The effect of different payment structures on the scale of individualism-collectivism : a study in which individual performance pay and team performance pay is analyzed in comparison with the scale of individualism-coll

(1)

The effect of different payment

structures on the scale of

Individualism-Collectivism

A study in which individual performance pay and team performance pay is analyzed in comparison with the scale of individualism-collectivism

Stef van der Schaaf - 10868697

Master Thesis Business Economics (Organisation Economics) Supervisor: Jeroen van de Ven

University of Amsterdam July 7, 2016

(2)

2

Statement of Originality

This document is written by Stef van der Schaaf who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3 Abstract

An extensive body of literature focuses on which payment structure is more effective or favoured in either an individualistic or collectivistic country. Also, explanations have been given for the differences in degrees of individualism-collectivism across countries. However, payment structures are not ‘on that list’. In this study, I want to investigate whether an individual payment structure makes you more individualistic and whether a group payment structure makes you more collectivistic. I tested this at 137 high school students, using a within-subject design. Results showed that after a group assignment within-subjects were significantly more individualistic than before. A possible explanation is that group members’ skills are unequally distributed, which is observed by all group members. This could have lowered the overall believe in the group, which made these subjects more individualistic. After the individual assignment, individualism-collectivism averages pointed at being more collectivistic than before the assignment – although not being significantly different.

(4)

4

1 Introduction

In the 1960s Hofstede started with categorizing countries at several dimensions (Beenhakker, 2011). One of those dimensions looks at how individualistic or collectivistic people tend to think. This degree of individualism-collectivism [INDCOL] has been tested extensively by other authors as well (e.g. Oyserman, Coon & Kemmelmeier, 2002; Vanello & Cohan, 1999). In the same way as Hofstede by comparing it across countries or cultures as well as in relationship with other fields, like environmental beliefs and behaviour (McCarty & Shrum, 2001) and aggression (Li, Wang, Wang, & Shi, 2010). Also, attention has been paid to the relationship between INDCOL and reward systems. Not unanimously, but the most common finding is that group rewards are favoured in more collectivistic countries, whereas individual rewards are favoured in more individualistic countries (Hofstede, 1994; Ramamoorthy & Caroll, 1998; Schuler & Rogovsky, 1998). However, in all these studies correlations are studied instead of causal relationships. Therefore, we do not know whether ‘the man determines the reward system’ or ‘the reward system determines the man’. This can be formulated as the following two questions as well: i) Does the structure of payment systems (and assignments) also contribute to people becoming more individualistic or collectivistic?; ii) Or do people that are more individualistic or collectivistic simply prefer different payment structures?

Whether ‘the reward system determines the man’ will be the main focus in this study. So in this study I will investigate whether a certain reward system changes one’s INDCOL level. Based on this information and motivation, the following will be the main research question: What is the change, if any, on the Individualism-Collectivism Scale when employees will be paid according to a team performance pay instead of an individual performance pay? Whether ‘the man determines the reward system’ will be addressed as well, but in a more brief way.

A practical implication of this matter might be the following. Fallick, Fleischman & Rebitzer (2006) describe that a cluster as a whole may profit from ‘the job-hopping phenomenon’. This is something that makes sense according to almost all economic principles which emphasize the importance of the allocation of goods and skills at the right places. However, what Fallick et al. also describe in the same paper, is that at the more individual firm level this job-hopping may have negative implications. An example they give is the lowered willingness to invest in human capital when the probability of separation between employer

(6)

6

and employee increases. This of course has an effect at the firm itself, since its employees become less re-trained over time. Job-hopping can in my opinion be considered as individualistic behaviour, since one puts itself ahead of its current firm by leaving and thereby making both capital and time investments in him of no more value. Hofstede (1994) also describes that individualistic countries favour job-hopping, where collectivistic countries do not. If a firm would be willing to change its employees’ morale, it might consider switching to a different reward system in order to for example create more commitment to the firm.

The above implication is quite a specific one, when focusing on job-hopping. More broadly this can be viewed as the problem sketched by Roberts (2004). He argues two (broad) sorts of behaviour are demanded from employees: initiative and cooperation. The former refers to amongst others pursuing one’s own goals and responsibilities, whereas the latter one promotes the opposite. When a shift among this frontier is desired, this might be accomplished by changing the reward system.

In this study I used a survey that determines the level of INDCOL. I ran a within-subject experiment, in which I tested the level of INDCOL twice. However, before the second test, the subjects either did an individual or a group assignment. Now I could check any possible differences which then could be attributed to the task, using both non-parametric tests and regressions. The tests were ran twice: once with the complete survey and once with a reduced survey (a selection of the questions). The survey consists of questions regarding for example the individual, his group, his family. Questions that apply to family issues among others were left out when testing with ‘the selection’, since they are expected to differ less after the assignment than questions about the person himself or his group. In the ‘complete condition’ the results are almost opposed to the hypothesized outcomes. In the group treatment subjects became significantly more individualistic after the assignment. In the individual treatment the subjects became less individualistic on average, but the differences between the two visits are in this case not significant. The overall effect was tested in a set of difference-in-difference estimations and those results were also significantly opposed to the hypothesis. The results from the ‘reduced survey’ tests show different results. The averages of INDCOL of course differ somewhat after both the individual and group treatments, but no significant difference between the two visits or relationship in general is found – in any test.

In the following section related literature will be described. In Section 3 the hypotheses will be derived that will be tested in this study. Section 4 then describes the methodology which is used to test those hypotheses and in Section 5 the results will be presented. Section 6 discusses and Section 7 summarizes the findings of this study.

(7)

7

2 Related literature

Triandis, Bontempo and Villareal (1988) describe in a clear way what distinguishes collectivistic and individualistic persons. According to them, collectivistic people have goals that are subordinate to the goals of their group, which is a stable ingroup (e.g., family, band, tribe). The behaviour of these persons is likely to be consistent with the goals of this ingroup. Individualistic people are often part of more ingroups, and also have goals that are consistent with various ingroups. However, when ingroups become too demanding individualistic people tend to drop out of the specific group, whereas collectivistic people stay in it. This means that individualistic people are not looking at themselves alone (they for example do care a lot about their close family), but they take less people into account when taking decisions and are willing to sacrifice less for others than collectivistic people do.

I did not find a paper directly testing my research question, or being very close, so I will describe the literature I think is relevant in the following section. These articles intersect parts of my study or give more background information about separate aspects.

2.1 Studies on individualism-collectivism in general

Individualism and collectivism has been measured widely throughout the years. Most often this measures have been used at the national or cultural level. This is a useful tool when for example using the obtained information to implement the right marketing strategy in a specific country (de Mooij, 2000). However, there are some critical notes regarding the generalizing within a country or culture (Brewer & Venaik, 2012, 2014). Schaffer and Riordan (2003) found that in the studies they examined 79% of the studies used the country as a proxy for culture, which can be considered a questionable one. Oyserman et al. (2002) namely did a meta-analysis to study within-U.S. differences among cultures (which they measured by taking ethnicities) and they found some significant differences on both the individualism and collectivism outcomes (they are measured separately, contrary to this study in which the measures are combined to one final value). According to Gerhart and Fang (2005) country is accounting for 2 to 4 percent in the variance in respondents’ values. Vanello and Cohan (1999) describe a list of factors related to social-, political-, economic- and religious practices and family-structure and living arrangements which are related to the degree individualism and collectivism. These factors narrow more down to the individual level than the studies that solely focus on nationality or ethnicity.

In the world we live in pluralism is the norm (Tung, 2008). She furthermore argues that it is imperative to explore the differences that exist within countries. Besides this, culture is not

(8)

8

static either: it evolves over time. Brannen and Salk (2000) argue the changes over time occur through an increase in interactions among people that originate from different cultures and nationalities.

This makes the measure of individualism and collectivism an interesting topic to study, which is never finished in my opinion. Studying this topic after special moments, disasters or just after several years it might be interesting to make a comparison with other moments in time or in trying to discover a certain pattern.

2.2 Studies on payment structures

Like there is an extensive body of literature focusing on measuring individualism and collectivism, there is a large amount of studies considering payment structures. These studies focus on all sorts of aspects. Lazear (2000) and Lavy (2009) focus on the changes in productivity when going from a fixed to a variable payment, where Kennedy (1995) determined the effect on worker morale after the same payment shift. Pouliakas and Theodosiou (2007) found that the increase in the use of performance payments did not rule out intrinsic work motivation of the British population. Besides this, they also do not discover any differences in the overall job utility between performance paid jobs or jobs that are paid according to other compensation schemes.

Dohmen and Falk (2001) studied which workers prefer which payment structure. Their basic results are that more productive workers would rather self-select in a performance payment than in a fixed wage contract. This is consistent with Weiss (1987) who claims that group payments decrease productivity, since in both the high and low productivity groups the quit rates are high which eventually lowers overall productivity. Torsvik (2011) conducted an experiment where he let subjects decide whether they preferred an individual or a team bonus scheme. The above average productivity workers most often chose the individual scheme, but also a fraction of the low productivity workers did so. This was attributed to the other regarding preferences ‘distributional fairness’ and ‘social emotions’. Besides the outcomes of their study presented in the previous paragraph, Pouliakas and Theodossiou (2007) did find that gainsharing incentives were enhancing employee well-being more than individual performance payment.

Weiss (1987) claims that a group payment instead of an individual payment decreases inequity. This lowers the ‘perceived unfairness’ and this might increase the cooperation. Gerhart and Fang (2014), however, claim that even when a payment structure with low pay dispersion (group payment) will lead to more cooperation or collaboration, this will not

(9)

9

necessarily lead to an increase in productivity. According to previous research, the authors claim that it is known that high productivity workers are more likely to join and keep working for organizations that have strong performance incentives. And these incentives happen to be larger for individuals than for groups.

This wage dispersion is more thoroughly investigated by Lemieux, MacLeod and Parent (2009). They attribute the increase in wage dispersion to the increase in the use of performance pay. Like Gerhart and Fang (2014) above claim, Trevor, Reilly and Gerhart (2012) found that in their sample of NHL teams the teams with the highest performance pay were best able to attract top players. This is supported by Weiss (1987) who claims that individual performance pay is a useful mechanism for a firm to incentivize individuals to work hard and at the same time keep its best employees. According to Gerhart and Fang (2014), high performance pay is simply not possible without pay dispersion. Frank and Nuësch (2011) studied the relation of soccer teams’ performance and wage dispersion. They concluded that the teams with either a high or a low pay dispersion appeared to perform the best, while the ‘medium pay dispersion’ teams performed worse. This contradicts the argument of Gerhart and Fang (2014) who claim that even when cooperation increases, performance will not due to loss of the better workers (players) when they think they do not earn enough. How the reward system has been designed also matters. This is what Beersma et al. (2003) tested in their study. They tested a cooperative and a competitive reward system, both in a team assignment. The cooperative reward system enhanced accuracy, whereas the competitive system enhanced speed.

2.3 Studies on teams

As performance pay, teams are an often studied topic as well. The topics investigated in relationship with teams vary widely. In this section I would like to address some studies focusing on teams, since the difference between individuals and teams is an important aspect of my study. I will list studies which I consider relevant in giving a general understanding.

McClurg (2001) listed three determinants that would make a team successful. She claims that the best programs are the ones that contain a lot of communication to employees about the details of the plan; let the workers be strongly involved in the design and implementation of the plan; and let the employees believe the payments to various employees are fair. Comparable with the study of McClurg (2001) is the study of Shea and Guzzo (1987). This study lists the following three factors as having a major role in the ‘effectiveness’ of teams: task and outcome interdependence and potency. The first and second apply to how closely individuals have to work together and how the rewards are based on group performance.

(10)

10

Potency means that members need to have a belief in the capability of the team. Hamilton, Nickerson and Owan (2003) concluded from their study that teams perform best when their members are heterogeneous. Besides this, they also found that high-productivity workers tended to join teams if they could, despite some loss in earnings. This contradicts the findings of for example Torskvik (2011) earlier this section, but it does indicate that other incentives than just pecuniary benefits play a role in the decision what to prefer. Eckel and Grossman (2005) found that when team members worked together on (unrelated) tasks before the actual experiment, the cooperation was more enhanced than when the subjects were assigned to a team and then started the assignment. This pre-assignment was designed to enhance the ‘team identity’.

As can be seen above there is a lot written about teams, from whatever perspective. Some complement each other and some contradict. What I did find is that free-riding behaviour might exist in team assignments, since studies agree unanimously on this. Hamilton et al. (2003) describe a large body of literature that focuses on this specific topic. Weiss (1987) argues that the larger the group, the less willing an individual will be to exert effort. This logically makes sense, since you have to share your efforts with more people. Wagner (1995) found the link between individualism and the free-riding phenomenon. He argues that free-riding provides insights in analyzing the cooperation in groups of individualistic persons, but not in the groups of collectivistic persons. This is an indication that collectivistic people are not willing to free-ride, since they care more about their group than individualistic people do. Moorman and Blakely (1995) studied organizational citizenship behaviour. Although it was a study that contained self-reporting, the results showed that collectivistic people more often than individualistic people reported that they performed organizational citizenship behaviour. The results may even have been more significant, since 12% of the employees of the surveyed firm did not fill in the survey. Based on the found correlation, it would not be unlikely that this fraction of employees is part of the more individualistic half and does barely perform organizational citizenship behaviour.

2.4 Studies on payment structures and individualism-collectivism

The last two studies mentioned took individualism and collectivism into account. In the following section I will describe studies that write about payment structures and individualism and collectivism. To start with the man who introduced the individualism-collectivism measure, Hofstede (1994) argues that in countries that tend to be more individualistic individual performance payments are favoured. Opposed to this, in more collectivistic countries people

(11)

11

favour payments based on group performance. Besides this, Hofstede (1994) argues that in individualistic countries easy job-hopping is a well-liked phenomenon, whereas in collectivistic countries more long-term initiatives like family-enterprises are liked. Bloom, Milkovich and Mitra (2003) write about the international compensation practices. They describe two perspectives from which the reward systems can be designed. One of these is the ‘national culture’ perspective, emphasizing that a multinational has to set its reward system such that it adapts to the local culture. The other one is ‘strategic alignment’, which emphasizes matching the organizational context (for example organizational structure, strategic orientation) and reward system. The former one is would then be in line with the argument of Hofstede (1994). Schuler and Rogovsky (1998) support the argument of Hofstede (1994) as well, by stating that their results indicate that “individual incentive compensation practices have a better fit in countries with higher levels of individualism” (p. 172). Schuler and Rogovsky (1998) managed to create a dataset of appropriate size when combining data from three datasets. However, in all datasets questionnaires were mailed to subjects. The data obtained was all clustered at the national level. These observations were linked to degrees of (among others) INDCOL which they retrieved from a different study. I think this is a shortcoming, since the data cannot be linked properly. As I concluded earlier this chapter, it is not sound to generalize to the national level. Chen, Chen and Meindl (1998) derived propositions on how cooperation can be fostered best in either a collectivistic or in an individualistic culture. These propositions were not tested empirically in their paper. Not all propositions are relevant in this section, but the ones about reward distribution are. The ‘equity principle’ plays a role in formulating their propositions. In this principle people are paid according to their individual contributions to the total value. According to the authors, individualistic people perceive this as ‘fair’, whereas collectivistic people state the equity principle enhances self-interest at the cost of collective interest. Chen et

al. (1998) in their propositions also state that in individualistic cultures “equity-based reward

allocation systems will be positively related to cooperation in both the short- and long-term work relations” (p. 298). Cooperation of collectivistic people is, according to the authors, positively related when paid on an equity-base in short-term relations, but are in long-term relationships willing to give up some equity to establish a good relationship. Therefore, in long-term relationships cooperation of collectivists would be positively related by paying on base of equality. Ramamoorthy and Carroll (1998) also state that individualism emphasizes performance payment on the notion of equity more than collectivism. Collectivism is more associated with group rewards and they also name job security as another associated phenomenon. Besides this, in ‘individualistic entities’, employees are not expected to commit

(12)

12

more to the organization than their contract implies and both parties are not connected on any long-term basis. The contracts ‘collectivistic entities’ are based on moral commitment, focus on the long-term and encourage group performance. This is in line with the literature regarding organizational citizenship behaviour, like the paper of Moorman and Blakely (1995) at the end of the previous section.

The studies above agree quite well with each other in separating individualistic and collectivistic cultures and what comes with it. Like mentioned in section 2.1, one cannot generalize too easily within countries or cultures. Bloom and Milkovich (1998) for example did a study considering reward systems in the individualistic United States and collectivistic Slovenia. Results appeared to overlap, which confirms the within-country variation. The authors also list studies that observed firms from similar competitive environments that used different reward systems, but that not necessarily performed differently. A comparable result comes from Fang and Gerhart (2012). They studied workers in the collectivistic Taiwan that were paid according to individual performance. Contradicting arguments of some other studies, they found that the firms that did use individual performance payments were not causing a diminished intrinsic motivation, neither did the firms perform worse. Moreover, both aspects were enhanced. However, intrinsic motivation in this study was measured by a survey which measured intrinsic job satisfaction. I do not know whether this is the best way to test this. First, this survey asked for answers on a scale from one to ‘something’. A cautious person might initially more often choose the middle answers than a more impulsive person. In my design I am working with surveys as well. However, I am measuring within-subject differences. This solves the ‘survey problem’, since it is not about absolute values anymore. Second, intrinsic job satisfaction was used as a proxy for intrinsic motivation. Although claimed as being a good proxy, I think a study like Deci and Ryan’s (1985) – who showed in an experiment that intrinsic motivation shrinks after dropping payments for a task – is more reliable in this sense, but which is contradicting Fang and Gerhart (2012). Gerhart and Fang (2014) argue that there is no good direct evidence whether the effectiveness of individual performance payments varies per country or culture, but that there is evidence on the (relative) usage of reward systems per country. ‘Conventional wisdom’ argues that individualistic countries rely more on individual performance pay (equity principle), whereas collectivistic countries rely more on group performance pay (equality principle). A meta-analysis by Fischer and Schmidt (2003) shows that the differences that exist are quite small. This meta-analysis uses 25 studies which had to fulfil some well described requirements before being included in the dataset. These studies compared two or more cultures. 21 of these have the United States as reference country and 22

(13)

13

based their sample on students. I think both the reference country and sample are showing a valid consistency, making this a reliable meta-analysis.

According to Bloom and Milkovich (1998) the literature on national culture is ‘implicitly prescriptive’: it more or less dictates that reward systems must be aligned with national cultures. This causes stereotyping. Bloom and Milkovich (1998) quote Schloss who wrote in 1898 that firms started to develop their own norms and that reward systems were related to these norms. Schneider (1987) came up with the attraction-selection-attrition (ASA) model that builds on the argument of Schloss. In this model firms create a unique environment, which is possible because ‘people make the place’. The firm specific culture is transferred to new employees. If the employees do not fit into the culture, they will leave if they’ve joined in the first place. This way, the firm only contracts employees with a homogeneous attitude towards the prevailing culture through which this culture can be maintained. Schneider (1987) did not build this model solely on his opinion. To make his model convincible, the author draws on findings and theories from authors in different fields of psychology.

Besides the culture an organization wants to create, it of course also has to deal with the regulation and legislation of a country. This has its effect on all sorts of aspects of business and therefore also on reward systems (Bloom and Milkovich, 1998).

As we saw in this chapter, a lot has been written about individualism and collectivism, effectiveness and preferences of reward systems, but to my knowledge no studies are giving answers to my research question. A lot of studies describe correlations between payment systems and INDCOL. I am going to investigate whether there also is a causal relationship between the two.

3 Hypotheses

In this study I test the degree of the subjects’ INDCOL at two moments in time. The difference between the two moments is that the latter one is preceded by an assignment to investigate whether possible INDCOL differences which then can be attributed to the assignment. For half of the subjects this assignment was an individual task, for the other half it was a group task (see the next section for a more detailed description). Given the factors that make a team successful (Shea and Guzzo, 1987)1, I think the ‘teamspirit’ will be such that a real difference between the

1_{My design does not really support the factors mentioned by McClurg (2001), but it does match the one}

(14)

14

individual- and the team treatment has been created in my design. Due to time constraints, I could not implement a pre-assignment as sketched by Eckel and Grossman (2005) to enhance the ‘team identity’. The individual treatment works the same for all subjects that fall under this treatment. A problem with the team treatment could be that not all teams function in the same way. The extensive body of literature on freeriding behaviour (Hamilton et al., 2003 for a review) gives some insights. Weiss (1987) stated that the more people are part of a group the less willing an individual is to exert effort. In the design I tried to keep groups as small as possible (three or four), but larger than two people.

However, there still might be different behaviour among groups. Wagner (1995) found that individualistic people tend to freeride more than collectivists. This might influence results of the INDCOL values: after a good cooperation one might be more collectivistic, but after a disappointing cooperation one might become more individualistic. I expect this result to be balanced so I therefore not base my hypothesis on this information.

Like mentioned in the previous chapter, the studies that would come closest – section 2.4 – to my study are not close enough to formulate a clear hypothesis from. Therefore, these studies can be considered as background material as well in my opinion. Based on the preferences for (Bloom et al., 2003; Hofstede, 1994; Ramamoorthy & Caroll, 1998) or effectiveness of (Chen et al, 2003; Schuler & Rogovsky, 1998) certain payment systems in relationship with the value of INDCOL we could reason in two ways: did the man change the reward system or did the reward system change the man? This is a dilemma like the world famous ‘chicken or the egg’ causality dilemma.

Based on the study of Bloom and Milkovich (1998), who quotes Schloss (1898), and the model of Schneider (1987) I think people are reshaping the place where they work and the country they live in. Firms started with implementing certain reward systems (based on the ideology of the founder) so these people started shaping places. In the beginning there were not as many firms as now of course, so people did not have a lot of different opportunities to switch to if they initially did not agree with the culture of the firm. Therefore, I think it makes sense that in the beginning firms were able to shape people over time and transfer their culture to its new employees. When working in and getting paid as a team I think one gets a more collectivistic attitude and vice versa. Besides that the models mentioned above make sense to me, they are also taking change of peoples’ attitude over time into account. This is not the case for the other articles described, they are more showing correlations observed at that moment. This leads to my first hypothesis:

(15)

15

H1. People that are paid according to their own performance become more individualistic than people that are paid according to their team’s performance.

Above I mentioned the preferences for certain reward systems (Hofstede, 1994; Bloom et al., 2003; Ramamoorthy & Caroll, 1998), which are quite unanimously. They all state that individualistic people prefer an individual payment and collectivistic people prefer a group payment. Of course individualistic people not only care about themselves, so when they sit next to some close friends they might prefer a group payment anticipating they will form a group with their close friends. However, I think the first effect is larger and it is not solely based on my intuition. Therefore, my second hypothesis is:

H2. People that are more individualistic prefer individual performance pay, where more collectivistic (less individualistic) people prefer team performance pay.

4 Methodology

In order to test the two hypotheses, I will use an empirical method to study within-subject differences. Two times I will visit the subjects to conduct several parts of the experiment. Two elements fulfill a central role in the experiment. Below I will per visit describe what I will do, how I will process the results and why I use this particular way.

4.1 Visit 1

In the first visit I will only determine the degree of INDCOL of the subjects by using a survey. The survey I use can be observed in Appendix 1a. This survey consists of statements that all measure either individualism or collectivism. Subjects indicate how much each statement applies to them, on a scale from 1 (not applicable at all) to 7 (very applicable). Eight statements measure collectivism and the remaining six measure individualism.

4.1.1 Processing the results of visit 1

To determine the final INDCOL score of a subject I will use the following formula: 𝐼𝑁𝐷𝐶𝑂𝐿 = 𝑇𝑜𝑡𝑎𝑙 𝑠𝑐𝑜𝑟𝑒 𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑣𝑖𝑠𝑚

8 −

𝑇𝑜𝑡𝑎𝑙 𝑠𝑐𝑜𝑟𝑒 𝑖𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙𝑖𝑠𝑚 6

All given answers to the statements are added for both the statements applying to collectivism and individualism. Those values are divided by the amount of statements considering both topics. The remaining values are the average scores given for both the collectivism and individualism statements. The average individualism score is subtracted from the average

(16)

16

collectivism score. The remaining INDCOL value can be interpreted the following: when negative, a person rates individualism statements higher than collectivism statements and can therefore be considered as more individualistic than collectivistic; when positive, a person rates collectivism statements higher than individualism statements and can therefore be considered as more collectivistic than individualistic; when 0, a person rates collectivism and individualism statements the same on average and cannot be considered as more individualistic or collectivistic.

4.2 Visit 2

In this visit again the list of statements to determine the INDCOL is a key element, preceded by an assignment. This assignment will either be an individual or a group assignment. The visit starts with explaining both variants and after the explanation the subject is asked: do you prefer the individual or the group assignment? This has been put in Appendix 2a. After the question is answered by all subjects I communicate to the subjects which assignment they actually will do. Upfront the visit I divided the assignments such that the observations were divided as equally as possible. The assignment lasts for 5 minutes. The individual assignment can be observed in Appendix 2b. The groups in the group assignment consist of three or four people and the particular forms can be looked at in Appendices 2c/2i. Immediately after the assignment the subjects will again fill in the survey to determine their second INDCOL value, following the same statements as in Appendix 1a, but the layout was modified to Appendix 1b.

4.2.1 Processing the results of visit 2

After this visit I have all the data I need and I can start testing my two hypotheses. The first hypothesis (after an individual assignment subjects become more individualistic and after a group assignment subjects become more collectivistic) can be tested in two broad ways. First, I will use non-parametric tests, the t-test, to compare the different INDCOL values. This t-test performs a test to determine whether two means are or are not equal. This is relevant, since I want to compare two INDCOL values and determine whether they differ and, if they do, whether these differences are significant. The t-test will be used three times: i) comparing the first and second INDCOL values of the subjects that did an individual task; ii) comparing the first and second INDCOL values of the subjects that did a group task; and iii) comparing the second INDCOL values of subjects that did an individual assignment with the second INDCOL values of the subjects that did a group assignment. I now can test whether differences between these groups of values are significant and then form a (preliminary) conclusion about the possible effect of the treatment. Second, I will use a difference-in-differences regression for

(17)

17

further testing. I use difference-in-difference estimations, since I want to determine the effect of the treatment on the outcome. However, in this study ‘the outcome’ is a change over time. It is the value with which the INDCOL values changes from the first to the second visit. This makes the main independent variable the treatment (either the individual or group assignment) and the dependent variable the difference between the two INDCOL values (INDCOLd). Control variables here would be gender, skill level and the class of which a subject is part. The dependent variable of this estimation will also be tested under iii) as mentioned above using a t-test and a Mann-Whitney test. I will use the Mann-Whitney test to determine whether two samples are from populations with the same distribution. The two samples are all observations, separated by the followed treatment (individual/group). This Mann-Whitney test can be considered as an extra test on top of the t-test to determine whether the changes in INDCOL differ among the treatments.

The second hypothesis (who prefers what treatment) will be tested by, again, using a non-parametric test. I now will use the INDCOL values of the first visit in combination with the given preferences of the second visit, since the INDCOL values of the second visit might be affected by the treatment. So I will compare the first INDCOL values of the subjects that prefer the individual assignment with the first INDCOL values of the subjects that prefer the group assignment.

4.3 Justification of the used survey and assignment

In this section I will justify the survey and assignment I used, starting with the survey.

4.3.1 Justification of the survey used

The statements in Appendix 1a come from the paper of Sivadas, Bruvold & Nelson (2008). I translated the statements to Dutch, where I obviously tried to stay as close to the original English translation as possible, with one exception. I changed the word co-worker for

classmate, here I expect the subjects to have more ‘feeling’ with since they do not have

(fulltime) jobs but are related to each other as classmates. The original English translation of the list used by Sivadas et al. (2008) can be observed in Appendix 1c.

The list of Sivadas et al. (2008) is one of many lists that has been used throughout the years measuring INDCOL. In their article, Sivadas et al. compare their list with lists from other authors. They test these for example by using Cronbach alphas, so they can determine the consistency in given answers per subcategory. I have to make clear that my experiment is not about measuring the degree of INDCOL. Therefore, I argue that it is less of less importance what exact list I use compared to studies that do study this degree. However, the list with

(18)

18

statements of course has to be internally valid when being used in a study. Sivadas et al. (2008) find their 14-point list most internally valid when comparing it to other lists and therefore this is the list I will use.

The list I use distinguishes between horizontal and vertical individualism and collectivism. However, Sivadas et al. (2008) state that the horizontal and vertical measures can be summated when one is interested in just the individualism and collectivism measures (p. 209). This is the case in my experiment, that is why my formula above is formulated in this way.

4.3.2 Justification of the assignment

The duration of the assignment might sound as too short and not long enough to make the comparison between the subjects of this study and employees with (fulltime) jobs. However, I do think that this will do since the treatment is quite intensive. During the five minutes one is either working purely on an individual basis or communicating and cooperating a lot with his group. Moreover, the assignment is not that intellectually challenging and therefore I do not think any extra minutes will make the outcomes more externally valid. My last reason is practical one, I namely only had 25 minutes for my second session. In this time span the assignment had to be explained, executed and afterwards the INDCOL value has to be determined again.

The assignment itself contains of counting letters, numbers and figures. This is something everyone should be able to do. This is the main reason for using this kind of assignment, it could have been any other easy assignment. I chose for an easy assignment, since I want the subjects to indicate their preference for either the individual or the group assignment solely based on whether they prefer to work alone or in a group. I do not want subjects to take skills into account and for example think they are (not) good at the assignment, and therefore prefer to work alone (group), when making the decision. This way, the subjects can indicate whether they really prefer working alone or in a group – irrelevant of the difficulty or nature of the assignment – and the second hypothesis can be tested.

I tried to create as large differences as possible between the individual and group assignment. In the individual assignment your performance (the amounts of points you can gather) and the possible payment solely depend on the subject itself. In the group treatment this is completely the opposite. Here the amount of points you can gather – and add to your group total – depends on both the subject and its group members, since they have to share information and decide what tactic to use. To clarify, I do not mind which information is shared or what

(19)

19

tactic is used, since I do not want to determine what is an efficient approach of the assignment or something in that direction. What I do want to examine, is whether this teamwork (the cooperating and communicating) leads to different INDCOL values after the assignment than before. And whether these possible differences are also different from the effect the individual assignment might have on the INDCOL values. Also the payment depends on the contribution of all group members. If my results show significant differences between the treatments, further research could for example only test the possible effect of one of the two dependencies (either on performance or payment). As Shea and Guzzo (1987) described three factors of team effectiveness: task interdependence, outcome interdependence and ‘potency’. The first two are met in my assignment and the third is irrelevant, since teams do not compete against each other, but only against themselves by achieving the highest possible score. I do not care about the effectiveness of the team regarding performance, but when the effectiveness factors are met I think the impact of working in a team might be larger. Besides this, Weiss (1987) argues that in group payments there is less ‘perceived unfairness’, which may increase the cooperation. The payments for group members are equal, so this may enhance the difference between the two treatments.

The group size is something I determined partly for practical reasons, but also based on literature. Since classes differ in size I chose for groups of three or four since with these two numbers you can easily form groups of all class sizes. Besides, I think that two people is not a real group, which is the reason I did not use this size. Three or four people is the maximum in my opinion, since Weiss (1987) states that the larger the group is, the less an individual is motivated to exert effort. Because of the task- and outcome interdependence of the assignment, I think free-riding behaviour is minimized and the effect of working in a team is achieved. From each of the eight participating classes, one student was drawn that was paid according to his (and his group’s) performance.

4.4 Sample details

My experiment takes place at a high school called the Christelijk Lyceum Veenendaal. Subjects’ age is between 15 and 17 years old on average and the share of males is 60%. Subjects come from three different groups: 63 follow classes at the level havo 4, 40 at vwo 4 and 34 at vwo 5, making 137 subjects in total. They all follow the course economics.

(20)

20

5 Results

In this section the results will be presented. After the first visit the sample contained 154 observations, from which 137 were left after the second visit. This loss in observations was due to students that were not attending class during my second visit, whereas they were during the first visit. Also some students did not completely fill in the questionnaire, therefore I excluded them from the entire dataset.2

5.1 Summary statistics

In Table 1 the summary statistics for both INDCOL variables (visit 1 and 2) are displayed. As one can see, the mean (2) and median (6) values are both negative. This indicates that the majority of the subjects perceived the individualistic statements as more applicable to themselves than the collectivistic statements.

Table 1: summary statistics of INDCOL values of both visit 1 and 2.

Men and women also score significantly different. The average scores for men in the first and second visit are -1.18 and -1.21, whereas women score -0.72 and -0.84. Differences in both sessions are significant at the 5% level according to the Mann-Whitney test. When comparing years (year 4 and year 5) and skill (havo and vwo) I do not find any significant differences for both INDCOL values. When testing differences within class categories – havo 4, vwo 4 and vwo 5 – I only take the first measure of INDCOL into account. This since there for example are only three havo 4 classes, which not all followed the same treatment in visit 2. The difference in the first INDCOL value is significant for havo 4 (-1.13) and vwo 4 (-0.75). The differences between vwo 4 and vwo 5 as well as the differences between havo 4 and vwo 5 are not significant.

2 I did test whether the excluded group is significantly different from the remaining subjects in the sample. I tested this with respect to both INDCOL values from the first visit (see section 5.4 for an explanation of this ‘‘second’ first visit INDCOL value’), gender, havo or vwo and year 4 or year 5. All differences were not significant, therefore the excluded and included subjects are not significantly different. This means the exclusion of the specific group does not harm the study.

(21)

21

The scores between men and women are significantly different, as well as there is one significant difference among the three class categories. However, any class can have a different atmosphere, for example created by the teacher or a student or group of students. In Figure 1 the results per class can be observed for both visits. As one can see, the results are varying among classes. As can be seen, the classes labelled 1-4 followed the group treatment, whereas the classes labelled 5-8 followed the individual treatment. Remarkable is the fact that these averages are not in line with my hypothesis, since the values from the second visit are more individualistic after the group assignment and vice versa. The only exception is the class labelled as 8, which does follow the expected change – albeit minimally.

Figure 1: averages of INDCOL values per class per visit.

5.2 Non-parametric statistical tests

Statistical tests were ran to determine whether the differences mentioned in Figure 1 are also significant. As mentioned in the methodology section, three times a non-parametric test will be used to test the first hypothesis. The first tests whether the two INDCOL values are significantly different for the subjects who followed the individual treatment. This has been tested by using a t-test. As we see in Table 2, the means of the two variables differ in the opposite way as hypothesized. Where I expected subjects that followed the individual treatment to become more

-1,6 4 , -1 2 -1, -1 8 , -0 6 -0, -0,4 -0,2 0 1 2 3 4 5 6 7 8 Group Individual

INDCOL averages per class per visit

INDCOL visit 1 INDCOL visit 2

(22)

22

individualistic, their average answers tend to be more collectivistic (less individualistic). This supports the values in Figure 1. However, according to the p-value of 0.256, the differences between the INDCOL values of the first and second visit are not significant.

The second test used examines the same as the first, but now for the subjects that followed the group treatment. Again a t-test is used and the values are summarized in Table 2. As seen in Figure 1, averages for all classes went in the opposite of the hypothesized direction. Subjects tended to be more individualistic after the group assignment. The values in Table 2 support the values in Figure 1 and are of statistical significance according at the 1% level (p-value of 0.009). From this I can conclude that the subjects that followed the group treatment filled in their questionnaire significantly more individualistic after the treatment than before.

Table 2: summary of values of non-parametric tests

The third test will test the differences between INDCOL values of the second visits of both treatment groups. The mean second visit INDCOL values for both groups do differ, but the difference is not significant at a relevant level (p value of 0.260 under 3.1 in Table 2). However, this does not mean the significant difference from the second test (group assignment) is meaningless. Namely, this study is about the within-subject differences. Therefore, I created a variable called INDCOLd, which is the INDCOL2 value minus the INDCOL1 value. When INDCOLd has a positive value, it indicates that your INDCOL1 value is more negative than

(23)

23

the INDCOL2 value and vice versa. So a positive value can be associated with getting less individualistic, whereas a negative INDCOLd value can be associated with getting more individualistic. In 3.2 in Table 2 I listed the INDCOLd values for both treatments. As we see there are clear differences. After the individual assignment the subjects become less individualistic (positive INDCOLd value), whereas the subjects become more individualistic (negative INDCOLd value) after the group assignment. The differences in shifts of the INDCOL values between the two visits are significant at the 1% level (p-value of 0.005. The Mann-Whitney statistic also indicates that the differences between the two groups are significant at the 1% level.

Taking above three tests into account, I can quite easily reject the first hypothesis. Clearly, one becomes neither more individualistic after an individual assignment, nor more collectivistic after a group assignment. At this point it looks more like the opposite. Since the second test did find a significantly more individualistic value after the group assignment than before. The first test did not find a significant more collectivistic value after the individual assignment, but the averages pointed in that particular direction. To determine whether the overall effect is significant – testing both treatments at the same – I will later perform a difference-in-difference estimation.

The second hypothesis considers the ‘self-selection’ of specific types into sorts of assignments. To clarify, both the individual and the group assignment were explained to the subjects during visit 2. After the explanation the subjects indicated which assignment they preferred. When all subjects indicated their preference, I decided which assignment the subjects would actually execute. With this preference for a specific treatment (which I called ‘self-selection’ at the top of this paragraph) and the INDCOL value from visit 1 I can test this hypothesis. In Table 3 the INDCOL values are presented, sorted by preference. As we see, the mean-, minimum- and maximum INDCOL values from the first visit indicate that the subjects with a preference for an individual assignment are more individualistic. However, the median is as good as equal (here it looks equal due to the fact it is rounded at three decimals). Besides that, the p-value arising from the t-test (7) and the p-value from the Mann-Whitney test (8) both show there is no statistical significant difference between the two groups. So even though the descriptive statistics might support the second hypothesis, it has to be rejected due to lack of significance. Therefore, I can conclude that subjects with a preference for the individual assignment are not significantly more individualistic than subjects with a preference for the group assignment. Using the terminology from the Introduction: the man does not determine the reward system.

(24)

24

Table 3: statistics in order to reject the second hypothesis.

5.3 Difference-in-difference estimations

The evidence in subsection 5.2 did reject the first hypothesis, but the overall effect is not made clear yet. I discovered that a group assignment made the subjects significantly more individualistic, but the individual assignment did not cause subjects to be significantly more collectivistic – although the means pointed in that direction. Besides that, by checking the INDCOLd values I found that both groups reacted significantly different at the two types of assignments. In this section I will perform difference-in-difference estimations to clarify the overall results, instead of separating per treatment like in the previous section.

In Table 4 I presented multiple forms of the regressions, taking INDCOLd as the dependent variable and the specific treatment as the main independent variable. The regressions differ in which control variables are used. In regression (1) only the treatment was used as an explanatory variable. As we see this coefficient is significant at the 1% level and supporting the evidence so far: subjects that followed the individual treatment have a higher INDCOLd than subjects that followed the group treatment, in other words: subjects become less individualistic after having executed an individual assignment and become more individualistic when having executed a group assignment.

In regression (2) I added two control variables that might affect the outcomes. As we have seen in paragraph 5.1, men and women have different starting values. This obviously does not say anything about the INDCOLd, but I think it is relevant to test whether this variable has influence. Therefore, this gender control variable was added. The variable concerning the preference for a specific treatment of a subject is added based on an experience while running the experiment. A few subjects thought their preference was the form they actually would get. This might affect their answers in the questionnaire after the actual assignment. However, as can be seen in (2) in Table 4 both the gender and preference variables are not having a significant influence on the dependent variable INDCOLd. The treatment variable is still significant at the 1% level with the sign in the same direction as under regression (1).

(25)

25

Table 4: difference-in-difference regressions with and without control variables.

In regressions (3) and (4) I still included the control variables as under (2), together with either the class categories (3) or the classes (4). This since in section 5.1 it also became clear that there was a significant difference between the categories havo 4 and vwo 4. Besides that, from Figure 1 it became clear that classes differ – although this has not been tested on significance. However, this all has to do absolute values of INDCOL, and not with the difference between the values of the two visits. I did still find it a relevant variable to control for in a regression. Because they partly overlap, I ran two regressions separately. As can be

(26)

26

observed in Table 4 both the class category variables and the class variables are not significant. In regression (3) the treatment variable is still significant, whereas this significance disappears in regression (4). Even though not all regressions support the earlier outcomes, I for sure can reject H1.

In section 5.2 it became clear that group assignments significantly made subjects more individualistic. The averages of subjects that did the individual assignments were pointing in the direction of getting less individualistic, although this difference was not significant. In this section I combined the two groups and checked the overall significance. In 3 out of 4 regressions the treatment variable is highly significant. Therefore, I can conclude that the treatment has the opposite effect as initially hypothesized. To again use the terminology from the Introduction: the reward system determines the man to a certain extent – albeit in the opposite direction as hypothesized.

5.4 Tests with a selection of the used questionnaire

The questionnaire of Sivadas et al. (2008) contains statements regarding the individual and people around him. However, one could argue that the statements considering co-workers for example will be more affected by the treatment than the statements considering family. Therefore, I again tested for differences between the treatments, now with the seven remaining questions. These are marked with an asterix (*) in Appendix 1c. The remaining selection of statements should all be questions that are completely relevant in this setting.

All tests which are shown in Table 2 were ran again with the selection of statements. The results become a lot less remarkable then. The averages for both the individual and the group treatment become a bit more individualistic after both the assignments. However, these differences are not significant at any relevant level. The averages of both treatments – and the value of the INDCOLd variable – both show that the individual treatment subjects change ‘more individualistic’ than the group treatment subjects. But as mentioned, not significantly more. The difference-in-difference regressions shown in Table 4 are also never showing a significant treatment coefficient, in neither of the four ran regressions. Therefore, the conclusion towards H1 remains unaffected – it still is being rejected. However, in the complete survey the results were opposite to the hypothesis, this is not the case when using the selection. H2 remains rejected as well. Opposed to the complete survey, the averages do not point in the hypothesized direction. Similar, however, is that this again is not a significant difference.

As mentioned, the differences between the two INDCOL values are not created in the seven questions I would expect the most (marked with an asterix in Appendix 1c). Apparently

(27)

27

the other seven questions are ‘causing the difference’ between the two visits. At the individual subject level, I could argue that a person at the morning of the first or second visit for example had a really nice conversation with his family members. Or the complete opposite. This could all affect differences between the two INDCOL values. Of course, this effect should be balanced out due to my sufficiently large amount of observations. Therefore, I do not think this ‘mood of the day’ is a really good explanation. What could be an explanation, is the following – although it may be a bit far-fetched. Since the subjects’ answers are all given on a scale from 1 to 7, they might take their first answer as a starting point and base their next answer on the answer(s) they already gave, by thinking: ‘does this statement apply more or less to me than the previous’ instead of ‘to what extent does this statement apply to me’. The first questions may be more answered in the sense of to what extent the statement applies to the subject, whereas the later questions may be more answered by comparing the applicability with earlier given answers. Obviously, this could have happened at both visits. However, the majority of the questions marked with an asterix are quite at the beginning of the survey. If the subjects were affected by the treatment, this effect then could be larger in the later stage of the survey than at the first few statements. Due to role of reference point the first few statements may get. This might cause a difference in outcomes between using the complete 14 statements survey and the reduced 7 statements version.

6 Discussion

The results of this experiment were quite remarkable. Both hypotheses were rejected and the main hypothesis even showed opposite results in the ‘complete survey condition’. Although averages pointed in the opposite direction of what was hypothesized, the individual treatment did not have a significant influence on the change in the value of INDCOL. In the group assignment, however, this did have a significant effect. The second hypothesis, considering self-selection into certain payment structures, did show averages pointing in the hypothesized direction. However, the averages from subjects preferring either the individual or the group assignment did not differ significantly. In the ‘selection condition’ both hypotheses were rejected as well, but no result was significant.

6.1 Explanation and discussion of results

A possible explanation of this opposite effect might be attributed to the fact that group members could quite good determine how much each group member contributed. If one group member was much faster than the others, he might consider himself as better than his group members.

(28)

28

He then might think he would have been better off alone. This at its turn might lead to a less collectivistic and more individualistic attitude at the moment of filling in the second survey. The same for group members who consider themselves as slower than the rest of their group members, they might feel a lack of sympathy or feel hurried up by their group. This may also lower their esteem of the group and make them less collectivistic and more individualistic when filling in the second survey. Leaving us with the average skill group members. I think one can assume that most of the subjects are of average speed. I think these group members might also get less collectivistic and more individualistic when they observe skill differences within their group. Since this could lower the feeling of unity for example.

The observation that subjects following the individual treatment got more collectivistic (less individualistic) – even though this difference was not significant – is somewhat harder to explain. A possible explanation for this might be the following. The individual assignment was designed such that it was ought to last longer than the time given to the subjects. This because I wanted subjects to be fully focused on themselves during the assignment and immediately afterwards fill in the survey. However, subjects might have gotten a (temporarily) lowered self-esteem due to the fact that they were not able to finish the assignment in time. They then could have reasoned that if they were in a group, they might have been able to finish the assignment – making them more collectivistic and less individualistic at the second survey.

The hypothesis regarding the preference for a certain treatment was supported by the average INDCOL values from the first visit, but these two preference groups did not differ significantly. This might also be explained by the construction of the assignment. In section 6.2 I will discuss the limitations of my experiment. The one I start with will be the one that might influence this particular result.

6.2 Limitations and future research

Before the subjects indicated which treatment they would prefer, both variants were explained. Although being described as easy as possible, more questions were asked about the explanation regarding the group assignment than for the individual assignment. This obviously could affect the results: if a subject has a question about the group assignment, but is not willing to ask this for whatever reason, it is simple to ‘prefer’ the individual assignment. A better alternative would for example have been to make the decision not about which treatment a subject prefers, but at what amount of money he wants to switch from the one to the other. If each point in the individual treatment stays worth 20 euro cents, subjects could indicate how much they want

(29)

29

each point to be worth to prefer the group treatment. This way they might be more challenged to fully understand both treatments and ask their questions if necessary.

External validity is always a critical point in experiments. In my study this is the case as well. First, the fact that I use (teenage) students instead of adult employees is not a problem, since the students make real choices for a (possible) payment. This is the same as employees would do. Second, however, the duration of the actual treatment was only 5 minutes. Although the treatments were quite intensive (a lot cooperation/communication in the group treatment and no cooperation/communication in the individual treatment), it might not be completely comparable with the working situation in a company. I am not completely sure whether an increase in length will affect the results, but the longer the treatment lasts the easier it will be to generalize. Third, the payments now were at random. One person per class got paid, due to my personal financial situation. This might not incentivize the students fully, since they might weigh their effort costs and possible benefits. If all subjects were paid, they were incentivized to exert effort and all treatments would have had the effect which was desired upfront. This since the functioning of a group is more important, since all group members definitely get paid, as well as the level of exerted effort in the individual treatment. This not necessarily affects the final results, but it does make sure that all subjects exerted (the same) effort and started the survey with having done the ‘same assignment’.

In this experiment the individual and group assignment differed in both performance (amount of points you score) and payment (the amount of money you earn). To determine whether both effects count or only one effect causes the differences, in a future research this might be separated.

Following the possible explanations (6.1) of the remarkable results, in a future research I would recommend to ask for feedback of the individual or group process immediately after the assignment and survey. At the time I analyzed the results, my second visit of the high school was too long ago to ask for detailed feedback. This feedback might consist of questions for example regarding your position in the group – based on skills, or being a leader or not – or a similar question like the preference question before the assignment, but now afterwards. The former question measures more detailed, the latter one is more general.

7 Summary

The literature in this study was not that well represented. I did not find any direct literature on which I could base my hypotheses. The literature I found was more background material and

(30)

30

described studies that measure the degree of individualism and collectivism in general; studies that focus on payment structures; studies that elaborate on teams. The end of the literature section consists of studies that focus on both individualism and collectivism and payment structures. However, these studies focus on whether group or individual payment structures are more effective or more favoured in either individualistic or collectivistic countries. I also found a theory describing that each firm tries to create and maintain its own specific culture. People that fit this culture stay at this company – either by already having the same norms and values or by adapting them. Therefore, I think the reward system can change the man, which led to my hypotheses. These stated that I expected an individual assignment to make subjects more individualistic; and that a group assignment makes subjects more collectivistic. Besides that, I tested the preferences for either an individual or a group assignment. I expected more individualistic people to prefer the individual assignment.

In my study I had 137 high school students who participated. They all followed the

economics course. I used a survey to indicate the degree of individualism and collectivism. In

my first visit to the high school I took the survey. Four weeks later I returned for the second visit in which I explained both the individual and group assignment, asked for the subjects’ preference, took the actual assignment and – last but not least – again took the survey. Now I could run tests to discover the differences between the two times the students filled in the survey, between which the only difference was the individual or the group assignment.

The results showed that the effect was the opposite of what I hypothesized: after the group assignment the students were significantly more individualistic and after the individual assignment the students were more collectivistic (although not significant). The difference-in-difference estimations supported the findings from the non-parametric tests: the overall effect was significant in three out of four regressions and in the opposite direction as hypothesized. I also ran a second set of tests, since the survey also takes questions into account which for example are related to family matters. I left out all the questions that were not necessarily affected by the individual or group assignment. The results then showed no significance at all, and means were closer to each other as well.

Possible explanations for the remarkable outcomes (of the complete survey condition) were that in groups people might have felt hurried up by their better performing team mates or felt slowed down by their worse performing team mates. The ‘middle group’ might have observed some inequalities in the team. All this could explain the becoming more individualistic of subjects that followed the group treatment. The getting less individualistic of subjects that followed the individual treatment was a bit harder to explain. It could be attributed

The effect of different payment structures on the scale of individualism-collectivism : a study in which individual performance pay and team performance pay is analyzed in comparison with the scale of individualism-coll