it is commonly agreed on in literature that in order to motivate group-orientated behavior, rewards should be linked to the group's performance (Gladstein, 1984; Spreitzer et al, 1999; Wageman, 1997). The strong positive relation is acknowledged by many and therefore it is unfortunate that due to a too low internal consistency the measure cannot be taken into account in this study. Training allows employees to develop knowledge and skills needed to perform their work, although according to Campion and Medsker (1993) overall evidence in support of team training is mixed (Gladstein, 1984; Spreitzer et al, 1999).

All of the variables above are set off against team performance. There is no single interpretation of performance, but it can be divided over four dimensions namely: operational, financial, attitudinal and behavioral (Guzzo and Dickson, 1996; Cohen and Baily, 1997; Delarue, 2007). Examples of operational performance measures include productivity, quality of a good or service, customer satisfaction and innovation. Financial performance measures comprise of for example revenues, value-added per employee and costs savings. Attitudinal performance measures are for example employee satisfaction, commitment and trust. Examples of behavioral performance measures contain turnover, absenteeism and safety. These performance measures can be measured at various levels: individual, team, divisional and organizational (Cohen and Baily, 1997). For some levels, some dimensions are more important than others. For example at the organizational level performance is most often measured using the financial dimension and at individual and team level most often the operational dimension is the ultimate measure. In this thesis team level is the level of interest, which is why mainly operational performance measures are used. Survey questions focused on perceptions of overall operational team performance and were answered by employees and managers creating two subjective performance measures team rated performance (three items, α = 0.68) and manager rated performance (three items, α = 0.61). Also the attitudinal performance measure satisfaction (two items, α = 0.71) is researched. Finally, an objective

19

performance measure is used of which the data comes from production records the company tracked at team level. The measure is effectivity, the percentage of customer requests which are taken care of according to the team’s target, which is taking care of the customer requests the same they the requests came in so Today-In-Today-Out. This percentage is taken per week over nine weeks from January 6th to March 7th 2014. An incoming customer request is scanned and its date and time are recorded. Once the request is taken care of teams scan it again, in this way TITO percentages of teams are followed. One drawback is that the company did not properly track this performance data for all 74 teams, but only for 43 teams. Of these 43 teams for four of them information about the inflow of customer requests, to construct the variable inflow discussed below, was not available due to problems with the system. Therefore the reduced sample used to test the relationship between managing behaviors and objectively measured performance consists out of 39 teams. In Appendix G the results of a two sample mean comparison test for unpaired data is performed to see if the samples statistically differ. The differences between the means of the two samples are not significantly different from zero for all of the variables. It is therefore assumed that the reduced sample is no different than the original sample. The reduced sample can be seen as ‘randomly drawn’ from the original one, the results of further analysis of the this sample will therefore be unbiased.

Since objective performance is a percentage of total incoming customer requests that are handled on time, inflow is added as a control variable in the analyses of objective performance. As the number of total to be handled requests rises, it gets harder to keep taking care of the same percentage of customer requests. Another variable only included in the analyses of objective performance is a time dummy, which accounts for differences in performance over time. Finally in all analyses a dummy variable with five categories indicating the department the specific team is in controlling for the differences that exist amongst the various departments.

20

4. Results

In this chapter the main findings of the research will be discussed. In the first paragraph some descriptive statistics will be given and the second paragraph includes the preliminary results. In the subsequent paragraph the regression methods will be discussed and in the last paragraph the results will be presented.

4.1 Descriptive statistics

Below in Table 1 the summary statistics of the variables of interest of the total sample containing 74 teams and the reduced sample containing 39 teams are given. For the summery statistics of the other explanatory variables see Appendix H and I for respectively the total sample and the reduced sample. The panel data set is an unbalanced panel, meaning that not all teams are observed in the same amount of weeks. A total of 37 teams are observed during 9 weeks, one team during 6 weeks and one team during 5 weeks, coming to a number of observations of 344. All variables except effectiveness, which is the percentage of customer requests taken care of according to the team’s target, were measured on a 7-point Likert scale in which higher values represented higher performance or agreement and in which 4 was the average performance or neutral agreement option.

Table 1: Summary statistics

Total sample (cross-sectional) Reduced sample (panel)

Obs. Mean Std. Dev. Min Max Obs. Mean Std. Dev. Min Max

Performance measures Team ratings 74 5.26 0.47 4.00 6.07 344 5.26 0.50 4.00 6.07 Manager ratings 74 4.77 0.75 3.00 6.67 344 4.89 0.74 3.00 6.33 Satisfaction 74 5.01 0.45 3.88 6.00 344 5.04 0.46 3.93 6.00 Effectiveness - - - 344 XXX XXX XXX XXX Managing behaviors Active 74 4.56 0.65 2.57 5.70 344 4.71 0.70 2.57 5.70 Encouraging self-management 74 5.48 0.56 3.49 6.73 344 5.60 0.56 3.49 6.73 Encouraging learning 74 4.61 0.89 2.00 7.00 344 4.72 0.90 2.00 7.00

In the total sample teams on average rated their performance between quite high and high, with average being the lowest and high being the highest performance rank. Managers on average rated their teams’ performance between average and quite high, with quite low being the lowest and between high and very high being the highest performance rank. It seems as though teams assess their performance as higher than do their managers and as can be seen in

21

Appendix F this difference is statistically significant. In the reduced sample effectiveness is the performance measure of interest. On average per week XX% of incoming customer requests are dealt with according to the target set, with only XX % being the lowest and XX % being the highest level of effectiveness.

On average all the managing behaviors are present above average. According to the teams there exist managers that do almost not show active behavior and on the other hand there are managers that show a certain amount of active behavior. For encouraging learning displays the same pattern as active behavior. Encouraging self-managing behavior concentrates more on the upper side of the scale. There were manager that exhibited between not so much and neutral encouraging self-managing behavior and on the other extreme there were managers exhibiting between a certain extent and completely encouraging self-managing behavior.

4.2 Preliminary results

Figure 1 to 3 display the means of the dependent performance variables, per Likert scale of the independent managing behaviors. Due to anonymity reasons the effectiveness lines are not depicted against the axis, but presented next to the graphs without numbers. Figure 1 concerns hypothesis 1: Active coaching behavior by the manager of a self-managing team will negatively influence team performance. At first sight support for this hypothesis cannot directly be derived from the graph. Contrary to the direction of the hypothesized relation, teams who rate their manager to be more active show have slightly higher mean performances for all four performance measures. Figure 2 highlights hypothesis 2: The supportive coaching behavior encouraging self-management by the manager of a self-managing team will positively influence team performance. It shows that the teams who score their manager to exhibit this supportive coaching behavior have higher mean performances for all four performance measures. Figure 3 regards hypothesis 3: The supportive coaching behavior encouraging learning by the manager of a self-managing team will positively influence team performance. As for hypothesis 1, at first sight support for this hypothesis does not directly show. Teams who score their managers to exhibit more encouraging learning behavior have higher means of team rated performance, lower means of manager rated performance and almost equally mean satisfaction and mean effectiveness.

22 1 2 3 4 5 6 7 0.5 v 0.0 v 5.5 v 5.0 v 1.0 v 6.0 v Active behavior v Satisfaction v Performance v

Manager rated performance v

Team rated performance v

4.8 5.1 5.6 4.6 4.8 4.9 1 2 3 4 5 6 7 0.0 v 6.0 v 5.5 v 5.0 v 4.5 v 4.0 v 3.5 v 3.0 v 1.0 v 0.5 v Satisfaction v

Manager rated performance v

Performance v

Encouraging self-management v

Team rated performance v

X 5.0 4.6 5.2 3.9 3.0 4.2

Figure 1: Mean performance per scale of active behavior

Figure 2: Mean performance per scale of encouraging self-management

Figure 3: Mean performance per scale of encouraging learning

1 2 3 4 5 6 7 4.5 v 5.0 v 1.0 v 6.0 v 5.5 v 0.5 v 0.0 v Satisfaction v

Manager rated performance v

Team rated performance v

5.0 4.3 4.8 4.9 5.0 5.1 Encouraging learning Performance v

23

Also from the correlation coefficients between the performance measures and the independent variables of interest an indication of the dependence between the variables can be concluded. Significant positive (negative) correlation indicates that high scores on the one variable go hand in hand with high (low) scores on the other variable. In accordance with the expectations encouraging self-managing behavior and encouraging learning show a significant positive correlation with satisfaction, indicating a positive association between the variables. Not in line with the expectations encouraging learning shows a negative significant correlation with the objective performance measure effectiveness. Furthermore none of the correlation coefficients are significant. Thus at first sight most associations expected by the hypotheses presented in paragraph 3.1 do not really show. If there is a correlation between two variables it does not have to be the case that one variable causes the other. A third or a combination of other variables may cause the two variables to be correlated. It is therefore important to take into account the other explanatory variables discussed in the previous chapter, since they also influence the performance of a self-managing team. To be able to accurately test the hypotheses in a complete model of self-managing team performance regression analyses will be performed.

Table 2: Correlations among dependent and independent variables

*

p<.05; **p<.01; ***p<0.001

4.3 Regression procedures

The three subjective performance measures are observed for different teams at one point in time, also known as cross-sectional data. The regression method used to test their relations to the different managing behaviors is the ordinary least squares (OLS) estimator, by far the most common way to estimate a regression line that is as close as possible to the observed cross-sectional data (Stock and Watson, 2012). Since the dependent variables are not dichotomous but continuous instead of a logistic regression there is chosen for a linear

1 2 3 4

Performance measures

1. Team ratings (total sample) 1.00

2. Manager ratings (total sample) .32** 1.00

3. Satisfaction (total sample) .56*** -.02 1.00

4. Effectiveness (reduced sample) 0.04 0.17** -0.17** 1.00

Managing behaviors

5. Active .22 .12 .22 0.09

6. Encouraging self-management .14 .02 .32** -0.05 7. Encouraging learning .21 .00 .29* -0.15**

24

regression. The OLS estimator is efficient among a certain class of unbiased estimators, but this efficiency holds under some special conditions. When there is heteroskedasticity in the model the OLS method does not longer have the smallest variance possible. Weighted least squares (WLS) becomes a better option. In all models of this thesis there is tested for heteroskedasticity, but it is never present. Several Breusch-Pagan/Cook-Weisberg tests for heteroskedastisity are performed. In all cases the null hypothesis of homoscedasticity cannot be rejected due to high Chi2-statistics (see appendix J). Furthermore the OLS estimator can be sensitive to outliers and therefore may not be efficient when large outliers are not rare. An estimator like the least absolute deviations (LAD) estimator is less sensitive to large outliers and is in that case more efficient. However, large outliers in this research are unlikely since the independent variables are rated on a 7-point Likert scale. Taking these two conditions and the data of this research into account OLS is the best linear (conditionally) unbiased estimator available.

Also the objective performance measure effectiveness is measured for multiple teams, but now each team is observed at multiple time periods during nine weeks as discussed in paragraph 4.1. Also the control variable inflow, which controls for the differences in inflow of customer requests, is observed at multiple times during nine weeks. The rest of the independent variables however change slowly and could therefore be considered constant during the research period. To test the relationships between effectiveness and the three managing behaviors a random effects model, suited for panel data, is used. This is because under the random effects approach the transformation involved in the generalized least squares (GLS) procedure will not remove the explanatory variables that do not vary over time, hence their impact on the dependent variable can be enumerated contrary to for example the fixed effects model (Brooks, 2008). A potential drawback from the random effects model is the stringent restriction that the model is only valid when the explanatory variables are not correlated with the error term, a thing that a fixed effects model does allow. Since per team most of the explanatory variables are constant over time the error terms will be positively serially correlated. The consequence is that the estimated standard errors will be smaller than the true standard errors, in that case null hypotheses will be rejected unjustified. Therefore in the GLS regression in this study standard errors are adjusted for 39 clusters in teams.

In both regressions the variables (for a list see Appendix E) of which the data was collected making use of the 7-point Likert scale was standardized. Holding that first the

25

variable was subtracted by its mean and then the result was divided by the standard deviation. What results is a variable with a mean of 0 and a standard deviation of 1. This is done for interpretation purposes, since now the standardized coefficients report how increases in the explanatory variables influence the relative position within the group. The standardized variables are indicated in this way, for example X’ is the standardized variable of X. Interpreting the standardized regression coefficients goes as follows. Increasing the independent variable with either one or one standard deviation in case of standardization, results in an increase in the dependent variable of its coefficient or in case of standardization of this variable its coefficient times the standard deviation of the dependent variable.

Before the regression analyses are performed potential multicollinearity in the model will be explored. Multicollinearity refers to two or more regressors that are highly correlated with each other. When severe multicollinearity is present then the coefficients on at least one of the regressors will be imprecisely estimated, they will have a large sampling variance. In Appendix K a table with the correlation coefficients between all the dependent and independent variables are included. There is no exact rule from which level of correlation multicollinearity exists, but when the coefficient exceeds 0.5 it starts getting serious (Stock and Watson, 2012). As can be seen there are several variables that have correlation coefficients of over 0.50. To check which variables might cause multicollinearity the variance inflation factors (VIFs) are calculated (see Appendix L), when a VIF passes 5 possible multicollinearity is plausible (Brooks, 2008). In the first calculations there are four variables with a VIF greater than 5. First to be eliminated from further research is communication with the highest VIF of 7.08. Again calculating the VIFs gives two variables with too high factors: Encouraging self-managing behavior (6.75) and cohesion (5.28). Since the first variable is needed to answer the hypotheses of this research, Cohesion is removed. Calculating the VIFs for the third time still gives a VIF of 6.75 for encouraging self-managing behavior. The second highest VIF is 4.60 for recognition. Recognition also shows a very high correlation of 0.71 with encouraging self-managing behavior, therefore also recognition is removed from the research. Once more calculating the VIFs shows that now all of the remaining variables have VIFs fewer than 5, indicating that when using these variables together multicollinearity is not present nor a problem.

26

4.4 Regression results

In Table 3 depicted on page 30 the abbreviated results of the four regression models are given, in Appendix J the extended regression results can be found. For every model the extended results present three regressions: One with only the three studied managing behaviors; one also including all other explanatory variables discussed in paragraph 3.5; and one where the insignificant other explanatory variables of the second regression are left out. The abbreviated results only present the second regression, the complete models of self-managing team performance. The following hypotheses are tested in this thesis:

 Hypothesis 1: Active coaching behavior by the manager of a self-managing team will negatively influence team performance.

 Hypothesis 2: The supportive coaching behavior encouraging self-management by the manager of a self-managing team will positively influence team performance.

 Hypothesis 3: The supportive coaching behavior encouraging learning by the manager of a self-managing team will positively influence team performance

In the first model, which uses the team rated performance as the performance measure, none of the three hypotheses are supported which is in line with the findings of Cohen et al. (1996). At first only the three managing behaviors, active, encouraging self-management and encouraging learning, are regressed at team rated performance. This results in the expected sign of only encouraging learning and no significant coefficients. The adjusted R2, a measure of how well the data fits the model, is with its value of 0.02 extremely low.

In the second regression all the explanatory variables discussed in section 3.5 are also taken into account. This is of great importance since omitting these variables may influence the relationships found. For example Cohen et al. (1996) find a significant effect of encouraging supervisory behavior on performance, but when taking into account controls for other predictors of performance, like group or task characteristics, this significant relationship does not show. This second regression has a substantially higher adjusted R2, 0.35. The adjusted R2 corrects the R2 for extra variables that are included in the model. Adding variables will genuinely increase the R2 even though the fit of the data might not always get better. Therefore, when looking at the adjusted R2 it is known that the higher value is not due to the increased explanatory variables, but due to a better fit of the data. In this regression analysis the coefficients of the three managing behaviors are lowered, indicating there was omitted