leader turnover and turnover effects - International Business [PDF]

Sep 1, 2011 - see, e.g., Maximiano (2006) and Lafuente and García-Cestona (2010). The consequences of a manager or coac

19 downloads 5 Views 2MB Size

Recommend Stories


2017 Q4 turnover Turnover like-for-like
Ask yourself: What do I need to change about myself? Next

Protein turnover and proliferation
If you want to become full, let yourself be empty. Lao Tzu

Turnover Minimization
I tried to make sense of the Four Books, until love arrived, and it all became a single syllable. Yunus

Organizational Demography and Turnover
Don't ruin a good today by thinking about a bad yesterday. Let it go. Anonymous

Manager turnover and firm performance
The only limits you see are the ones you impose on yourself. Dr. Wayne Dyer

and Turnover Intention:A Proposed Framework
I tried to make sense of the Four Books, until love arrived, and it all became a single syllable. Yunus

Job Openings and Labor Turnover
Don't count the days, make the days count. Muhammad Ali

Antecedents to Retention and Turnover
Don't fear change. The surprise is the only way to new discoveries. Be playful! Gordana Biernat

Bone Turnover Status
What we think, what we become. Buddha

Revelations From Workforce Turnover
Be grateful for whoever comes, because each has been sent as a guide from beyond. Rumi

Idea Transcript


Doctorate in Economics, Management and Organization (DEMO) Departament d’Economia de l’Empresa Universitat Autònoma de Barcelona (UAB)

LEADER TURNOVER AND TURNOVER EFFECTS MASTER THESIS

Supervisor: Prof. Jordi Brandts Bernad Academic year 2010/11

Handed in by: Christina Rott Despatx B1-123 Departament d’Economia de l’Empresa Fac. CC. Econòmiques i Empresarials – Edifici B Universitat Autònoma de Barcelona 08193 Bellaterra (Cerdanyola del Vallès) Barcelona (Spain) E-Mail: [email protected] Due to: 01/09/2011

LEADER TURNOVER AND TURNOVER EFFECTS! (WORK IN PROGRESS, RESULTS ARE PRELIMINARY) Jordi Brandts Bernad*, Christina Rott**, Carles Solà Belda***

ABSTRACT In this paper, the effects of an external leader turnover are analyzed with an experimental laboratory study using the sequential voluntary contribution game in a finitely repeated form. The form of leadership is leading by example, i.e., first, the leader of a group decides, and then the followers are informed about the leaders decision and take a decision on their turn. The results suggest that (1) contributions by leaders and followers do not increase significantly after an external leader turnover takes place, (2) the leader turnover increases the heterogeneity of group contributions, (3) leaders do contribute significantly more than followers, and (4) leaders’ and followers’ contributions are highly correlated, but the following behavior decreases significantly over the rounds if a turnover takes place and if the leader can lead a group only by setting an example.

1. INTRODUCTION Almost all kind of institutions, firms, departments and (sport) teams are organized in some kind of hierarchical structure and guided by a leader. Societies are lead by politicians or ideological leaders, firms are conducted by managers, departments are guided by directors and teams are headed by coaches. Previous experimental studies have shown that the introduction of leading by example in form of a sequential game structure yields higher

!

We would like to thank Arthur Schram, Gary Charness, Orsola Garofalo, Julian Rode, Daniel Schunk and the participants in the discussion lunch at the Business department at UAB for constructive comments on how to improve the paper. * Universitat Autonoma de Barcelona - Department of Business Economics and Instituto de Analisis Economica (CSIC). Address: Campus UAB, 08193 Bellaterra (Barcelona), Spain. Email: [email protected] ** Universitat Autonoma de Barcelona - Department of Business Economics. Address: Campus UAB, 08193 Bellaterra (Barcelona), Spain. Email: [email protected] *** Universitat Autonoma de Barcelona - Department of Business Economics. Address: Campus UAB, 08193 Bellaterra (Barcelona), Spain. Email: [email protected]

1

contributions to the public good then simultaneous contributions, see for example Güth et al. (2007), Rivas and Sutter (2009) and Gächter et al. (2010). Potters et al. (2007) find this positive effect only if returns are not commonly known. The importance of leadership, and in particular of leading by example, becomes clear when thinking about important guides in social and ideological movements like Mahatma Gandhi or Martin Luther King. But also in every-day situations, leadership is an important form of organization and motivation of a group of individuals. Leader turnover is a common process, whereby the circumstances of a leader turnover differ from situation to situation. Leaders are replaced because they are retired, or because they resign voluntarily, or because they are fired. The later is often the case when firm performance or the leader’s performance is low. A turnover can take place internally, i.e. an employee from within the institution, firm, department or team substitutes the previous leader, or the leader is replaced externally, i.e. by a person from outside the institution, firm, department or team. In this paper, we will concentrate on external leadership turnover. Besides the replacement of managers and CEOs in firms, the maybe most appealing and illustrative example of external leadership turnover is the turnover of coaches in (professional) sports teams, see, e.g., Brown (1982), Audas et al. (1997), Audas et al. (2002), Bruinshoofd and Ter Weel (2003), Ter Weel (2005) and Maximiano (2006). The most common situation, in which a leader turnover in a firm or team takes place, is when the group performance is low. The negative relationship between firm or team performance and leader turnover is widely confirmed in the corporate governance literature, see, e.g., Maximiano (2006) and Lafuente and García-Cestona (2010). The consequences of a manager or coach turnover on performance are however rather inconclusive. Whereas some studies on leader turnover find negative effects on firm or team performance,1 others show significant improvements in performance.2 The third type of findings is insignificant performance changes.3

1

See, e.g., Khanna and Poulsen (1995) for stock performance, Khuran and Nohria (2002) for accounting performance, and Audas et al. (1997), Bruinshoofd and Ter Weel (2003) and Ter Weel (2005) for professional sports team performance. 2 See, e.g., Bonnier and Bruner (1989), Weisbach (1988), Denis and Denis for financial performance, and Hotchkiss (1995) for accounting performance. 3 See, e.g., Warner et al. (1988), Dedman and Lin (2002), Danisevska et al. (2006) and Cools and van Praag (2005) for strock performance, Ortín-Angel and Cannella (2004) for accounting performance, and Brown (1982) and Maximiano (2006) for professional sports team performance.

2

Huson et al. (2004) find that relative performance improvements are greater when successor CEOs are hired from outside the firm than when they are insiders. Similarly, Borokovich et al. (1996) show that, for forced dismissal, abnormal stock returns are significantly positive for an external turnover announcement and significantly negative for an internal replacement. Because of these ambiguous results, we study the effects of an external leader turnover on the contribution of leaders and followers to a public good (i.e. cooperation in the group) with a laboratory experiment. The group in the experiment represents a firm or team. The leader of an experimental group simulates the manager (coach) of a firm (team) and the followers the employees (team players). The followers of a group influence the group outcome together with the leader through their contribution to the public good. The leading by example mechanism is represented by a sequential game structure, i.e. leaders decide first on their contribution, followers are informed about their leader’s decision and choose the own contribution level. As summarized by Albert Schweitzer (“Example is leading”), this form of leadership is supposed to be a strong one. The idea of this paper is that, after converging towards a low contribution situation, a leader turnover could be perceives as a “restart” by leaders and followers and yield a similar increase of contributions as provoked by the “restart effect”. In Andreoni (1988) and Croson (1996), participants play the voluntary contribution game in the repeated form and, after the initially announced rounds are over, they are informed that there will be another couple of rounds of the same game. Contributions go up after the prolonged experiment is announced and this is called the restart effect. We call the effect of the external leader turnover on contributions “turnover effect”. The control treatment is designed in such a way that it will be controlled for the pure restart effect when measuring the turnover effect. In the following, we address four research questions. (1) Does an external leadership turnover ceteris paribus lead to an increase in effort by leaders and followers, i.e. better cooperation? With this question we address the ambiguous results in the corporate governance literature on the effect of a leader turnover on firm or team performance and try to find answers with an experimental study. (2) How does the turnover affect the contributions between groups? After a couple of rounds, followers and leaders have an idea about the average following behavior of the other group members. The new leader of a group might 3

adapt faster to the contribution behavior of the followers crystallizing thereby the average contribution preferences of a group, which leads to bigger differences between group contributions. (3) Is (after an external leader turnover) too much of the group performance attributed to the new leader (“leader attribution bias”)? One strand of corporate governance literature studies whether on average too much of the firm performance is attributed to the manager or CEO of a firm.4 In our simple leading by example setting, we want to see whether the contributions of leaders and followers differ and whether leaders react in particular to the turnover. (4) Do followers go along with their leader’s contribution over time? Leading by example is a powerful leadership tool if the followers’ contributions are influenced by the leader’s contribution. Therefore, we check for the correlation between the contributions by the leader and the followers. Furthermore, it matters whether followers go along with the leader over time because most relationships such as working relationships are of repeated rather than nonrecurring nature. The results suggest that, first, the external leader turnover does not cause a significant increase in contribution by leaders and employees in the leading by example setting. Second, we find that the external leader turnover increases the heterogeneity of group contributions whereas, in the treatment without turnover, the heterogeneity of group contributions decreases. Third, we confirm an earlier finding by Güth et al. (2007) amongst others that leaders contribute significantly more than followers. The public good contribution of leaders after a turnover does not increases significantly. All together, we do not find evidence for a leader attribution bias in the leading by example design. The fourth question whether followers go along with their leader over time is confirmed in that contributions by leaders and followers are strongly correlated, but the following behavior decreases significantly over time if an external leader turnover takes place and if a leader can lead a group only by setting an example. In Güth et al. (2007), leaders possess in one of the treatments formal authority in form of exclusion power. This change in the design results in higher contributions by both, leaders and followers, and closer orientation of followers’ contributions to the leaders’ ones. For the moment, leader turnover effects on the level of contributions seem to be bigger than the ones caused by a restart, but not significantly. The average contribution preferences of the followers of a group become more evident if leaders change the group as the group contribution heterogeneity increases. Since the difference between the leader’s and the

4

See, e.g., Green and Mitchell (1979), McElroy (1982), Martinko and Gardner (1987), Doukas et al. (2007) and Billet and Qian (2008).

4

followers’ contributions increases significantly over time in the turnover treatment, additional tools are required in order to render leading by example more efficient. The remaining of the paper is organized as follows: In section 2, a review of the corporate governance literature related to manager and coach turnover is presented as motivation for the experimental design and a short introduction to the theory of the voluntary contribution game is given. Section 3 provides an overview on existing experimental studies related to the restart effect and leading by example in the voluntary contribution game. In section 4, the contribution of the research project, the research questions and the methodology are presented. In section 5, the experimental design is explained. After the presentation of the results in section 6, we propose two possible extensions in sections 7 and conclude with a summary and outlook in section 8. 2. LITERATURE REVIEW AND THEORETICAL FRAMEWORK 2.1. Literature Review Manager Turnover In the corporate governance literature, there exists a wide empirical documentation of the relationship between firm performance and CEO turnover and a clear consensus that firm performance and CEO turnover are negatively correlated with each other.5 The relationship can be summarized by a remark in Jensen and Murphy (1990, p. 238): “… managers are more likely to leave their firms after bad years than after good years and therefore are disciplined by the threat of termination”. This statement suggests that managers are made responsible for firm performance, which precedes the causal relationship between firm performance and manager turnover. Another strand of literature, consistent with the previous research question, focuses on the effects of a leader turnover on firm or team performance. Because of the multiple factors 5

See, e.g., Becht et al. (2002) and Hermalin and Weisbach (2003) for surveys; Coughlan and Schmidt (1985), Warner, et al. (1988), Weisbach (1988), Jensen and Murphy (1990), Denis and Denis (1995), Kim (1996), Borokhovich, et al. (1996), Parrino (1997), Huson, et al. (2001), Engel et al. (2003) and Farell and Whidbee (2003) for the US; Cosh and Hughes (1997), Conyon (1998), Franks, et al (2001), Conyon and Florou (2002), Dahya, et al. (2002), Hillier, et al. (2005), and Florou (2005) for the UK; Kaplan (1994a, 1994b, 1995) for Germany, the US, and Japan; and Danisevska et al. (2006) for Holland; Volpin (2002) and Brunello, et al. (2003) for Italy; Lausten (2002), as well as Neumann and Voetmann (2005) for Denmark; Maury (2006) for Finland, and Crespi, et al (2004) and Lafuente and García-Cestona (2010) for Spain.

5

influencing a leader turnover, previous and subsequent performance, studies cover a wide range of situations from CEO turnovers in firms to coach turnovers in professional sport teams. In contrast to the conclusive findings concerning the relationship between preceding performance and manager turnover, the results on the consequences of a leader turnover on performance are rather ambiguous. First of all, different levels of leader turnover are considered. In firm studies, CEO and top manager turnovers are the focus, representing leadership on a very high hierarchical level. In team studies, coach rather than club manager turnovers are studied representing a more direct leadership of a group of individuals similar to a department director. Second, there are different measures of performance. On the firm level, there is financial performance, i.e. stock market performance, and accounting performance. On the team level, on-field performance is used rather than the financial performance of a club. Results in all three types of studies (effect of turnover on firm/team performance) are inconclusive. Whereas some studies find no significant effects of top executives turnover on stock performance,6 Bonnier and Bruner (1989), Weisbach (1988) and Denis and Denis (1995), among others, show that manager turnover information results in significant increases in stock prices. In contrast, Khanna and Poulsen (1995) report a significant negative reaction of financial performance. Considering forced turnover, i.e. replacement of fired CEOs, Borokhovich et al. (1996) find that abnormal stock returns are significantly positive for an external turnover announcement and significantly negative for an internal replacement. With accounting performance, a more reliable measure of performance, results are also ambiguous. Ortín-Angel and Cannella (2004) find no significant improvement of firm performance after a manager turnover takes place, which is in line with the theoretical prediction of the efficiency wage model by Shapiro and Stiglitz (1984). Denis and Denis (1995) show that forced managerial replacements are followed by large improvements in performance. However, the level of industry-adjusted ratio of operating income before depreciation on total assets shows a monotonic and statistically significant decrease. Khurana and Nohria (2002) get similar results when the turnover is not voluntary and the manager is replaced from outside the firm. Hotchkiss (1995) finds that management turnover improves performance of firms that emerge from bankruptcy. Huson et al. (2004) find that accounting 6

See, e.g., Warner et al. (1988) for american top executive turnover, Dedman and Lin (2002) for CEO turnover in Great Britain, Danisevska et al. (2006) and Cools and van Praag (2005) for studies in Holland.

6

measures of performance relative to other firms deteriorate prior to a manager turnover, and improve subsequently. They argue that this improvement seems to be due to the turnover and management improvements. According to Huson et al. (2004), relative performance improvements are greater when successor CEOs are hired from outside the firm than when they are insiders. Analyses with sport teams show different results, too. Brown (1982) finds no difference in performance change between American football teams where a coach turnover took place within a season and teams where the coach was not replaced. Both groups of teams showed similar declines in output. Audas et al. (1997) for English football, and Bruinshoofd and Ter Weel (2003) and Ter Weel (2005) for the Dutch premier league, show that teams with the same coach improve faster, on average, than teams with a replaced coach. Maximiano (2006) uses match-level team performance data from soccer teams and shows that, on average, teams perform better with the new coach. However, when controlling for teams in a similar performance situation and with a similar likelihood of coach dismissal, the improvement disappears, leading to a null effect of the coach turnover. As the review on the corporate governance literature shows, the negative relationship between previous firm performance and turnover is widely confirmed. Concerning the effects of a leader turnover on the firm or team performance, the results are rather ambiguous motivating this experimental study. 2.2. The voluntary contribution game The voluntary contribution game, also called public goods game, represents a game, in which an individual’s payoff depends on the own contribution and the contribution of the other members of a group. Each group member decides how much of the initial individual endowment she wants to contribute to a public good whose return is beneficial for all group members. The remaining endowment is beneficial only for oneself. The payoff " i, t of the voluntary contribution game of individual i in round t is given by n

" i, t = rp # ( E $ hi, t ) + 14243 Payoff from private good

rV # % h j, t j=1 142 4 3! !

!

Payoff from public good

!

7

where E is the individual endowment in each round, which is fix across individuals and over time, hi, t is the contribution of individual i in round t to the public good, h j, t is the contribution of individual j in round t to the public good with j = 1,...,n , n is the number of ! subject in the group, rP is the return rate !of the private return rate of the ! good, and rV is the ! ! public good. For 0 < rV < r! P < n " rV , the dominant strategy of!each group member is to ! ! contribute zero to the public good. This is the case in the stage game, i.e. when the game is ! ! played for one round only, as well as in the finitely repeated game. The later fact can be ! shown by backward induction. The socially optimal solution is that each group member j = 1,...,n contributes in each round the entire individual endowment E to the public good

leading to an individual round payoff of rV " n " E . !

In the leading by example setting, each round is composed by three sequential stages, ! see also Güth et al. (2007). The role of the leader is assigned to one group member and the ! remaining group members are followers. In stage 1, the leader of each group decides how much of the own endowment to contribute to the public good. In stage 2, the n "1 followers are informed about the decision of the group leader and decide how much of the own endowment to contribute to the public good. In stage 3, all players are informed about the ! average contribution by the other group members, the sum of contributions by all group members and the individual payoff. The equilibrium contribution of leaders and followers in the sequential structure of the game is the same as in the simultaneous game, i.e. zero. This holds for the stage game as well as for the finitely repeated game, which can again be shown by backward induction. Therefore, the equilibrium contribution in the finitely repeated sequential voluntary contribution game is zero, too. The socially optimal solution is just the same as in the finitely repeated simultaneous game: Each group member j = 1,...,n contributes in each round the entire individual endowment E to the public good leading to an individual round payoff of rV " n " E .

!

!

3. CONCEPTUAL EXPERIMENTAL FOCUS ! 3.1. Restart effects in the voluntary contribution game A general finding in public goods experiments, is that participants contribute voluntarily positive amounts to a public good, which conflicts with the free-riding hypothesis of public economics and which was first tested by Marwell and Ames (1979, 1980, 1981). The voluntary positive contribution is not in line with theory because the optimal strategy of an 8

individual is to contribute zero assuming independent decision taking and purely selfish interests, i.e. monetary payoff oriented behavior. However, the voluntary amounts contributed decrease over time, or, in other words, free-riding increases. 7 Andreoni (1988) conducts such a public good experiment in the laboratory and tries to find answers to this observation. He discusses two potential hypotheses and designs an experiment to separate them. The first hypothesis is based on simple learning and suggests that subjects learn the incentives of the game throughout the experiment. The learning hypothesis assumes that subjects learn the optimal strategy, which is contributing zero to the public good, over the rounds of the experimental session. Since they are assumed to do so at different speed the average contribution should decay over time. The second hypothesis suggests that subjects use strategies and play with the objective to influence the other group members' actions, i.e. they take into account the repeated play and use contributions as signals about future contributions. This induces that it is consistent with the strategies hypothesis that subjects contribute positive amounts though they already learned the optimal strategy (zero contribution to the public good under the assumption of rational behavior by the other group members). Croson (1996) replicates Andreoni’s (1988) experiment and the two hypotheses. In order to test for the learning hypothesis, participants are announced that there will be ten rounds of the voluntary contribution game. After the ten rounds are over, participants are informed about a restart, i.e. that there will be another ten rounds of the public goods game. After the restart, contributions to the public good raise in the partners condition (repeated play of the public goods game with the same group members), but not under the strangers condition (repeated play of the public goods game with different group members). This effect is called the restart effect and both, Andreoni (1988) and Croson (1996), observe it in their experiment. The raise in contributions in the partners treatment is not compatible with the simple learning hypothesis, but it is with the strategies hypothesis: Subjects contribute positive amounts to the public good in the beginning of the repeated game in order to influence the contribution behavior of other group members in subsequent rounds. In order to 7

See Davis and Holt (1994) and Ledyard (1995) for reviews. See, e.g., Isaac et al. (1984), Andreoni (1988, 1995), Weimann (1994), Laury et al. (1995), Croson (1996), Burlando and Hey (1997), Gächter and Fehr (1999), Ockenfels and Weimann (1999), Sonnemans et al. (1999), Keser and van Winden (2000), Fehr and Gächter (2000), Park (2000), Masclet et al. (2003), Croson et al. (2005), Carpenter (2007), Sefton et al. (2007), Egas and Riedl (2008), Gächter et al. (2008), Herrmann et al. (2008), Nikiforakis and Normann (2008), Neugebauer et al. (2009) and Fischbacher and Gächter (2010). Also in prisoner’s dilemma games, cooperation declines in experimental studies, see, e.g., Selten and Stoecker (1986); Andreoni and Miller (1993) and Cooper et al. (1996).

9

provide support for the strategies hypothesis, contributions should be significantly larger in the partner than in the stranger treatment. In both studies, partners contribute more than strangers do and these differences are significant in Croson (1996). Therefore, Croson’s (1996) results are consistent with strategies, unlike in Andreoni's (1988) original experiment. Because of the restart, Croson’s (1996) results are however not consistent with simple learning. Andreoni's (1988) data supports neither the simple learning nor the strategies hypothesis. 3.2. Leading by example in the voluntary contribution game Another finding in voluntary contribution experiments is that having a leader in a group who contributes first to the public good increases contributions by the group members. The other group members, the followers, are informed about the leader’s decision before taking their own decision about the contribution to the public good. This design simulates a management mechanism, which is called leading by example, see for example Potters et al. (2005), Güth et al. (2007), Potters et al. (2007), Rivas and Sutter (2009) and Gächter et al. (2010). The difference to authority is that followers contribute voluntarily, i.e. they are not forced to contribute a certain amount to the public good, see also explanations in Hermalin (1998). Güth et al. (2007) study three treatments. First, they examine how providing the leader of a group with exclusion power influences contributions and behavior of followers. Second, they look how internally rotating leaders affect the contributions of all group members as compared to fix leaders. These two treatments create a 2x2 experimental design in the first part of the experiment. Third, they examine the effect of endogenously established leaders as compared to no leadership creation. A leader is endogenously determined if the leader herself and all followers want the leader to remain leader (treatment with fix leaders) or a single member within a group is commonly accepted (treatment with rotating leaders). As to the first two treatments, Güth et al. (2007) find that leading by example increases contribution especially if it goes along with authority in the form of granting the leader exclusion power. Whether it is combined with fix or internally rotating leadership does not make a difference in contributions. As to the third treatment, the results suggest that the endogenous implementation of a leader produces higher contributions compared to groups in which the members vote to not install a leader and to take the decision simultaneously. 10

Gächter et al. (2010) investigate in their paper the characteristics that make a leader a good leader whereby they focus on the individual’s cooperativeness and the individual’s beliefs about the cooperativeness of others. Cooperatively inclined subjects result to be better leaders, which partly reflects the fact that they seem to be more optimistic about the follower’s cooperativeness. 4. INVESTIGATION OBJECTIVES AND JUSTIFICATION 4.1. Contribution and Research Questions As mentioned in the introduction, the most common circumstance, in which a leader turnover takes place, is low firm or team performance. This situation can be simulated in the experimental laboratory with the voluntary contribution game. As it can be seen in wide range of studies, average contributions start far above zero and converge over the rounds of the session towards the equilibrium level. Thus the end of a experimental sessions simulates a situation, in which the group gets stuck in a low contribution situation. The idea of this paper is that, after converging towards a low contribution situation, a leader turnover could be perceives as a “restart” by both, leaders and followers, and yield a similar increase of contributions as provoked by the restart effect in Andreoni (1988) and Croson (1996). We call this outcome “turnover effect”. The intuition behind this reasoning is that the leader turnover makes all group members change their beliefs about the contribution of the other group members. This expected behavior is similar to the strategical hypothesis in Andreoni (1988) and Croson (1996). The control treatment is designed in such a way that it will be controlled for the pure restart effect when measuring the turnover effect. The concrete research questions are: (1) Does an external leader turnover ceteris paribus lead to an increase in contribution by both, leaders and followers, i.e. better cooperation? Or in more applied words: Does an external manager (coach) turnover ceteris paribus lead to higher effort provision by both, managers (coaches) and employees (team members)? (2) How does the turnover affect the contributions between groups? (3) Is (after an external leader turnover) too much of the group performance attributed to the new leader which is simply due to the change in leadership (“leader attribution bias”)? (4) Do followers go along with their leader’s contribution over time? 11

The first research question is the central focus of this study, i.e. whether an external leader turnover increases contributions of group members. Leaders are replaced externally, that means from outside the group, and the new leader has not been leader of the same group in a previous part. We expect that through the change in leadership, beliefs about the contribution of the other group members increase by more than when a new part starts with the same group leader. Fischbacher and Gächter (2010) search for an explanation of differing contribution behavior of subjects in voluntary contribution games. On the one hand, individuals contribute on average more than then pure self-interest level (zero). On the other hand, free-riding is a wide-spread observation in experimental studies. Fischbacher and Gächter (2010) find that the most common preference types are conditional cooperators (55 per cent) and free riders (23 per cent). They conclude that individuals are “imperfect conditional cooperators”, on average, i.e. they prefer to contribute less than the other group members. The second research question investigates a second-order effect, which is the standard deviation of the sum of contributions by all group members on the group level. Potter et al. (2007) study whether the success of leading by example is driven by reciprocity or by signaling. In the treatment, in which leaders are provided with private information about the returns from contribution (asymmetric information), leading by example increases contributions and earnings. With common information, this effect disappears. Potter et al. (2007) conclude that signaling rather than reciprocity drives the success of leading by example in their setting. At the end of part 1, leaders and followers have an idea about the following behavior of the other group members. In the turnover treatment, leaders have no information about the following behavior of followers in the new group. The leaders send the followers with their contribution signals about the other followers’ contribution. Depending on the beliefs of followers about the other followers’ contribution, some groups might be able to increase the group contribution in part 2 whereas in groups with low average following behavior, the group outcome will be quickly lowered because leaders might adapt faster to the following behavior. Therefore, the differences between groups might become stronger if a leader turnover takes place. The third research question addresses an issue of corporate governance. One strand of literature addresses the question whether in general too much of firm performance is attributed to the manager or CEO of a firm, see for example Green and Mitchell (1979), 12

McElroy (1982), Martinko and Gardner (1987), Doukas et al. (2007) and Billet and Qian (2008). In our voluntary contribution game, the analysis is limited to leading by example. Leaders have no authority tool such as exclusion power or control mechanism. We want to check for the finding in Güth et al. (2007) that the leaders’ contributions are significantly higher then the followers’ ones. Additionally, we want to see whether subjects in the position of a leader react particularly to the turnover. The fourth research question deals with the behavior of followers. Leading by example is a useful tool only if the leader’s behavior has an impact on the followers’ behavior. The first question is thus if followers go along with leaders in general or whether their contribution behavior is not influenced by leaders. As it was shown in the second section of this paper, the contribution that maximizes the individual payoff in each round of the game is zero and followers should not be influenced by the leader’s contribution. Güth et al. (2007) among others have shown that this is not the case and that the followers’ contributions are highly correlated with the leader’s contribution. The second step is to see whether followers go along with the leader’s contribution over time. To make leading by example a strong leadership tool, this is crucial because most interactions such as hierarchical working relationships are not of one-shot, but of repeated nature. To our knowledge, this research question has not yet been addressed. We want to know whether the following behavior differs if an external leader turnover takes place as compared to repeated play with the same group leader. 4.2. Methodology The main result in several econometric studies that investigate the correlation between the performance of a firm and the turnover rate of managers is that previous (absolute and relative) performance or performance changes are negatively correlated with CEO turnover, see for example Lafuente and García-Cestona (2009), Kaplan and Minton (2008), Lausten (2002) and Kesner and Dalton (1994). Isolating a turnover effect with empirical data is hardly possible. Therefore, one would need to control for all possible confounding factors because a turnover in real world settings is in general an endogenous decision influenced by a wide spread of factors. That this is a difficult task is confirmed in the empirical literature on manager turnover, see for example Lafuente and García-Cestona (2009).

13

The second best option would be a field experiment, which is hardly realizable. A firm or team or group would need to let an experimenter determine a manager turnover randomly. Only through the randomization, the observed overall turnover effect would be exogenous. If there was any willing firm, team or group running such an experiment would be very costly, risky and eventually from the ethical point of view irresponsible. The main critique to laboratory experiments is that the external validity due to a nonrepresentative sample is not given. Additionally, the transmission to real world settings should be done with caution because of the strong simplification in the game setting. These are two valid critiques that can be relativized if the effect that we study in the experiment is a basic behavioral principle. For all this reason, a laboratory experiment has been chosen which has the advantage that the turnover effect is controllable through randomization of the treatment assignment (externally rotating vs. fix leaders), which means that the causal effect of leader turnover on contributions and cooperation in a group can be studied. Furthermore, cost concerns and ethical problems such as harming a firm, a team or group are not a problem in a laboratory experiment. 5. EXPERIMENTAL DESIGN 5.1. Procedure / Game The general instructions were handed out to the participants in hard copy (see appendix) and read aloud by one of the experimenters. Additional part-specific instructions were shown on the computer screen just before the corresponding part started (see appendix) and also announced aloud by one of the experimenters. The game chosen is the sequential voluntary contribution game consisting of three stages and representing leading by example. Each group was composed by four group members ( n = 4 ), one leader and three followers. The role (leader vs. follower) was assigned randomly in the beginning of the experiment and was the same throughout the entire experimental session. The group composition was created randomly at the beginning of the ! experiment. Followers remained in the same group, whereas leaders did so in the control treatments without turnover and changed the group in the turnover treatments subsequently. 14

In the first stage, the leader of each group decided how much of the endowment to contribute to the public good. In the second stage, followers were informed about their leader’s decision and decided how much of their individual endowment to contribute to the public good. In stage 3, all players were informed about the average contribution by the other group members, the sum of contributions by all group members and the individual payoff. The payoff function was the same for both, leaders and followers. The individual endowment was E = 40 , the return rate of the private good was rP = 1, the return rate of the public good was rV = 0.5 yielding the following payoff function of individual i in round t :

!" = i, t !

!

4

+ 0.5 $ % h j, t j=1 Payoff from private good 142 43 # hi, t ) (1E42 4 3

!

Payoff from public good

!

!

The game was repeated finitely (36 and 24 rounds, respectively) and each experimental session was divided into parts with 12 rounds each. In figure 1, the chronological order of an experimental session with 36 rounds is represented. Each part was preceded by the part-specific instructions, i.e. the information whether there was a leader turnover or not and the information that the group composition would not change for the following 12 rounds. The part-specific instructions were provided on the screen and also announced aloud by one of the experimenters. After the experiment finished, participants were required to fill out a questionnaire and were paid the earnings. The experimental sessions were conducted in the experimental laboratory at Universitat Autònoma de Barcelona (Spain) and programmed with the experimental software z-Tree, Fischbacher (2007). 5.2. Treatments A total of four treatments were run so far. In the following, we will call them, treatment T36, control treatment NoT36, control treatment T24 and control treatment NoT24. Table 1 provides a list with the characteristics of each (control) treatment. In treatment T36, 36 rounds were played and leaders changed groups after part 1 and 2 (after round 12 and 24, respectively). In control treatment NoT36, 36 rounds were played and the group composition

15

remained the same throughout the 36 rounds (no leader turnover).8 In control treatment T24, 24 rounds were played and leaders changed groups after part 1 (after round 12). In control treatment NoT24, 24 rounds were played and the group composition remained the same throughout the 24 rounds (no leader turnover). By ensuring the same chronological structure in the turnover and the control treatment and providing the same information before the start of the experiment, the control treatment controls for the pure restart effect. Differences between turnover and control groups are supposed to be solely due to the turnover. Control treatment T24 and NoT24 were run in order to check for the “last round effect”, which is often observed in experimental repeatedly played voluntary contribution experiments. During the last rounds of an experimental session, participants tend to converge to the free-riding equilibrium.9 At the beginning of part 3 in treatment T36, there are two different effects: (1) the leader turnover effect and (2) the last round effect. Since the last round effect is supposed to be strong the two control treatments T24 and NoT24 were run in order to test whether it is dominant. In that case, part 3 with the last 12 rounds of the experimental sessions with 36 rounds should be excluded from the analysis of the turnover effect and the data analysis should be limited to the first 24 rounds of the 36-roundtreatments. 5.3. Participants Participants were mainly undergraduate students from Universitat Autònoma de Barcelona (Spain) and were recruited using the online recruitment system ORSEE, Greiner (2004). A total of 104 subjects took part in six experimental sessions composed by 57 women and 47 men. Treatment T36 was run with a total of eleven groups, control treatment NoT36 was run with six groups, control treatment T24 was run with five groups and control treatment NoT24 was run with four groups.

8

After round 12 and 24, participants in this treatment were informed that their group composition had not changed and that the group composition would be the same throughout the following 12 rounds. 9 See also the literature list in section 3 to the increase in free-riding over the rounds of a repeatedly played voluntary contribution experiment.

16

Participants were aged between 19 and 49 with an average age of 23.7.10 The average earnings per person were 18.82 Euro in the sessions with 36 rounds and 14.47 Euro in the sessions with 24 rounds (including a show-up fee of 5.00 Euro, respectively). 6. RESULTS 6.1. Contribution behavior in general 6.1.1. General remarks to the observations Figure 2 presents the average contributions of leaders and followers separately in part 1, 2 and 3 for treatment T36 and control treatment NoT36 and in part 1 and 2 for control treatment T24 and NoT24. The difference of contributions in part 1 between groups under control treatment NoT36 (without turnover) and treatment T36 (with turnover) are not significantly different (p=0.4214, N=11 for treatment T36 and N=6 for control treatment NoT36, two-sided two-sample Wilcoxon rank-sum (Mann-Whitney) test).11 However, the relatively low number of observations (6 groups in control treatment NoT36 and 11 groups in treatment T36) could lead to small sample biases and needs to be ruled out by more observations.12 So far, we try to control for this when analyzing the data and focus on the contribution changes within each treatment. A second observation is that the change in contribution within each treatment is very similar when comparing part 1 and part 3. The increase in contributions from part 1 to part 2 in groups without leader turnover is smaller than in groups with an external leader turnover. In part 2, the turnover effect is the only observable effect. In part 3, there are two independent effects: (1) the leader turnover effect and (2) the last round effect. The last round effect is often observed in repeatedly played games and is characterized by a decrease in contribution at the end of the experimental session.13 Since the last round effect is supposed to be strong the two control treatments T24 and NoT24 were run in order to test whether it dominates in part 3. The contribution changes from part 1 to part 3 in both treatments with 36 rounds and 10

Due to a technical problem the data in the questionnaire was not saved in the first session. Therefore, the information about the age of participants only refers to the information from the 88 participants in the remaining sessions. 11 For the comparison, the average contributions per part by all group members are caluclated in order to get one independent observations per group. 12 Further experimental sessions will be run in September 2011. 13 See also the literature list on the increase in free-riding over the rounds of a repeated voluntary contribution experiment in section 3.

17

from part 1 to part 2 in both treatments with 24 rounds are similar (see Figure 2). The contributions of leaders, aggregated over the rounds of each part, do not change significantly comparing part 1 and part 3 in both treatments with 36 rounds and part 1 and part 2 in both treatments with 24 rounds (p>0.46 in any treatment, two-sided Wilcoxon signed ranks tests, N=11 for treatment T36, N=6 for treatment NoT36, N=5 for treatment T24 and N=4 for treatment NoT24).14 The aggregate contributions of followers do not decrease significantly comparing part 1 and part 3 in control treatment NoT36, and part 1 and part 2 in control treatment T24 and NoT24 with 24 rounds (p>0.14 in any treatment, two-sided Wilcoxon signed ranks tests, N=6 for treatment NoT36, N=5 for treatment T24 and N=4 for treatment NoT24).15 In groups with an external leader turnover, the contributions of followers decrease significantly comparing part 1 and part 3 (z=1.778 and p=0.0754, two-sided Wilcoxon signed ranks tests, N=11 for treatment T36). These results support the supposition that the last round effect is indeed strong in the last part. Because of this confounding factor, part 3 of the treatments with 36 is not taken into consideration when analyzing the external leader turnover effect.16 6.1.2. Contribution behavior over rounds in each part We also want to check whether contributions to the public good converge towards the equilibrium contribution (zero) at the end of each part. This behavior would simulate a firm or team that gets stuck in a low contribution situation. Participants know that they will play the game for 36 rounds when the experiment starts and that they will get the part-specific instruction just before the corresponding part starts. So it is not clear that this incomplete information creates a decrease in contribution at the end of each part, as it is usually the case in voluntary contribution experiments at the end of the experiment. 14

For the comparison, the average contributions per part of leaders are calculated on the individual level, i.e. for part 1, it is the average contribution of each subject over round 1 to 12, for part 2, it is the average contribution of each subject over round 13 to 24, and for part 3, it is the average contribution of each subject over round 25 to 36. 15 In order to get independent observations for the statistical tests, the average contributions per part of followers are calculated on the aggregate group level. For part 1, it is the average contribution of all followers of a group over round 1 to 12, for part 2, it is the average contribution of all followers of a group over round 13 to 24, and for part 3, it is the average contribution of all followers of a group over round 25 to 36. 16 Güth et al. (2007) repeat the voluntary contribution game in their experimental study 24 times. For the analysis of leading by example in the exogenous part (especially the analysis of the internal turnover effect), they take only round 1 to 16 into account, which is similar to restricting our analysis of the external turnover effect to round 1 to 24.

18

In Figure 3, the aggregate contributions of leaders and followers in round 1 to 24 are presented for treatment T36 and control treatment NoT36 separately. Across the 12 rounds of each part, there is a clear decrease in contributions in both treatments, i.e. with and without leader turnover. The regression results with pooled OLS are presented in table 3. Clustering for group is done in order to control for the correlation of contributions within a group. The observations in the regressions are the observations from round 1 to 24 in treatment T36 (with external leader turnover) and in control treatment NoT36 (without turnover). In regression 2, “Round” (with values from 1 to 24) and a dummy variable for part 2 are used as explanatory variables, whereas in regression 3, “Part-round” (with values from 1 to 12) is used as explanatory variable instead. The regression results do not change by much when comparing regression 2 and 3 and the coefficients for all three variables “Round”, “Part 2”-Dummy and “Part-round” are highly significant (p

Smile Life

When life gives you a hundred reasons to cry, show life that you have a thousand reasons to smile

Get in touch

© Copyright 2015 - 2024 PDFFOX.COM - All rights reserved.