Monitoring, routines and social dispositions and their effect on team members' intention to cooperate

Monitoring, routines and

social dispositions and their

effect on team members’

intention to cooperate

COEN KLEIPOOL (10990038)

coen.kleipool@student.uva.nl

Supervisor

DR. IR. J. KRAAIJENBRINK

Master Thesis

Amsterdam Business School

UNIVERSITY OF AMSTERDAM

July 15

, 2017

ABSTRACT

The purpose of this research is to deepen the knowledge on team members’ intention to cooperate in situations of high task interdependence. Three constructs are described, and their interaction with and effect on cooperation are investigated. The first construct is monitoring, used in organizations as a coordination mechanism. Monitoring is split up in this research into outcome-based and behaviour-based. The second construct is routinization, which is posed here as an expression of the nature of work – split in high levels of routinization versus low levels of routinization. The third construct has a more social approach: social dispositions are a relatively stable trait, and an expression of how individuals approach a situation with high interdependency – prosocial or proself. These three constructs are presented to the participants (students of the EPMS programme of the University of Amsterdam) in a vignette-based design. The results show no direct effects, nor interaction effects between type of monitoring, level of routinization and intention to cooperate. Significant effects are found between social dispositions and intention to cooperate, indicating that individuals classified as prosocial have a significant higher intention to cooperate than individuals classified as proself. The implications that can be drawn from this research are that thorough research is required to understand why, how and when individuals engage in cooperative behaviour, because there is an increasing amount of evidence suggesting that the directive way of doing so is quite limited. The results of this should be used to reconsider how firms are designed and how they support their employees in cooperating with each other.

TABLE OF CONTENTS

INTRODUCTION

In the first section of this thesis, the topic is introduced, and based on a brief analysis of the current literature a research gap is presented.

Framework

Cooperation within firms is the foundation for performance (Bridoux, Coeurderoy, & Durand, 2011; Gottschalg & Zollo, 2007; Marshak & Radner, 1972; Poortvliet, Janssen, Van Yperen, & Van de Vliert, 2009; Qin, Johnson, & Johnson, 1995). For as long as humanity inhabits earth it has used its cognitive capabilities to collaborate in order to achieve its goals. Finding ways to improve cooperation has therefore always been appealing to scientists. While initially researchers interested in cooperation within firms focussed on an economic or supply and demand approach to cooperation (please review Weintraub (1993) for a extensive summary), recently the emphasis of economic and business-schooled researchers has shifted to a more social and psychological angle, focussing on how for example personal traits influence cooperation in firms. This research is trying to contribute to the business literature by deepening the knowledge about social and economic approaches to cooperation. Specifically, by combining the type of monitoring, the level of routinization and an individual’s social dispositions and tying it to the intention to cooperate in team based situations with high inter-dependence. In doing so, it is trying to establish the interaction between a company’s coordination mechanisms – the monitoring -, an employee’s nature of work – the level of routinization – and a component of an employee’s personality – the social dispositions -. The goal is to contribute to current literature through the deepening of knowledge on cooperation, factors that improve it and design that influences it.

Current standings

Transactions are a fundamental part of our life. Adam Smith (1776) first described transactions by stating that the division of labour could maximize social welfare and that the price mechanism or the market would be the source of coordination of transactions. Coordination is the crucial aspect here: coordination is necessary to guide – complex – transactions among actors, and many authors see firms as the solution for coordination problems the market cannot solve (Coase, 1937; Williamson, 1979, 1981). Alchian and Demsetz (1972) discuss a specific coordination problem, namely the coordination in complex team production processes: when combining their efforts, team members can produce more than when they work alone, so they have an incentive to cooperate. If there is a high interdependence between the work of the individuals, or in other words when an individual’s input cannot be clearly measured based on the collective output - the so-called metering problem -, a potential problem arises. Individuals may have an incentive to shirk (“opportunistic behaviour, ranging from cheating to not really putting

in effort” (Barney, Hesterly, Clegg, Hardy, & Nord, 1996, p. 116) because their specific

output cannot be measured. The solution Alchian and Demsetz (1972) propose is to assign a monitor – with the power to punish and/or reward team members – that holds residual claimancy, hereby reducing the likelihood of shirking by team members. This monitor can be a person, a group of persons or even an instrument.

At the foundation of this approach by Alchian and Demsetz (1972) lies the assumption that man acts as homo economicus: “… a being who desires to possess wealth, and who is

capable of judging the comparative efficacy of means for obtaining that end” (Persky,

1995, p. 321), or put otherwise, an actor is always seeking to get maximum rewards for minimal effort. Ever since the seventies, extensive research has shown that many other non-economical factors are also influencing team members’ likelihood to shirk or

willingness to cooperate, hence indicating that incentive systems that are based solely on financial rewards are not the only solution to this team production problem. Examples of alternative methods of influencing team members’ actions are cultural values (Chatman & Barsade, 1995), group characteristics (Cohen, Ledford, & Spreitzer, 1996), organizational climate (Glisson & Hemmelgarn, 1998), team composition (Beersma, Hollenbeck, Humphrey, Moon, & Conlon, 2003), group size (J. P. Carpenter, 2007) and team heterogeneity (Hamilton, Nickerson, & Owan, 2003).

This research

From a business point of view, this research is trying to contribute to the knowledge on cooperation, monitoring, routines and social dispositions, while also trying to connect them to each other, making them better applicable to daily operational situations. As discussed later, the individual topics are all very relevant to organizations, but there is not much knowledge on how they work together and interact. Ultimately, this research tries to suggest how the type of monitoring, the level of routinization and the social dispositions together influence an individual’s intention to cooperate. By doing so, it could help managers to shape their organization in order to maximize cooperation. One of the factors of this research is monitoring. As originally suggested by Alchian and Demsetz (1972), monitoring is an instrument to increase performance, due to its inhibiting effect on shirking. This effect is found in many forms by other researchers (Brewer, 1995; Brewer & Ridgway, 1998; de Jong & Elfring, 2010; Komaki, 1986; Salas, Sims, & Burke, 2005; Sundaramurthy & Lewis, 2003) and is further discussed in the literature review in this thesis. Less emphasis has been on how different types of monitoring – behaviour-based or outcome-based – affect firm performance through individuals’ intention to cooperate.

The second concept of this thesis are routines: routines are known to be the result of organizational learning (Argote, 1999) and to reduce complexity in an organization (March & Simon, 1958), thereby increasing a firm’s potential for increased performance (Feldman & Pentland, 2003). A lot of scientific effort has been put in routines: for example how they store and generate knowledge (Nelson & Winter, 1982), how they should be designed (Dyer & Singh, 1998) and how they relate to the firm’s environment (Eisenhardt & Martin, 2000). The effect of the amount of routinization on how willing individuals are to cooperate with each other is surprisingly neglected in research. Sandra Ohly, Sonnentag, and Pluntke (2006) approach this topic the closest with their research on how routinization positively influences proactive behaviours.

Besides the organizational design factors monitoring and routinization, social dispositions are the third building block in this research. In current literature, there is a strong emphasis on the fact that further research is necessary to unveil social and personal aspects that influence cooperation (Balliet, Mulder, & Van Lange, 2011; C. Boone, De Brabander, & van Witteloostuijn, 1999; Coff, 1997; Klotz, Hmieleski, Bradley, & Busenitz, 2014; Ostrom, 2014). One of the many ways to approach this is through social dispositions. This is a classification used to describe how individuals systematically prefer the distribution of outcomes in situations with high interdependency (Fehr & Fischbacher, 2002; Messick & McClintock, 1968), or in other words, do individuals consider themselves or ‘the others’ first in distributing outcomes? Why this model is interesting for this research is because it adds a dimension to an individual’s intention to cooperate: one might for example help another because the collective benefits from it, or on the opposite one might refuse to cooperate because there are no personal advantages to be gained. Additionally, measuring social dispositions is quite straightforward – please refer to the method section of this

research for an elaboration of this -, but their effect on how individuals approach cooperation can be profound. Knowing an employee’s intrinsic approach to situations of high interdependency could be an important tool for any manager in influencing cooperation in a firm.

To bring the previous sections together, there are convincing amounts of research indicating that both monitoring and routines are very relevant for firms, but their joint effect on individuals’ intention to cooperate is not yet determined. Secondly, the moderating influence of social dispositions on this effect is also unknown, despite the fact that the relevance in daily situations can be considerable.

The research question designed to capture this is the following: what is the effect of

monitoring and routinization on an individual’s intention to cooperate and to what extent is this effect moderated by social dispositions?

To answer this question, two methods are applied. First, an extensive literature research is conducted to get a good grasp of current literature standings and to further specify what gaps are there to be filled. Secondly, field research is carried out, designed specifically to answer the research question. The design of this research is the quasi-experimental vignette design, a variation to a survey based research. Rather than measuring the variables monitoring and routinization, their manipulations are presented to the participants, asking them to consider a proposed situation from framework presented to them. Please refer to the method section of this research for further specifications on this.

LITERATURE REVIEW

In this section, the constructs and hypotheses of this research will be explained based on past and current literature. The constructs that will be reviewed are Team Production, Cooperation, Monitoring, Routines, and Social Dispositions. The review is concluded with a graphical representation of the conceptual model.

Team Production

Team production theory finds its origins in the conceptual article by Alchian and Demsetz (1972). They introduce the concept, stating that it is a production process in which “several types of resources are used and … the product is not a sum of separable

outputs (p. 775)”. This creates the so-called metering problem: it is necessary to measure

an individual’s input to be able to grant him with a reward, but due to the nature of the process it is difficult to do so (at least without high costs, potentially offsetting the benefits of monitoring). At the base of this problem lies the assumption of homo

economicus: a completely rational, outcome-oriented, uncompromisingly thorough,

a-moral man that is seeking to maximize his own welfare, including “ex ante and ex post

efforts to lie, cheat, steal, mislead, disguise, obfuscate, feign distort and confuse” (Bowles &

Gintis, 1993; Gintis, 2000; Henrich et al., 2001; Williamson, 1985, p. 51). As long as there is no system to monitor the individual output and thus no individual reward system can be used, and since homo economicus likes rewards and dislikes effort, he has an incentive to shirk (also known as free-riding): “opportunistic behaviour, ranging from

cheating to not really putting in effort” (Barney et al., 1996, p. 116). The solution offered

by Alchian and Demsetz (1972) is to install a monitor who rewards the team members based on their individual input, and who has a claim on all remaining funds after all members have been paid. Holmstrom (1982) goes even a step further by suggesting that

making all team members the residual claimant of their own actions can even further improve efficiency.

Research deepening the concept of team production theory suggests a few disadvantages: the monitor for example, has an incentive to minimize the rewards for the team members, leaving a greater cut for himself (Miller, 1993). Additionally, it assumes that all actors act only in their own self interest – this is further elaborated in the next paragraph - and that the way an organization is designed will influence the potential shirking (Perrow, 1986). Among other things, Perrow (1986) also argues that the measurement of individual effort increases self-interested behaviour, the opposite of what a monitor is meant to achieve. Prendergast (1999) elaborates on this, suggesting that when it comes to rewards, “agents are also capable of actions that are privately

beneficial at the cost of overall efficiency” (Prendergast, 1999, p. 55). Larkin, Pierce, and

Gino (2012) conclude that to determine pay, pay-for-performance is a more efficient system than subjective evaluations, but that this system is not frequently applied in firms. Other issues involve the loss of freedom of choice by team members (Van Vugt & De Cremer, 1999) and the system of allocating the residual proceeds of team production (van der Heijden, Potters, & Sefton, 2009).

Relevant for this research is the fact that people often act contrary to their – expected – self-interest. William Ouchi (1980) for example already suggested clans to be another fundamental form of coordination. Additionally, manners of reciprocity and the willingness to bear costs to punish free-riders (Fehr & Gächter, 2000) are mentioned as possible reasons for this effect, as well as (perceived) fairness (Cohen-Charash & Spector, 2001; Husted & Folger, 2004), morality (Minkler, 2004), work motivation

(Tomohara & Ohno, 2016) and team identity (C. C. Eckel & Grossman, 2005). Another interesting theory is the Stewardship Theory: Davis, Schoorman, and Donaldson (1997) discuss this theory as an alternative explanation of why an individual would put in effort to achieve organizational or the principal’s goals, even when that does not primarily serve his or hers own interest. They state that so-called Stewards “whose behaviour is

ordered such that pro-organizational, collectivistic behaviours have higher utility than individualistic self-serving behaviours” (p. 24) act in the best interest of the organization,

even though that might not be their own: they simply perceive contributing to the organization – and its stakeholders – to be of higher value than contributing just to themselves. Possible explanations of this behaviour of stewards could be that they are pursuing growth, that they are intrinsically motivated or that they identify themselves with the organization. A major advantage for organizations would be that the presence of stewards can allow them to reduce agency cost associated with controlling the agents, simply because the agents already behave the way the principals want them to behave. Concluding, there are many factors or theories that provide explanations for behaviour not in-line with the expectations of homo economicus. Many of these are in some way linkable to social or psychological aspects of human behaviour, just as the social dispositions as discussed later in this research.

Cooperation

A team production setting, with high task interdependence, calls from an economic angle for a system that converges all actors (both agent’s and principal’s) goals (Alchian & Demsetz, 1972). The fact that there is always the possibility for shirking in these kinds of situations has influence on individuals’ cooperation. Ouchi (1980) described three different forms of coordination mechanisms to facilitate corporation/efficiency - markets, bureaucracies and clans - and stated that different levels of goal incongruence

and performance ambiguity are the factors that determine the suitability of each of the forms. What is problematic about situations with high task interdependence is that it often leads to a social dilemma. As originally stated by Olson (1965, p. 2) “rational,

self-interested individuals will not act to achieve their common or group interest. … even if all of the individuals in a large group are rational and self-interested, and would gain if, as a group, they acted to achieve their common interest or objective, they will still not voluntarily act to achieve that common or group interest”: the situation puts individuals

in a position where there is a conflict between their self-interest and collective interests. These kinds of situations can be found everywhere and in many contexts: in laboratories they are simulated in public good games such as the Prisoners’ Dilemma (Tucker, 1950), Tragedy of the Commons (Gibbons, 1992), Public Good Games (Ledyard, 1997) and the Ultimatum Game (Güth, Schmittberger, & Schwarze, 1982). In more recent years, field studies have been conducted to test the clinical laboratory results in reality. Inquiries were done into Japanese fishermen who collectively pool their fishing proceeds and divide the returns (J. Carpenter & Seki, 2011) and exchanges in e-commerce situations (Bahbouhi & Moussa, 2017). The ‘classic’ example by Alchian and Demsetz (1972) is also worth mentioning: they sketch a situation where two men lift cargo into trucks. Just by observing the total weight put into the truck, one can never distinguish who did what. If one of the two men would decide to shirk - thus harming the group interest of cooperating to maximize the total weight, while serving his own interest of putting in as little effort as possible – he could do this without anyone noticing. One could imagine this sort of behaviour also taking place in more sophisticated situations, where teams composed of a substantial amount of people jointly work on a project – of which they all benefit once completed – but where individual members still feel an urge to shirk and consider their own interests above that of the group as a whole. Again, hard-line

economists would expect individuals to always act in their own self-interest if that is the most beneficial option, a statement that could be debated, but it shows that social dilemmas and the serving of ones own self-interest is thus a crucial part in cooperation. Other research by Poortvliet et al. (2009), looks at how a cooperation is influenced by achievement goals. They find that the intention to cooperate is influenced by whether individuals are comparing their performance to others or to their own earlier performance, and that this effect is moderated by an individual’s rank: performance goals decrease the intention to cooperate, especially for high or low ranks (Poortvliet et al., 2009).

What all these studies have in common is that they try to connect different situational, social or personal variables to cooperation in situations with high interdependence, showing that regardless of what supporters of homo economicus would expect, people often do consider the collective interest (Van Lange, Joireman, Parks, & Van Dijk, 2013). Closely related to the later discussed social dispositions, cooperative behaviour depends on the valuation of the self-interest and the cooperative goal, and the expectation of cooperative behaviour of others involved (Pruitt & Kimmel, 1977).

Monitoring

The third construct in this research is monitoring. As discussed, monitoring is a proposed solution to the metering problem of team production theory (Alchian & Demsetz, 1972). A monitor (in the broadest sense of the word) is a system of organizational control that is designed to “to increase the probability that people will

behave in ways that lead to the attainment of organizational objectives” (Flamholtz, 1979,

p. 51), or to offer control through the reduction of information asymmetries between transaction parties (Eisenhardt, 1985). According to Flamholtz (1979), a control system should exist of four basic components, being (1) goals for performance, (2) standards of

performance, (3) method of measurement of performance and (4) method of administering rewards to motivate and reinforce performance.

Team production theory is not the only economic theory that suggests that the team production problem might be solved through monitoring, but also the related transaction cost (Williamson, 1981) and agency theory (Jensen & Meckling, 1976) propose monitoring to be the solution to opportunistic behaviour. Although monitoring seems to be a suitable solution to observe and improve employee behaviour (Chalykoff & Kochan, 1989), motivation (Stanton, 2000) and effort (Dickinson & Villeval, 2008), there is also research indicating otherwise. Bruno Frey (1993) for example, suggests that monitoring can also be conceived as a signal of distrust, hence leading to reduced work effort. This is called the crowding out effect, and is also confirmed by Costa (2003). Niehoff and Moorman (1993) state that monitoring can lead to team members not performing extra duties that are not rewarded, Deci and Ryan (1985) found that monitoring could undermine intrinsic motivation, and Holman, Chissick, and Totterdell (2002) concluded that in some cases monitoring can lead to decreased well-being. Other relevant research on monitoring describes the legitimacy or fairness of the monitoring: Charles Halaby (1986) introduces authority cost, a premium added by the subordinate to compensate for the fact that he is being monitored. If the governance mechanism is perceived to be legitimate, employees will assign less authority cost to their contract; on the contrary, if the governance mechanism is perceived to be a representation of the will of the employer, authority cost will increase (Halaby, 1986). Hence monitoring might be perceived as fair if it is in line with the agreement between the employer and the employee (Heide, Wathne, & Rokkan, 2007).

There is an obvious link between monitoring and rewards. Agency theory predicts that individuals work harder if they are paid for performance (e.g. Eisenhardt, 1989), but

this consequently requires some form of input assessment (Larkin et al., 2012). Team-based compensation is a way to get around the cost involved in individual monitoring, but it also leaves plenty of possibilities for shirking with the individuals involved (Bonin, Dohmen, Falk, Huffman, & Sunde, 2007).

When looking at the implementation of monitoring, a distinction can be made between output-based and behaviour-based monitoring: the first emphasizes largely on objective measures of results (such as the number of products sold), the latter focuses more behaviour observed during the process (Anderson & Oliver, 1987). Research indicates that output monitoring decreases opportunism and that behaviour monitoring in some cases could actually promote opportunism, particularly if it is not supported by micro level social contracts (Heide et al., 2007). Outcome-based monitoring is especially suitable for work environments where tasks are not highly programmable (Eisenhardt, 1985), but more problematic in situations of high task interdependence (Mahoney, 1992). Selviaridis and Wynstra (2015) state that outcome-based monitoring can be deployed to align goals and incentives, to spread risk across the actors involved and to emphasize bi-directional performance. Because the risk of the performance is transferred to the individual, heavy reliance on outcome-based contracts decreases behavioural relations in a firm (Hopwood, 1972).

In case the output of an individual is not easily observable, the behaviour should be observed (Barzel, 1982). Behaviour-based monitoring on the other hand, looks at the process by which an employee achieves its goals (Anderson & Oliver, 1987). It forces the monitor to be much more involved with the activities of an individual in order to be able to evaluate (Aulakh, Kotabe, & Sahay, 1996), where as technological developments have enabled them to increasingly do so (Alder, 1998). The problem with this type of

monitoring is that it is almost never perfect – as in all behaviour is observed -, it is much more complex than outcome-based monitoring (Barney et al., 1996), and it can be perceived as intrusive (Heide et al., 2007) and as limiting autonomy and self-control (Atuahene-Gima & Li, 2002). When comparing this to outcome-based monitoring, behaviour monitoring is a much more suitable option for solving the metering problem, because it unveils an individual’s contributions to the collective (Barney et al., 1996). Because individual behaviour is taken into account with behaviour-based monitoring, there is an incentive for individuals to be engaged in behaviour that benefits the collective, because they can be punished or rewarded for that behaviour (Christophe Boone, Declerck, & Kiyonari, 2010). Based on this assumption the first hypothesis is proposed:

H1: An individual’s intention to cooperate is higher when behaviour is monitored than when output is monitored.

Routinization

Routines are generally defined as “repetitive, recognizable patterns of interdependent

actions, carried out by multiple actors” (Feldman & Pentland, 2003, p. 94) and have been

a significant part of strategic organizational literature. They are a result of organizational learning (Argote, 1999) because for example organizational members will prefer easier solutions to situations over harder ones (Pentland & Rueter, 1994) and exist because they reduce complexity (March & Simon, 1958); they are a way for organizations to store knowledge (Huber, 1991; Nelson & Winter, 1982). Nelson and Winter (1982) further state that for routines to be effectively used in organizations, all individuals involved must be familiar with the routines relevant for their jobs, and know when to use them. If organizations are able to use routines this way, they can become involved in very complex processes, while still remaining coordinated (Argote, 1999).

The problem with routines is that because of their knowledge and skill retaining character, they reduce variance (March, 1991). In dynamic environments variance reduction can be harmful because the reduction might lead to a misfit between a firm’s current capabilities/competences and those required by the environment (Eisenhardt & Martin, 2000). Current literature sees routines as coordination mechanisms that provide the people involved with a common perspective (Edmondson, Bohmer, & Pisano, 2001; Okhuysen & Bechky, 2009; G. A. Okhuysen, 2005), or as a mechanism that allows for easy integration of individuals’ highly specific knowledge into the organization (Grant, 1996).

As behaviour becomes more automatic, it routinizes (Ohly et al., 2006). Performance of the task becomes faster (Wickens, Hollands, Banbury, & Parasuraman, 2015) and requires less cognitive processing capacities (Norman & Bobrow, 1975). Especially this last bit is relevant for this research and will be discussed later. What is interesting to note about routinization is that it can be interpreted in two ways. On the one hand, routines reduce variation: the range of behaviour by employees is reduced (Ford & Gioia, 2000); on the other hand, Kanfer and Ackerman (1989) state that due to routinization, resources are freed allowing them to be spend for other purposes. A positive relationship is found between routinization and creativity for example (Sonenshein, 2016), especially when employees have high levels of job control (Ohly et al., 2006). Routines are thus an important asset for cooperation in firms: they “give

structure to collaborative exchanges, and so provide a means by which particular exchanges are integrated into a broader collaborative enterprise” (Lazega & Pattison,

Routines can happen at all sort of levels. Much research has been done into for example interorganizational routines (Dyer & Singh, 1998; Zollo, Reuer, & Singh, 2002), but this research focuses on how routinized or automatized the work process is. Routinization was first defined in terms of standardization of work procedure as well as work pattern through the introduction of rest pauses (Taylor, 1947). Bargh (1994) classified four features of automaticity: unintentionality, uncontrollability, lack of awareness and efficiency. A routine is formed – and reinforced through positive feedback – in three steps. First the repetitive execution of tasks or behaviour that meets the features of automaticity is the primary source for routinization (J. R. Anderson, 2000; Betsch, Haberstroh, Glöckner, Haar, & Fiedler, 2001). Subsequently, repetitive past behaviour contributes to future intentions in a comparable context, which makes it an indicator for future behaviour (Ouellette & Wood, 1998). Lastly, when a familiar task is presented to an individual, it automatically triggers the behaviour pattern associated with the task: the individual does not consciously starts the action (Ohly et al., 2006). These three principles combined form the process of creating the routine. Baba and Jamal (1991) make a distinction between routinization in job content (based on variety in job characteristics by Hackman and Oldham (1975)) and job context (for example permanent working hours or rotary shifts). Routinization of job content, based on the four features by Bargh (1994) will be used as a variable in this research.

Above summarization of current literature supports the fact that high levels of routinization imply high levels of automaticity. In turn, this leads to lower levels of cognitive load. Schulz, Fischbacher, Thoni, and Utikal (2014) found that cognitively high loaded individuals are more generous in resource allocation in dictator games. They ascribe this to the fact that “basic social preferences are fundamental” (p. 84) and thus

higher cognitive load forces individuals to rely on their routinized or automated behaviour. Roch, Lane, Samuelson, Allison, and Dent (2000) and Cornelissen, Dewitte, and Warlop (2011) found similar results, stating that low cognitive load leaves individuals with enough cognitive capabilities to adjust the situation to their advantage. Important to note here is the fact that the effect found in these studies is mainly applicable to one-shot situations.

Considering the above review, one could argue that it is possible that individuals engaged in highly routinized work – and thus low cognitive load - have more ‘spare’ cognitive capacities and are therefore more likely to cooperate with other individuals. This does not necessarily say that individuals engaged in less routinized work would not be willing to cooperate, but it could be plead that in those situations the possibility of helping another could have a negative impact on ones own work which might be a reason not to be helpful. This leads to the following hypothesis:

H2: An individual’s intention to cooperate is higher when the individual’s work is highly routinized.

An important aspect of this research is the interaction between monitoring and routinization. It is proposed here that these two factors can amplify or weaken each other. As stated in the previous section about routinization, when an individual is engaged in none-routinized work, it has little ‘spare’ cognitive capacities to help another individual. Even if he or she would like to help, the high cognitive load would impede this. This inability to help is be exacerbated when the individual’s output is monitored. His or her superior is not likely to ever find out that help was given to a fellow employee, so no indirect benefits based on this helpful act could be achieved. This reasoning can also be applied the other way around: an individual with highly routinized work – and thus a low cognitive load – whose behaviour is monitored, might have an extra incentive

to be engaged in cooperative behaviour; not only has he or she cognitive capacity to share, personal benefits can be gained when the monitor registers this behaviour.

Continuing on this reasoning, it could be stated that some combinations may have a weakening effect on each other. Expected is that these effects will be a lot less distinct though. An individual engaged in none-routinized work – with a high cognitive load –, but with behaviour-based monitoring could be tempted to help another employee, even when it might have a negative impact on his or hers own work. It is to note though, that there are many conditions imaginable that would influence this. One can imagine for example that a monitor that emphasizes personal deadlines might be less sensitive for cooperative behaviour than one that focuses more on team deadlines. The last combination is that of an individual performing highly routinized work who is monitored based on output. The managerial appreciation for his or hers work is not likely to be influenced by cooperative behaviour, but the individual does have the ‘spare’ cognitive capacities to help another. Again, the last two weakening effects are probably much less distinct than the first two, which is why only the first will be specified in a two-sided hypothesis:

H3: an individual’s intention to cooperate is lower when the individual’s work is not routinized and when output is monitored, than when routinization is high and behaviour is monitored.

Social Dispositions

Nuancing the strict a-morality of homo economicus, social dispositions (or social preferences or social value orientation) are a classification used to describe how individuals systematically prefer the distribution of outcomes in situations with high interdependency (Fehr & Fischbacher, 2002; Messick & McClintock, 1968). They have been identified as a strong indicator of an individual’s cooperative motives, choice

behaviour and strategies (Kollock, 1998). A three-category distinction based on this orientation can be made: Prosocials maximize outcomes for both themselves and other while trying to maintaining maximum levels of equality; individualists maximize their own outcome without regarding other’s outcomes; and competitors seek to maximize the relative difference between them and others (De Cremer & Van Lange, 2001; Van Lange, De Bruin, Otten, & Joireman, 1997). Individualists and competitors are often combined into one group, called proselfs (Smeesters, Warlop, Van Avermaet, Corneille, & Yzerbyt, 2003; Van Lange & Liebrand, 1991). Other classifications are by Thibaut and Kelley (1978) (seven orientations) and MacCrimmon and Messick (1976) (ten orientations). The two value distinction (prosocial/proself) is used here, because it has the most empirical background (Au & Kwong, 2004). Research by Wallace, Cesarini, Lichtenstein, and Johannesson (2007) showed that at least 40% of an individual’s social preference is assignable to genetics and a relatively stable trait (Eisenberg et al., 1999; Sheldon & Elliot, 1999).

Research on the difference between prosocials and proselfs has some interesting outcomes. First of all, there is a fundamental difference in how individuals approach a social dilemma in their ex ante motivation to cooperate and the way the interaction is perceived and evaluated (Bogaert, Boone, & Declerck, 2008): when it comes to cooperation, prosocials regard cooperative behaviour as rational when it increases the collective outcome (they reason from collective rationality); proselfs naturally tend to shirking behaviour, because that leads to the best personal outcome (they assume individual rationality) (De Bruin & Van Lange, 1999; Ostrom, 1998). In line with this, prosocials – who value fairness, honesty an equality – reason more from a moral standpoint, where proselfs – who value dominance, personal values and potency – act more from a power dimension (Liebrand, Jansen, Rijken, & Suhre, 1986). These

differences in cooperation between prosocials and proselfs have been proven both in public good games as in the real world (Levitt & List, 2007; Pruyn & Riezebos, 2001; Wallace et al., 2007). As for expectations of behaviour, proselfs have a more homogeneous view of others and consequently expect others to behave individualistic or competitive as well (just as they would), while prosocials adopt a more diverse perspective, hence expecting much more cooperation from their partners (Smeesters et al., 2003). What is important to note though, is that cooperation by prosocials is not unconditional: an (expected) non-cooperative partner activates a sense of reciprocity and decreases the levels of cooperation for prosocials (De Cremer & Van Lange, 2001). Co-operators are even willing to decrease their own benefits in order to punish those that free-ride (Fehr & Fischbacher, 2002).

Based on the previous theoretical frame, the social dispositions can be tied to an individual’s intention to cooperate. In general, it is expected that individuals that are classified as prosocial are more likely to cooperate than individuals classified as proself:

H4: An individual classified as prosocial is more likely to cooperate than individuals classified as proself.

When the dispositions are added to the effect of monitoring on intention to cooperate, individuals classified as proself are expected to cooperate less in situations with outcome-based monitoring: it is not in their best interest to help someone else, because they can never be rewarded for that specific behaviour when it cannot be observed. Hence, when the specific actions of an individual are monitored (behaviour-based), more cooperation is expected from proself individuals. As said, prosocial individuals are expected to be more likely to cooperate. They have a fundamental tendency to cooperate, and so the impact of their disposition on their intention to cooperate in any of

the two monitoring situations is less than that of the proself individuals. When considering the impact of social dispositions on the effect of routinization on cooperation, it is expected that prosocials are more willing to cooperate than proselfs, both for situations of low and high routinization. Most interestingly is the impact of social dispositions on the combined effect of monitoring and routinization. Predicted is – see also the reasoning for hypothesis H3 - that none-routinized work, combined outcome-based monitoring has a negative effect on the intention to cooperate: an individual needs all of his cognitive abilities to perform his own task, and helping someone else without ever being rewarded for it is therefore not interesting. This effect is expected to be even stronger when an individual is classified as proself. On the opposite, high routinization, behaviour-based monitoring and a prosocial individual is a situation where a high intention to cooperate is expected. To summarize the above statements in hypotheses:

H5a: Social dispositions interact with the effect of monitoring on an individual’s intention to cooperate. Namely, an individual classified as proself is more likely to cooperate in situations coordinated with behaviour-based monitoring than in situations with outcome-based monitoring.

H5b: Social dispositions interact with the effect of routinization on an individual’s intention to cooperate. Namely, an individual classified as prosocial is more likely to cooperate in situations of high routinization than in situations with low routinization.

H5c: Social dispositions interact with the combined effect of monitoring and routinization on an individual’s intention to cooperate. Namely, the effect of hypothesis H3 is amplified when an individual is classified as proself and reduced when an individual is classified as pro-social: it is expected that an individual classified as pro-social will have a high intention to cooperate, even when routinization is low and output is monitored

Methodology

The following section contains an explanation of the chosen research design, a description of the sample, a report on all the variables and an annotation of all the manipulations applied to the dataset.

Design

This research is investigating the effect of type of monitoring, level of routinization and social dispositions on individual’s intention to cooperate in situations of high interdependence, and how these factors all relate to each other. Such a setup is described by Saunders, Lewis, and Thornhill (2014) as a research with an explanatory nature, which is examined very well with a survey based research design. A survey is a method “most frequently used to answer ‘what’, ‘who’, ‘where’ (…) and ‘how’ questions (…)

and allow the collection of standardised data from a sizable population (…) allowing for

easy comparison” (Saunders et al., 2014, p. 176). These properties are very closely

related to the necessities of this research, hence making this research method very suitable. Surveys come in many forms and shapes, one of them being a so-called vignette study (also referred to as factorial survey approach), the design chosen for this research. This vignette design, originally proposed by Rossi, Sampson, Bose, Jasso, and Passel (1974), uses standardized vignettes (fictive descriptions) in which “selected

characteristics describing the objects to be judged by respondents are simultaneously manipulated” (Wallander, 2009, p. 505). It is a quasi-experimental study that resembles

to traditional survey research, but with one big difference: rather than requiring the subject to imagine the proposed situation as with a survey, the situation is presented to them in simple, direct and abstract details (…) thereby standardizing it across respondents (Alexander & Becker, 1978, p. 94). After reading the vignette, the subjects are asked to fill in a survey with questions about their intention to cooperate (the dependent variable), as if they were in the situation just described to them. Since the frame the subjects refer from is set by the vignette they just read, the judgments and opinions of the subjects can be reliably measured (Alexander & Becker, 1978; Rossi & Anderson, 1982). A disadvantage of a vignette study is that in principle it is a hypothetical study: the participants are asked to answer the question as if they were in a particular situation, when in reality they are obviously not (Weibel, Rost, & Osterloh, 2007). The complexity of a real life situation however, makes it hard to control for all variables, which is an advantage of a vignette study (Weibel et al., 2007), because it allows the researcher to describe – and thus control – as many variables as he or she sees fit. Additionally, this design requires a large number of participants, because sufficient participants have to be assigned to each of the conditions (Kirk, 1982).

The survey is drafted using the online survey tool Qualtrics, which is freely available for UvA students. This online tool is chosen because it allows for easy creation of the survey. The survey itself is distributed both online and offline (printed). The complete survey, and a link to the digital one can be found in Appendix 1.

Sample

The survey is distributed among all students currently enrolled in the Executive Programme for Management Studies (EPMS) of the Faculty of Economics and Business of the University of Amsterdam. Both the premaster and the master students are approached, resulting in a population of about 350 persons. Based on the most recent information (September 2016) provided by the University’s Secretarial Service, the profile of this sample is 35% female and 65% male, with an average age of 31 (26% being younger than 26 years old, 25% older than 34). The majority (74%) of this sample has over 6 years working experience, but based on the profile of students who are generally attracted to this programme, it can be safely assumed that all individuals in this sample have some form of working experience (this statement is tested in the survey). This is the primary reason why this population was chosen, because they allow for probability sampling (the population is known before the distribution of the survey (Saunders et al., 2014)), but at the same time come from different sectors so that possible biases within sectors could be corrected for. Zooming in on these sectors, 17% of the population work in financial services, 34% in trade and industry, 30% is active in business services and 19% in the non-profit or government sector.

The participants were approached when they were attending classes at the University. They were presented with a printed copy of the survey – and a link to the digital survey if they prefer to fill in the survey digitally - and were asked to complete it during the break of their classes or in their own spare time. To encourage the students to complete

the survey, they ere offered a cup of coffee after completion of the survey. This method of survey distribution is chosen because a list of email-addresses of all students could not be obtained due to the University’s privacy restrictions. Based on this – physical - way of data collection, Saunders et al. (2014) predict a response and completion rate of around 50 per cent, resulting in approximately 165-185 completed surveys, but over 200 completed surveys were preferred. Assuming that all participants are randomly (but equally) allocated to one of the four vignette conditions, each condition is expected to have about 45-50 respondents. The data collection has taken place in April 2017.

221 Participants completed the survey, three surveys were not analysed because they were empty and three were removed because the particular participants indicated they were not familiar with the situation described to them in the vignette, so the total sample size is 215, of which 126 are males (58,6%) and 89 females (41,4%). Further details on age distributions, years of working experience and sectors can be found in Appendix 3. All participants were then randomly assigned to one of the four vignette conditions, 58 to vignette 1, 53 to vignette 2, 52 to vignette 3 and 52 to vignette 4. The conditions were split into the new variables monitoring (behaviour/output) and routinization (high/low), see also Appendix 4 for the distribution of the variables.

Variables

In this section, the measurement of the variables will be presented. Some modifications were applied to some variables after the participants completed the survey. The

documentation of this can be found in the next section “Manipulations”. The reason that the variables are described both pre- and post-measurement, is to be able to give a complete representation of the data collection process.

Independent variables

In this study, the independent variables (type of monitoring and level of routinization) were manipulated. The partitioning of the variables is made based on the literature review in this research. A manipulation of each of the two variables is assigned to each vignette. For example, vignette 1 describes a situation with behaviour based monitoring and high levels of routinization: see Table 1. Both the independent variables are thus expressed as categorical variables with a nominal sub-distribution (Jasso, 2006; Saunders et al., 2014).

High levels of

routinization routinization Low levels of Behaviour based

monitoring Vignette 1 Vignette 3 Outcome based

monitoring Vignette 2 Vignette 4

Table 1 Partitioning of variables

The vignettes were assessed and corrected by three independent reviewers to check their validity and reliability in advance. Participants were first introduced into the fictive company “Alpha”. Secondly, they were explained that they work in an interdependent team situation. After this, the first variable ‘level of routinization’ is presented (high or low), followed by the second manipulated variable ‘type of monitoring’ (behaviour based or output based). The explanation of variable ‘level of routinization’ is based on the definitions in the literature, consequently consisting of terms like “awareness and efficiency” (automaticity by Bargh (1994)) and repeated execution (Betsch et al., 2001). For the manipulation of monitoring, emphasis is put on how a supervisor is assessing the participant’s performance: is the primary concern of the supervisor the output of the participant or the process by which any output is achieved? To further emphasise the manipulation, it is explicitly explained that the opposite type of monitoring is not

relevant, so for the behaviour-based manipulation, output is not relevant and vice versa. After the participants have read the vignette, they continued to the section of the survey where their intention to cooperate is measured. The complete vignettes can be found in Appendix 1.

Dependent variable

The dependent variable “intention to cooperate” was measured using a four-item questionnaire, where the subjects indicated their intention to cooperate on a seven-point Likert-scale considering the situation just described to them in the vignette. The first two items inquired how likely and motivated to cooperate the participant would be, the third and fourth item were reversed coded, asking how unlikely cooperation would be (see Appendix 1 for the complete measure). This measure for cooperation is based on research by Poortvliet et al. (2009) and Bridoux and Stoelhorst (2013), who adapted the measure from Constant, Kiesler, and Sproull (1994). The small adaptations were made to increase the readability and the suitability to this thesis. After completing this section, the participants continued to the measurement of their social dispositions.

Moderator

After the vignette section of this research, all participants were presented with a ‘game’ to measure the moderator “social dispositions”. The same game is used as presented in the model described by Van Lange et al. (1997, p. 746) and is reliable for measuring an individual’s social dispositions. First the participants read a short explanation of the ‘game’, including an example, where they are explained that they will have to distribute points between themselves and another person they have never met and never will meet (see Appendix 2 for the original measure). The way participants distribute the points in nine different cases is a valid indication of their social dispositions in one of three categories: prosocial, individualist or competitor. As discussed earlier in this

research, the categories ‘individualist’ and ‘competitor’ are combined into a new category called ‘proself’. The participants now reach the last section of the survey, where the control variables are measured.

Control variables

In the last section of the survey, the participants were asked for their age (Sveiby & Simons, 2002), gender (Catherine C Eckel & Grossman, 1998), years of working experience (Sveiby & Simons, 2002) and the sector they work in. This is done to be able to control for the fact that these factors may influence the effect of the dependent on the independent variables (Saunders et al., 2014) and to see if a representative sample of the population is tested – compared to the data of the population provided by the University. Additionally, the subjects were asked if the situation they have been presented to in the vignette is something they encounter in their daily life, in order to test the reliability of the measurement.

Manipulations

To ensure a correct analysis of the data, a few manipulations have to be applied to the data set. The first manipulation was to remove all surveys in which the participant indicated that they had not been (and never were) familiar with the situation described in the vignette. Secondly, a variable had to be created to represent the social disposition category a participant was classified as. This proved to be a difficult procedure to perform in SPSS and was therefore conducted in EXCEL software, in which a formula was constructed which identified the number of “prosocial”, “individualist” and “competitor” answers given by the participants. The category that had six or more answers was the category assigned to the participant. Borderline cases were manually assigned or discarded. This resulted in a new variable “SD CAT” in which the numbers 1, 2 or 3 became a representation of one of the three social dispositions categories the

participants were assigned to, with 1 being the prosocial, 2 the individualist and 3 the competitor category. This classification was later adjusted by combining categories “individualist” and “competitor” into a new category “proself”, that combined with “prosocial” formed the new variable “Social Dispositions”. The third step in preparing the data was recoding two of the four items that compile the dependent variable Intention to Cooperate. The third and the fourth item were recoded, reversing the answers given on the seven-point Likert scale. Hereafter, a new variable “DV MEAN” was created, to represent the mean of the four items of the dependent variable. A reliability analysis of this new variable resulted in a Cronbach Alpha of α = .784, which was increased by removing the third and fourth item (new α = .823), resulting in the new dependent variable “Intention to Cooperate”.

Results

This results section holds an extensive description of the analyses of the data. It starts with a correlation analysis of the dataset, followed by more specific analysis of variance of the hypotheses. In the last part the control variables are reviewed. All analyses are conducted using the IBM’s statistical software SPSS, version 24.

Correlation analysis

The first analysis is a correlation analysis of the entire model, to measure the strength of the relationship between all the variables (Field, 2015). The complete correlation matrix, including numbers, means and standard deviations can be found in Table 2.

N M SD 1 2 3 4 5 6 7 1. Intent Cooperate 215 _{4.57 1.29} 2. Monitoring 215 _{1.49 0.50 0.048} 3. Routinization 215 _{1.53 0.50 -0.035 0.001} 4. Social Dispositions 210 _{1.49 0.50 .383** 0.056 0.01} 5. Age 215 _{2.81 0.87 0.039 0.048 -.151* 0.059} 6. Gender 215 _{1.41 0.49 0.051 0.048 0.012 -0.063 -0.026} 7. Work experience 215 _{2.93 1.23 0.09 -0.009 -.255** .153* .800** -0.087} 8. Sector 215 _{2.94 1.52 0.093 .137* 0.011 0.002 .142* .276** .167*} Table 2 | Correlation Matrix. Note that at **, correlation is significant at the 0.01 level (2-tailed) and at *, correlation is significant at the 0.05 level.

A significant positive correlation (.383, p<.01) can be found between social dispositions and the dependent variable intention to cooperate. The variable social dispositions is coded in a way that the category prosocial would correspond to a positive correlation, so based on this it can be determined that there is a positive correlation between higher intention to cooperate and a classification as a prosocial. This correlation is the only correlation to be found between the main independent and dependent variables. There are a few correlations though to be found among the control variables. The positivity or negativity of the correlations depends on the coding of the categorical variables and – if relevant – is discussed here. An overview of the coding of all categorical variables can be found in Appendix 5. The correlation between Age and Routinization (-.151, p<.05) is based on coincidence. Routinization was not measured - but manipulated in the vignettes -, and even though the participants were randomly assigned to the manipulations, an effect appears, suggesting a correlation between age and routinization. Because the participants had no influence on the variable routinization (and neither on age of course), this correlation can be attributed to coincidence. The same goes for the correlation between working experience and routinization (-.255, p<.01). Another, quite logical, correlation is the one between age and working experience (.800, p<.01). Intuitively, it makes sense that participants in higher age

categories have a higher probability of having more working experience, which is confirmed by this high positive correlation. The last correlation to be discussed here is the correlation between gender and sector (.276, P<.01). A Pearson chi-square test confirms this correlation (χ (4) = 19,890 p = .001), the frequencies can be found in Table 3.

Sector

Total Financial

Services Trade and Industry Business Services

Non-profit and Government Other Gender Male 44 17 27 23 15 126 Female 17 8 15 17 32 89 Total 61 25 42 40 47 215

Table 3 | Gender and sector cross tabulation

Analysis of Variance

To analyse the specific hypotheses, two different methods of analyses were conducted: to test specific isolated effects one-way ANOVA analyses are run, without controlling for other variables. Additionally factorial ANOVA analyses were conducted, analysing the entire model including interaction effects and control variable influences. The specific values of the effects can be a little different for one-way and factorial ANOVA, but no differences in significance are found. At first, Levene’s test for equal variances is ran for the entire model – with all variables and control variables. No significant result (p = .083) is found, so we assume that the variance of all the variables are equal, allowing us to compare them without further transformations. Having determined the equal variances between all variables, we can start testing our hypothesis, the results of which will be discussed in the next section.

The first hypothesis (H1) is proposing that there is a difference between participants’ intention to cooperate when behaviour is monitored than when output is monitored. This hypothesis is tested using the one-way ANOVA. The results show no statistically significant difference between the groups (F (1,213)= .500, p = .480)).

The second hypothesis (H2) is comparing the effects of level of routinization on intention to cooperate. No significant differences are found between the two conditions, indicating that participants experiencing high or low level routinization indicate no statistically significant intentions to cooperate (F (1,213) = .261, p = .610)).

The third hypothesis (H3) is testing the combined effect of the previous two variables, to see whether the combined effect of type of monitoring (behaviour or output) and level of routinization (high or low) has an effect of intention to cooperate. The results of the factorial ANOVA indicate no statistically significant effect of the combined variables on the intention to cooperate (F (3,211) = .018, p = .893). When looking at the means of the interaction effect, this statistically very non-significant effect can be confirmed, because it can be seen indeed that all the means are very close to each other (see Table 4). The other specifications of the results of H1, H2 and H3 can be found in Appendix 6.

Monitoring Routinization Mean Std. Error Lower Bound Upper Bound 95% Confidence Interval

Behaviour Low 4,548 ,180 4,194 4,902

High 4,483 ,170 4,147 4,818

Output Low 4,696 ,181 4,339 5,054

High 4,583 ,176 4,236 4,931

The fourth hypothesis (H4) is predicting an effect of social dispositions on intention to cooperate: individuals classified as prosocial are expected to indicate more intentions to cooperate than individuals classified as proself. The one-way ANOVA test indicates an effect of social dispositions on intention to cooperate (F (1,208)= 35.694, p = .000), and further elaboration (means plot (Figure 2) and contrast test (see Table 5, equal variances assumed – Levene p = .065) of this effect shows that the direction of the hypothesis is confirmed as well; participants that are classified as prosocial have a statistically significant higher (t = 5,974, p = .000) mean score on intention to cooperate (M=5.098, SD = 1.099) than individuals classified as proself (M = 4.107, SD = 1.29), see also Appendix 7.

Figure 2 | Means plot H4

Contrast Tests Contrast Value of Contrast Std. Error t df Sig. (2-tailed) Intention to Cooperate Assume equal variances 1 ,9916 ,16597 5,974 208 ,000

Does not assume equal variances

1 ,9916 ,16521 6,002 205,805 ,000

Hypothesis H5a, b and c are also tested using factorial ANOVA and are variations to hypotheses H1, 2 and 3, with the addition of the moderator “social dispositions”. The combined effect of the moderator and monitoring on the dependent variable (H5a) resulted in F(7,202) = .735, p = .392, the combined effect of routinization and the moderator (H5b) produced F(7,202) = .936, p = .334. Finally, the results neither indicate any statistical reason to assume H5c, measuring the combined effect of the two variables and the moderator on individual’s intention to cooperate (F(7,202) = .825 , p =.4365). The variance in the model explained by the interaction effects of H5a, b and c is also very marginal. H5a produces an η2 _{of .004, H5b of .005 and H5c of .004. The complete results}

can be found in Table 7. What is to consider about the hypothesis H5a,b and c is that, though not significant, a tendency can be found when looking at the means of participants’ scores on intention to cooperate: in all conditions the means for individuals classified as prosocial are higher than those classified as proself (see also Table 6 and Appendix 8). The interpretation of this can be found in the discussion section.

Monitoring Routinization Social Dispositions Mean Error Std.

95% Confidence Interval Lower

Bound Bound Upper

Behaviour Low Proself 4,000 ,233 3,541 4,459

Prosocial 5,140 ,242 4,662 5,618

High Proself 4,000 ,214 3,578 4,422

Prosocial 5,120 ,242 4,642 5,598

Output Low Proself 4,135 ,237 3,666 4,603

Prosocial 5,292 ,247 4,804 5,779

High Proself 4,348 ,252 3,850 4,846

Prosocial 4,875 ,229 4,424 5,326

Table 6 | Distribution of means for H5c. Mean of intention to cooperate.

To summarize all the above results, a significant effect of social dispositions on the intention to cooperate is found (H4), the other hypotheses are not confirmed.

Tests of Between-Subjects Effects Source Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared Corrected Model 55,953a _{7 7,993 5,452 ,000 ,159} Intercept 4428,634 1 4428,634 3020,538 ,000 ,937 Monitoring ,492 1 ,492 ,336 ,563 ,002 Routinization ,162 1 ,162 ,111 ,740 ,001 Social Dispositions 50,574 1 50,574 34,494 ,000 ,146 Monitoring * Routinization ,109 1 ,109 ,075 ,785 ,000 Monitoring * Social Dispositions 1,078 1 1,078 ,735 ,392 ,004 Routinization * Social Dispositions 1,373 1 1,373 ,936 ,334 ,005 Monitoring * Routinization * Social Dispositions 1,209 1 1,209 ,825 ,365 ,004 Error 296,167 202 1,466 Total 4772,750 210 Corrected Total 352,120 209 a. R Squared = ,159 (Adjusted R Squared = ,130)

Table 7 | Test of between-subjects effects for H5a, b and c

This last part of the results section will deal with the control variables. For completeness, a factorial ANOVA is run to see the individual effects of the control variables on intention to cooperate. Results show a significant effect of working experience on intention to cooperate (F(4,213) = 2.642, p = .035). The results indicate no other effects of the other variables on intention to cooperate. Further specifications through contrast tests of the specific levels of working experience indicate that having between 11 and 20 years of working experience is an indicator for significantly higher intention to cooperate (t = 2.353, p = .020), as compared to the other groups (see also Figure 3 and Appendix 9).

Figure 3 | Means plot for working experience

Discussion

This research is intending to answer the question how individual’s intention to cooperate is influenced by the type of monitoring, the level of routinization and individual’s own understanding of team based situations – expressed through the “social dispositions” construct. Based on the literature review conducted earlier in this thesis, seven hypotheses were drafted, suggesting effects between and among type of monitoring (behaviour or output), levels of routinization (high or low), social dispositions (proself or prosocial) and team members’ intention to cooperate. This research was only able to appoint a statistically significant effect between social dispositions and intention to cooperate, where individuals classified as prosocial were to have more intentions to cooperate than individuals classified as proself.

When looking at the effect of monitoring, fundamental research suggests that cooperative behaviour is stimulated by the installation of a monitor (e.g. Alchian & Demsetz, 1972). Continuing on this statement, this research has found no evidence that the type of monitoring – whether output-based or behaviour-based – significantly influences an individuals’ intention to cooperate, despite the fact that research makes a

clear distinction between the two types (e.g. Anderson & Oliver, 1987; Selviaridis & Wynstra, 2015). The same goes for routinization and cooperation. Routines have been proven to be an essential aspect of any organization (e.g. Feldman & Pentland, 2003) and by testing the effect of routinization (how routinized behaviour is, how much cognitive capabilities it requires (Norman & Bobrow, 1975; Ohly et al., 2006)) on cooperation this research has tried to deepen the knowledge on routines. Though previous research has found some effects of surplus cognitive capabilities having an effect on cooperation, these findings are not supported by this research – please refer to the limitations section of this chapter for some remarks to this.

In line with what most research predicts, this research found a significant difference between prosocial and proself individuals and their intention to cooperate. Due to their fundamentally different approach to social dilemmas, prosocials are expected to demonstrate much more cooperative behaviour than proselfs do (Bogaert et al., 2008). Rather than putting their own personal interests first, prosocials consider the best option for the group before considering their own (Liebrand et al., 1986). In the context of this research, it can be concluded that prosocial individuals are significantly more likely to help a colleague, not necessarily for their own benefits, but because the group as a whole will benefit from that act.

Implications

One of the scientific implications of this research is that it is contributing to current literature with a very specific examination of individuals’ intention to cooperate. It is trying to tie well known and much described topics (monitoring, routinization, social dispositions and cooperation) to each other in an effort to make them less abstract and more usable in the daily operations of an organization. A few very interesting remarks can be made. First of all, regardless of what agency theory expects, no effect is found