2015-2016
How does demographic homogeneity influence organizational
citizenship? An analysis among the “Fortune, 100 best companies to
work for”
Author: Camille Mayoly Student ID: 11087153
MSc in Business Economics – Managerial Economics and Strategy Supervisor: Adam Booij
2 Statement of Originality
This document is written by Student Mayoly, Camile 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
Table of Content
Abstract ... 4 Introduction ... 5 Literature Review ... 7 2.1 Organizational Citizenship ... 7 2.1.1 Job Satisfaction ... 8 2.1.2 Organizational Commitment ... 82.2 Organizational Demography Theory and Social Identity Mechanism ... 9
2.2.1 Organizational Demography Theory ... 9
2.2.2 Social Identification Mechanism ... 10
2.3 Empirical findings and Hypotheses ... 11
2.4 Summary of the Hypothesis ... 13
3. Context, Data and Design ... 14
3.1 Sample Construction ... 14
3.2 Control Variables ... 16
3.3 Summary Statistics ... 17
3.4 Design of the Model ... 19
3.4.1 Fixed Effect Model ... 19
3.4.2. Random Effects Model ... 20
3.4.3 Type of homogeneity tests ... 21
3.4.4 Statistical Problems ... 22 4. Results ... 23 5. Mechanisms ... 29 6. Concluding Discussion ... 31 7. Appendix ... 33 8. Bibliography ... 36
4
Abstract
Gender and Ethnicity are the main types of homogeneity that are influencing team’s dynamics. To study the different impacts of homogeneity in terms of gender and ethnicity, we tested the relationship between the ranking of the companies from the “100 best companies to work for” and their level and type of homogeneity. Our study analyzed the relationship between the ranking of 234 companies and their organizational compositions over 10 years. We find moderate results concerning the relationship between gender homogeneity and organizational citizenship. We found that homogenous groups composed mainly by men are leading to higher organizational citizenship. The results between ethnic homogeneity and organizational citizenship showed a positive relationship between the two variables. The theory proposes that this relationship is due to higher level of social identification and organizational attachment within ethnic homogenous groups.
5
Introduction
Nowadays companies are vastly operating globally, labour is becoming more flexible, women are taking an increasing role within organizations and governmental policies are reinforcing gender and minority quotas. These developments are leading to an increased attention from the literature analysing the benefits and stakes of diversity at workplace. On the other hand, there is also a growing literature showing that job satisfaction at work is correlated with an increase of productivity level, a better customer orientation and creativity. Such novelties influence companies’s policies targeting high job satisfaction of their employees. However, it is commonly though that being in a group of similar people will increase the wellbeing of employees. In this context, we would like to empirically test whether these two variables have an impact on each other and could be simultaneously used for the application of a smooth diversity management.
Although a lot of research has already been conducted on this topic, most of the studies have focused on the organizational behaviour of individual companies, such as the impact of gender and minority ratio within one organization. Therefore, our study will be conducted on an aggregate level comparing a large set of company’s working climate. The precise goal of this study is to analyse how the organizational composition in terms of gender and ethnic homogeneity influences job satisfaction and organizational commitment, that we are going to sum up as organisational citizenship. More specifically, we would like to study the impact of the level and type of homogeneity from the “100 best companies to work for” published by Fortune on their organizational citizenship. Our study is based on the rankings over the ten last years, which will allow us to have a panel data over 10 different years. This is also a unique characteristic of our thesis in comparison to what has been done until now.
The first part of this thesis will focus on the literature that studied the different dynamics of organizational citizenship and organizational demographics. We will present how the literature is linking organizational demography and social identity to diversity and organizational citizenship. The literature review will allow us to picture our main hypothesis. The two first hypotheses, based on past empirical research, are in connexion with the level of homogeneity. We are going to explain why we believe that homogeneity in terms of gender and ethnicity leads to higher organizational citizenship. After that, we are going to present two other hypotheses that reflect the type of homogeneity. We are going to demonstrate why we believe that homogenous groups composed
6 mainly by woman and minorities have more chance to have a higher organizational citizenship than groups composed mainly by man and majorities.
After presenting our hypothesis we are going to formulate our different models. Our data will mainly be analyzed as a panel data model. This will allow us to control for omitted variable without actually observing them. The first two hypotheses are going to be scrutinized with a fixed effect model and a random effect model. To address to the third and fourth hypothesis we are going to use the same fixed and random effect model and add an interaction variable between the level of homogeneity and its type.
In the result part we will show that while homogenous groups in terms of gender seems to have an insignificant impact on organizational citizenship, there seems to be a significant positive correlation between homogenous groups in terms of ethnicity and organizational citizenship. We are also showing that the type of homogeneity seems to play a role only when it comes to the type of homogeneity in terms of gender. Indeed, groups composed mainly by women implies higher organizational citizenship than groups composed mainly by men. On the other hand, the type of homogeneity in terms of ethnicity doesn’t show any impact on organizational citizenship.
Our thesis is framed in four parts. The first part will present the literature review and the hypotheses. After that, the second part is presenting the data set and the different models that will allow us to analyze our hypotheses. Once this has been done, the third part will reveal our results and show if the hypotheses are supported or not. The last part will discuss our results, discuss further mechanism and draw the main conclusions and restrictions of the thesis.
7
Literature Review
The impact of Diversity on Organizational Citizenship can be explained using the Indicator explanation model introduced by Lawrence, B., (1997, p. 3). In this model, demographic variables are considered as indicators of subjective concepts that produce some outcome. Lawrence, B., (1997, p. 3) stipulates that when the reliability and validity are high, demographic variables are reasonable substitutes for subjective concepts that produce outcome. We will start our literature review by introducing the definition of the outcome: organizational citizenship. We will than study the link between diversity and organizational demography and social identity.
Figure 1: Indicator explanation model
2.1 Organizational Citizenship
Organizational Citizenship (OC) is defined by Organ, D. (1988, p. 4) as the “Individual behavior that is discretionary, not expectedly directly or explicitly recognized by the formal reward system, and in the aggregate promotes the efficient and effective functioning of the organization”. This definition shows that OC is a mechanism that influences the intrinsic behaviors of workers. Many
8 authors (Organ, D., 1988; Williams, L & Anderson, E., 1991; Shappe, S., 1998; Feather, N. & Rauter, K., 2004) are defining job satisfaction and organizational commitment as the key drivers of organizational citizenship. For this reason, we will follow the section with a detailed definition of these two concepts.
2.1.1 Job Satisfaction
Job satisfaction is “the extent to which people like (satisfaction) or dislike (dissatisfaction) their jobs’’ (Spector, 1997, p.2). Moorman, R. (1993) defines job satisfaction as “what job evokes in an effectively feeling-oriented measure.” Moreover, he adds that it shows respondent's feeling on the job and their feelings when working” (p.763). Locke (1976, p. 300) defines job satisfaction as being a “pleasurable and positive emotional state resulting from the appraisal of one's job or job experience”. Locke (1969, p.314) stipulates that job satisfaction comes from, first, the comparison between the worker’s expectations to the perceived job and, second, from the interaction between the workers and their environment. Locke (1976) and Moorman, R. (1993) are making a difference between affective and cognitive job satisfaction. While cognitive job satisfaction appears from a more logical and rational evaluation of the job conditions, the affective job satisfaction comes from a more emotional and compartmental state of the workers. For example, attitude of workers toward their task shows their cognitive job satisfaction. On the other hand, happiness of workers at work shows the affective job satisfaction. Wiliken (2013, p. 34) argues that job satisfaction has a large attention from scholars and from companies because it is recognized to have a direct impact on performance, contextual performance, and turnover.
2.1.2 Organizational Commitment
Organizational commitment is “an individual's psychological and behavioral involvement in a social group or unit of which he or she is a member” (Tsui et al., 1992, p.550). Scholl's (1981, p. 593) defines organizational commitment as “the relative strength of an individual's identification and involvement in an organization.”. For Weiner (1982), organizational commitment shows in which extent workers are preoccupied by the organization. organizational attachment is solely influenced by the work environment and not by the punishment and reward system of an organization (Williams, L. and Anderson,S., 1991, p.763). Moreover, following Weine (1982) model, organizational commitment shows the relative strength of an individual's identification and involvement to an organization.”. O'Reilly and Chatman (1986) introduces three levels of organization attachment. The first one is the involvement of the workers for extrinsic rewards, the
9 second is the willingness to be affiliated to a company and the third is the relation between the individual from the organizational values. Tsui et al., (1992, p. 554) identifies two main factors that are causing low organizational attachment. The first factor is described as arising from social isolation and low interpersonal attraction because of demographic dissimilarity. The second factor comes from non matching self-categorization of the group and its actual demography. Many studies have been studying the impact of organization commitment on several organizational outcomes, such as turnover (Golthelp et al., 2003, absenteeism (Tsui et al., 1992) or firm productivity (Tsui et al., 1989; Milliken, J. & Martins, L., 1996).
2.2 Organizational Demography Theory and Social Identity Mechanism
Most of the authors study the effect of groups characteristics on Organizational Citizenship basing their hypotheses on two theories: 1) Organizational Demography Theory and 2) Social Identification Mechanism. In this sections, we will thus, define these two theories. Moreover, while giving their definitions we will try to understand how they are connected with diversity and Organizational Citizenship. We believe that it is important to analyze them in detail in order to understand with precision why job satisfaction and organizational commitment are influenced by diversity.
2.2.1 Organizational Demography Theory
Organizational Demography Theory studies the effect of demographical composition of an organization on the organization’s behavior. Following this theory, the range of individual characteristics inside a group defines the property of an organization (Pfeffer, 1985). To understand how demography is going to influence the behavior of the organization, it is important to analyze the property of the group (Pfeffer,1985). These properties can be categorized by the composition of different demographic groups inside an organization. It can also be seen, as “how diverse an organization is”. For this reason, we can consider diversity as the key stone of organizational demography theory. There are two main types of diversity. The observable and the unobservable type of diversity (Tsui et al., 1992; Milliken, F. & Martins, L., 1996). The first subgroup of diversity includes observable characteristics such as gender or race (Tsui et al., 1992, p. 404). As it is visible, it directly influences the interaction among the workers. Even though these characteristics could be considered as very superficial, the literature is considering them as a type of diversity
10 influencing the interaction and the environment of an organization on the long term. The main reasons for this relation are stereotypes, biases and prejudices. The non-observable characteristics take more time to play a role but they have an important impact on how workers are interacting on long term. Milliken, F. & Martins, L., (1996, p. 413-415) are taking functional background and occupational background as non-observable characteristics.
The demographic composition of a group influences many organization’s behavioral patterns such as job satisfaction, organizational attachment, turnover and costumer-oriented behaviors. Jackson, S. et al. (1991) show that differences in a group impacts people's attitude. Based on this conclusion, they add that diversity will have an impact on organizational phenome such as job satisfaction or organizational attachment.
2.2.2 Social Identification Mechanism
Asforth, B. & Mael, F. (1989, p. 20) are defining social identification as follow: “Social Identification is a perception of oneness with a group. This social identification stems from the categorization of individuals, the distinctiveness and the prestige of the groups, the salience of outgroups, and the factors that traditionally are associated with group formation.” Based on this, Hogg, M., (2006, p. 112) is considering social identification theory as “the social psychologic analysis of the role of self-conception in group membership, groups process and intergroup relations”. From a cognitive point of view, categorizing helps individuals to interact in a social environment has it gives them a systematic means of defining others. This systematic way of thinking, helps individuals to classify themselves into a social environment. The social identification mechanism can be defined as “idiosyncratic characteristics based on social categories such as gender, race, age, educational background or religions” (Asforth, B. & Mael,F, 1989, p. 21). Turner et al. (1987) are summarizing this by writing “Social categorization is the cognitive basis of self-identity process.” The typical questions that are derived from the self-identification processes are “Is this my kind of organization?” “Do I belong here?”. Turner (1987) shows that when people are able to socially identify themselves to the others, they will tend to find the group more trustworthy, leading to a smoother interaction among the members of an organization. From this conclusion, many authors are stipulating that satisfaction and attachment among organizations are increasing when diversity is low (Ashforth, B, 1989; Jehn, K. et al., 1999; Tsui et al., 1992). Hofman, C. and Hurst, N. (1990) are taking gender diversity as an example to illustrate the mechanism of social identification and self-categorization. They analyze that if gender is a factor
11 for self categorization, the satisfaction among an organization will be higher when there is a large amount of workers from the same gender. Turner (1987) is observing that when social identity in terms of group membership is low, the worker will leave the group psychologically or physically. Based on this we can conclude that the theory would predict that high diversity leads to lower social identification and, thus, to lower job satisfaction and organizational attachment.
Many authors made a link between low social identification to a group and turnover (Godthelp, M. and Glunk, U., 2003; Jackson, S. et al., 1991). However, Tsui et al. (1992) make the very interesting comment that it is often complicated to leave the company physically. For this reason, they argue that low social identification could also bring the worker to leave the company psychologically. Leaving the company psychologically could be expressed by absenteeism (Tsui et al., 1992), poor performance (Tsui et al., 1989), low customer orientation (Joshi, A., 2006), conflicts or low intent to stay within the company (Tsui et al., 1992).
2.3 Empirical findings and Hypotheses
In this section we are going to present our hypotheses and show how previous empirical findings are supporting them. It is important to state that we will make two major distinctions: we are first going to analyze how the level of homogeneity in terms of gender and ethnicity influences organizational citizenship. After that we are going to analyze how different types of homogeneities are influencing organizational citizenship.
Hypothesis 1: Homogenous Work Organization in term of gender will increase Organizational Citizenship
We believe that either way, when working groups are composed with a majority of woman or man, it will result in a higher organizational citizenship. This one-sided hypothesis is based on multiple empirical findings. Bender, K., (2005) through a survey, analyzed the relationship between gender diversity and job satisfaction. After controlling for earnings, size of the company and performance, they found a monotonically increasing pattern between job satisfaction and the ratio of woman. The key finding of their research is that “The two coefficients for the highly female and highly male workplace are statistically higher than in groups with heterogeneous workplaces” (Bender, K. et al., 2005, p.492). Lee and Farh (2004) made a field experiment among 260 students of an organizational behavior course that had to work on a project in groups. They studied how the
12 independent variable gender diversity was correlated with the dependent variable group cohesion. The group cohesion variable was measured with surveys asking the students how close to the other group members they were feeling. Yet again, in this study, they found a negative relation between group cohesion and gender diversity (Lee, C. and Fahr, J.L., 2004, p. 146).
Hypothesis 2: Homogenous Work Organization in terms of ethnicity will increase Organizational Citizenship
This second hypothesis stipulates that a group composed homogenously in terms of minorities or majorities will have a higher organizational citizenship. This one-sided hypothesis relies on David, T. (1999) research. His conclusions are based on a group compartmental analysis of undergraduate students from 14 different countries. He showed that the number of different nationalities within the group had a negative relationship with the performance and collectivism level of the groups. Tsui et al. (1991) made an empirical research among 151 work units from three big companies. They found that there was a negative correlation between the level of heterogeneity within the group and organizational commitment. Moreover, they did further investigation and found out that that there was a negative relation between turnover, absenteeism and racial heterogeneity. On the other hand, Jehn, K. et al., (1999) did a study on 92 workgroups within a single company. They explored the influence race diversity on the worker's. They measured the motivation of the workers in terms of individual satisfaction, commitment to the groups and intent to remain in the company. They found the opposite of what the theory would have predicted. Indeed, their findings showed that diversity of social category had a positive impact on every chosen identifying aspect of workers’ motivation. In their discussion they stipulate that their contradictory findings may be due to the cross-sectional nature of the data and due to the fact that, as their study was based on only one company, they were not able to control for performance (1999, p.757).
Hypothesis 3: Asymmetric effect of Female Homogenous Groups Works and Male Group Works on Organizational Commitment
We believe that the type of homogeneity in terms of gender will not have the same effect depending on if there is a majority of woman or man in the group. This hypothesis is introduced by Chatman (2005, p.193), who says that “different group could react differently on being the same or not and
13 that not all differences are the same”. In his study, Chatman (2005, p.193) found that homogenous group solely composed of women had higher group commitment than homogenous group composed solely of men. Tsui et al. (1992, p.569) reversed the relationship by showing that “being different in gender has more negative effect on attachment for men than women”.
Hypothesis 4: Asymmetric effect of Minority Homogenous Groups Works and Majority Group Works On Organizational Commitment
This hypothesis is mainly based on hypothesis 4 and on Tsui et al. (1992, p. 569) research that are showing that homogenous groups with minorities have higher group commitment than homogenous groups composed by mainly majorities. There is a lack of literature on this topic. However, we want to test if we can find similar patterns than in hypothesis 4.
Our two first hypotheses test the direction of the relationship between organizational citizenship and homogeneity, for this reason we will have to consider those hypotheses as one sided. On the other hand, the two last hypotheses, looking for a relationship between two variables can be consider as two sided hypotheses. This will be important to keep in mind while testing our coefficient in the statistical part.
2.4 Summary of the Hypothesis
Organizational Characteristic Effect on Organizational Citizenship References 1)Homogenous Groups in Term of Gender
Positive effect on Organizational Citizenship
Bender et al. (2005); Lee and Farh (2004)
2)Homogenous
Groups in Term of Ethnicity
Positive effect on Organizational Citizenship
David, T. (1999) ; Tsui et al. (1991)
14 3)Homogenous
Female Group VS. Homogenous Male Group
Homogenous female group have higher organization citizenship than homogenous male group.
Tsui et al. (1991); Chatmann (2005)
4)Homogenous
Minority Groups VS. Homogenous Majority Groups
Homogenous Group have higher Organization Citizenship than Homogenous male group.
Tsui et al. (1991)
3. Context, Data and Design
3.1 Sample Construction
To test our hypotheses empirically, we decided to take the “100 Best Companies to Work for in America” (BCW) as a proxy to the organizational citizenship level of a company. We choose to take this proxy based on other papers that analyzed organizational citizenship using this ranking (Fulmer et al., 2003; Bernardi et al. 2006; Simon and Devaro, 2006). BCW is a yearly list that ranks the 100 US based companies accordingly to their level of organizational citizenship. The BCW is published by Fortune, an American business magazine publishing a large set of rankings. Fortune’s rankings are highly globally recognized, we can thus consider their ranking as a gauge of the business markets and a largely recognized research and consulting company when it comes to Human resources matters.
The companies are mainly rated in respect to an employee survey that is reflecting the level of job satisfaction and commitment of the employees. The survey is provided by the “Great Place to Work Institute”, defining itself as “The business leaders and researchers to establish the standard that defines a great workplace”. A great place to work had to fulfill three main criteria: 1) Trust for the organization 2) Pride of doing what they do and 3) Enjoy the people they work with. (Fulmer, 2003). The companies that are part of the survey campaign must count more than 500 employees and be at least 10 years old. The BCW exists since 1998, but we could only find the data from the ranking from 2006 until 2015.This gives us 1000 ranks data about the level of organizational citizenship of 286 different companies.
15 To simplify our analysis, we inversed the position between the best and the worse ranked companies. This means that a company that has the rank 100 has the highest level of organizational citizenship of the sample in a given year. On the other side, the company that has the rank 1 has the worse level of organizational citizenship within the sample in a given year. Our two main independent variables assessing the demographic homogeneity of an organization are based on the ratio of woman and ethnic minorities within the company. We found this data on the fortune website. Fortune is giving the data about the minorities and woman ratio for the companies that are inside the ranking. This will allow us to use panel data for our analysis. However, as the we have only data about the organizational composition of the companies that are inside the ranking, our panel will be unbalanced. From these ratios we created two homogeneity variables. Our homogeneity variables are going from 1 to 50 and are calculated as follow:
Gender Homogeneity Variable = |𝑟𝑎𝑡𝑖𝑜 𝑤𝑜𝑚𝑎𝑛 − 50|, this means that when there are 50 % of women and 50% of men, the variable is equal to 0 and indicates a perfectly heterogeneous group. On the other hand, when there is 0% or 100% of women, the variable will be equal to 50 and will indicate a completely homogenous group. We decided to take the absolute value to only show the level of gender homogeneity in the organization. This allows us to isolate the effect of the level of homogeneity from the type (male or female) of homogeneity.
Ethnic Homogeneity Variable = |𝑟𝑎𝑡𝑖𝑜 𝑒𝑡ℎ𝑛𝑖𝑐 𝑚𝑖𝑛𝑜𝑟𝑖𝑡𝑖𝑒𝑠 − 50|, this means that when there are 50 % of minorities, the variable is equal to 0 and indicates perfectly heterogeneous groups. On the other hand, when there is 0% or 100% of minorities, the variable will be equal to 50 and will indicate a completely homogenous group. We decided to take the absolute value to solely show the level of gender homogeneity in the organization. This allows us to isolate the effect of the level of homogeneity from the type (minority or majority) of homogeneity.
Moreover, we create two dummy variables that are reflecting the kind of homogeneity. The dummy variable is equal to 1 when there are more women than men. This means that when the dummy variable is equal to 1, we have a woman type of homogeneity. We are applying the same scheme for minorities. When there are more minorities than majorities in the company our dummy variable is equal to 1. In this case, when the dummy variable is equal to 1 it thus indicates that we have a minority type of homogeneity.
16 To summarize, we have the ranking from 1 to 100 that indicates the level of organizational citizenship, with 100 that shows the maximum level. Moreover, we have the gender homogeneity variable and the ethnic homogeneity variable from 0 to 50 indicating the level of homogeneity within an organization. The maximum level of 50 shows a completely homogenous organization. We are also adding dummy variables indicating the type of homogeneity. Our dummy variables are equal to one when it is a type of homogeneity with mainly women or minorities.
3.2 Control Variables
The control variables are not our main concern in this study, however we have to add those control variables because the ‘’regressor included to hold constant factors that, if neglected, could lead the estimated effect of interest to suffer from omitted variables bias’’ (Stock and Watson, 2003, p. 280). We need to add the control variables to make sure that the analysis we get from the homogeneity variable and dummy variables only show their correlation with homogeneity and that our analysis doesn’t suffer from omitted variable bias. The control variables will thus help us to analyze the precise correlation between organizational citizenship and demographic homogeneity without any noise from other uncontrolled variables. Our models will include the following control variables: company size and type of sector. We decided to include these variables because based on the literature we have reasons to believe that they will influence organizational citizenship. Our model would have been more accurate if we could have added more control variables, however due of lack of scope and availability these are the only control variables we could obtain. Moreover, based on the previous study testing the effects of different demographic variables, they also represent, in a big majority, what previous studies have used.
We decided to include the size of the company based on the existing literature we have reason to believe that the employees’ involvement will decrease with the size of the company, the bigger the company, the lower the involvement (Hodson & Sylivan, 1985; Tsui et al., 1992; Fulmer et al., 2003). For example, in the empirical research of Hosdson & Sulivan (1985), that studied the different level of job commitment and job satisfaction in various kind of companies, they found that the level of job commitment and satisfaction was greater in small companies than in large companies. Our module will include the number of employees as proxy for the size of the company. This same proxy as already been used by Tsui et al. (1992) while studying the relational between
17 relational demography and organizational attachment. We are going to use the number of employee in a given company and in a given ranking based on the information on the fortune website. The types of industry have been added to our model based on the empirical research of Kontoghiorghes, T. & Bryant, K. (2004) and Tepeci, C. & Bartlett, M. (2002). The industry type can influence the ranking in two different ways. First of all, based on the field research of Kontoghiorghes, T. & Bryant, K. (2004) it seems that industry differences are linked to different work values and very different work environment that might influence the level of satisfaction of the employees. Moreover, Jackson et al. (1991) stipulate that the type of industry influences “the number of firms, productivity growth, rate of the industry and the volatility of the economics environment” (p. 678) and that this has a direct impact on the employees’ turnover, thus on employees’ attachment toward their companies. On the other side, as shown by Tepeci, C. & Bartlett, M. (2002), certain kind of industries might attract certain kind of people. For example, more competitive industries might attract more hard working employees that care less about their work conditions. Those different intrinsic people’s characteristics might influence the way they are answering the survey formulated by a “Great Place to Work Institute”. We collected the information about the kind of industries of the companies on the website of the company.
3.3 Summary Statistics
As we can see in the table 3, we have the data on 997 rankings of over 10 years. We have 3 data missing but as the mean of the ranking is 50.447 we can conclude that it doesn’t have much impact on the overall mean. By analyzing the companies’ size we can see that there is a big difference between the companies. This large difference in control is a good point for our models because it will allow us to analyze organizational attachment in various ways. When focusing on our main independent variable we can see that there is more ethnic homogeneity than gender homogeneity. Moreover, the standard deviation is about the same in both level of homogeneity. We can also note that there are missing values due to lacking information about the level of gender (12 missing values) and ethnic homogeneity (30 missing values). When we encountered a missing value, we sat them to the mean of the value. This shouldn’t distort our results but make them more internally consistent. Moreover, when looking at the dummy variable showing if there is a majority of women or minorities in the company or not, we can see that for ethnic homogeneity, only 77
18 companies on 998 (11 %) available data have a minority type of homogeneity. On the other hand, when looking at the gender homogeneity we can see that 388 companies (40%) have a woman type of homogeneity.
Table 3: Summary Statistics
Variable Observation Mean
Standard
Deviation Minimum Maximum Data Source
Y
RANK 994 50.45 28.9051 1 100 Fortune 100
X
Level of Gender Homogeneity 994 15.235 10.62133 0 46 Fortune 100
Level of Ethnic Homogeneity 994 22.510 11.88821 0 49 Fortune 100
Minorities Dummy 994 0.11 0.31 0 1 Fortune 100
Woman Dummy 994 0.4 0.49 0 1 Fortune 100
W
Company Size 994 15557.62 28722.64 114 233457 Bloomberg
Electronics 994 0.010 0.099 0 1 Company Website
Healthcare 994 0.033 0.179 0 1 Company Website
Media 994 0.004 0.063 0 1 Company Website
Pharmaceutical 994 0.0135 0.114 0 1 Company Website
Retail 994 0.165 0.3713 0 1 Company Website
Transportation 994 0.008 0.089 0 1 Company Website
Engineering 994 0.053 0.225 0 1 Company Website
Entrainment 994 0.004 0.063 0 1 Company Website
Advertising 994 0.009 0.0948 0 1 Company Website
Biotechnology 994 0.021 0.144 0 1 Company Website
Construction 994 0.0591 0.236 0 1 Company Website
Financial 994 0.140 0.350 0 1 Company Website
Food 994 0.007 0.0837 0 1 Company Website
Healthcare 994 0.120 0.325 0 1 Company Website
Information 994 0.132 0.338 0 1 Company Website
Manufacturing 994 0.033 0.179 0 1 Company Website
Professional Services 994 0.187 0.389 0 1 Company Website
19
3.4 Design of the Model
We decided to consider our data set as a panel data set. Panel data are used when each entity is observed over two or more years. In our sample the entities are represented by the firms and we have data about them over the last 10 years (from 2005 to 2015). This type of data set is particularly interesting because it will allow us to observe how the experiences inside the entities and over the years are influencing organizational citizenship (Stock & Watson, 2015, p. 57-58). The main advantage of panel data is that it allows us to control for variables without observing them (Stock & Watson, 2015, p.396). We believe that it makes more sense to use panel data than the Least Squares Estimating Method (OLS) as the theory says that the OLS is only determinative when there is a low correlation between the independent variables. Indeed, when there is a high correlation between the independent variables, it will lead to very high stand error and thus to low t-statistic. It may also lead to wrong estimations of the coefficients and also contradictory results between the significance and R-squares results. Panel data has the merit to reduce this estimation bias (Stock & Watson, 2015, p. 398).
It is possible to analyze panel data using two different models. The first model is the fixed effect model and the second model is the random effect model.
3.4.1 Fixed Effect Model
The fixed effect model analyses the correlation between the dependent and independent variables within one entity and is use as a method for omitted variables. We use this model when omitted variables vary across entities (companies) but don’t change over time (Stock,J. & Watson,M., 2015, p.403). This means that there is a correlation between the entities error term and the predictor variables. The main advantage of this model is that it will allow us to control for omitted variables that vary from an entity to another but that remain constant over time without even observing them. For example, the geographic location of a company, the type of industry and the year may influence the ratio of woman or minorities and job satisfaction without we can observe them.
To asses our hypotheses, with the fixed effect model, we established following model:
(1) RANKit= 𝛽0+ 𝛽1GenderHomogeneityit+ 𝛽2EthnicHomogeneityit + 𝛾1 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝑆𝑖𝑧𝑒𝑖𝑡+ λt
20 RANKit is the ranking of a given company “i” in year “t” in the fortune 100 ranking, it represents
the dependent variable, and is used in our analysis as a proxy for organizational citizenship. The main drawback of taking a rank as dependent variable is that it is not showing the magnitude between the companies. This means that we cannot measure the exact difference between two companies in terms of organizational citizenship. We believe that it will still be a good predictor for the organizational citizenship of the companies. However, we will have to keep this point in mind while examining the results.
𝛽1; 𝛽2; 𝛾1; are the regression coefficient. 𝛽1;2 𝛾1 are non- binary variables and represent respectively the coefficient of the level of homogeneity (𝛽) and the control variable (𝛾). ∝i (i=1,…,n), is the variable that makes this model unique as it represents the time invariant specific intercept of each company. And 𝑢𝑖𝑡 represents the error term.
One of the problem from the fixed effect model is that there might be some variation from one entity to another that are actually changing over time. For example, a company might want to implement a new maternity leave policy, or a quota of minorities or women within the company. These policies might have an impact on both, the regressor and the dependent variable, without actually capturing the effect we want. For this reason, this model could suffer from omitted variable bias. However, we can assume that these company policies are directly influenced by governmental policies and that those trends are the same across entities. To control for the trends we thus add a time fixed effect (∝𝑖) , that controls for variables that are constant between entities but are changing
over time (Stock,J. & Watson,M., 2015, p.403). 3.4.2. Random Effects Model
The main difference between the fixed effect model and the random effect is that the variation crosswise the different entities is considered to be random and not correlated with the independent variables. In this case, we believe that there is no common trend across the entities. The RE model is hence controlling for variables that are changing across entities but not across time. For this reason, the RE model requires more control variables than the fixed effect model, leading to higher treat for omitted variable bias.
(2) RANKit= 𝛽0+ 𝛽1GenderHomogeneityit+ 𝛽2EthnicHomogeneityit + 𝛾1 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝑆𝑖𝑧𝑒𝑖𝑡 + 𝛿I (𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑠𝑒𝑐𝑡𝑜𝑟)i + 𝜀𝑖𝑡 + 𝑢𝑖𝑡
21 RANKit is the ranking of a given company “i” in year “t” in the fortune 100 ranking, represents the
dependent variable, and is used in our analysis as a proxy of organizational citizenship. 𝛽1, 𝛽2, 𝛽3 are the non-binary regression coefficient. 𝛾1, 𝛾2 are non- binary control variable. As the RE requires control for time invariant and company specific variables, we are adding a dummy control variable for the type of industry 𝛿I. The specificity of this model is the apparition of two different error term. The first error term (𝜀𝑖𝑡) represents the between-entity error and the second (𝑢𝑖𝑡) is the
within-entity error term.
3.4.3 Type of homogeneity tests
Finally, we are going to build a new model to test how the type of homogeneities are correlated with organizational citizenship. First, we are going to apply Tsui et al. (1992, p. 570) model by dividing our sample in two subgroups. We are going to compare the difference in means between a homogenous women group and a homogenous men group and between a homogenous minority group and a homogenous majority group. To test the significant difference between the two subgroups we are going to use a T-student statistic test. It will allow us to test whether the means are significantly different or not. If we fail to reject the null hypothesis it will mean that the means are significantly different between the two subgroup (Stock & Watson, 2015, p. 123). This first test is a good test to get a general overview. However, we will have to be careful with the interpretation of the results as this simple mean comparison is lacking control. The second way to make a conclusion regarding our hypotheses, is to add interaction variables to the model we presented in the last sections. Interaction variables are used when there are arguments that the effects of an independent variable are influenced by another independent variable (Stock & Watson, 2015). In this case, we believe that the type of homogeneity will influence how homogeneity will impact the organizational citizenship. The model with the interactions will be presented as follow:
(3) RANKit= 𝛽0+ 𝛽1GenderHomogeneityit+ 𝛽2EthnicHomogeneityit +
π1GenderHomogeneity*D1+π2EthnicHomogeneity*D2 + 𝛾1 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝑆𝑖𝑧𝑒𝑖𝑡+ πt(yeart)+∝𝑖 +
𝑢𝑖𝑡
(4) RANKit= 𝛽0+ 𝛽1GenderHomogeneityit+ 𝛽2EthnicHomogeneityit +
π1GenderHomogeneity*D1+π2EthnicHomogeneity*D2 + 𝛾1 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝑆𝑖𝑧𝑒𝑖𝑡 + 𝛿I (𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑠𝑒𝑐𝑡𝑜𝑟)i + 𝜀𝑖𝑡 + 𝑢𝑖𝑡
22 Equation (3) is the fixed effect model and the equation (4) the random effect model. π1 and π2 are the two regressors showing the effect of the two dummy variables (D1 and D2) in interaction with the level of homogeneity on the correlation with organizational citizenship.
3.4.4 Statistical Problems
We need to be careful with the exact interpretation of our results. Indeed, our analysis seem to possibly be affected by different statistical biases. The first one is that our sample might be biased by reversed causality. This would mean that the dependant and independent variables are simultaneously influencing each other. On a practical level, it would indicate that the demographic configuration of an organization is influenced by its organizational citizenship. For example, women might be more willing than men to work for a company that presents a good work atmosphere. If this is the case it would mean that our model has a simultaneous causality and that our estimators are biased and inconsistent (Stock, J. & Watson, M., 2015, p. 373). To address this problem, we need to measure how strong the influence of the level of homogeneity is on the chance of a company to be in the sample. Our solution for measuring this relation is to evaluate the correlation between the mean ratio of woman and minorities within a company and the number of times the company is appearing in the ranking. The appendix 1 addresses in detail reversed causality. The results from appendix 1 are showing that there seem to be no relationship between the ratio of woman and minorities within a company and the number of times the company is appearing in the ranking. This allows us to believe that the model is not biased by reversed causality.
Another statistical issue could arise from the sample selection bias. Stock,J. & Watson, M. are defining sample selection as “arising when a selection process influences the availability of data and that process is related to the dependent variable, beyond depending on the regressors” (2015, p.372). In our sample, this would mean that the level or type of homogeneity influences the fact that a company is inside the fortune ranking or not. In this is the case it will lead to bias and inconsistencies of the OLS estimator (Stock, J. & Watson, M., 2015, p.372). To address this problem, we need to measure how strong the influence of the level of homogeneity is on the chance of a company to be in the sample. As showed in appendix 2, the results are once again very small and insignificant. This lead us to believe that our data set is not suffering from sample selection bias. However, we need to be very careful with this conclusion from appendix. Indeed, as we only
23 have data about the level of homogeneity of the companies when they are inside the ranking, the test we used is thus only an approximation and an indirect way to assess sample selection bias.
4. Results
We can apply the Haussmann test to decide which of the two models are the best fit for analyzing our data set. The aim of this test if to check if there is a common trend between the entities. It evaluates if the errors (ui) are correlated with the regressors. In our case, the null hypothesis is that there is no correlation between ui and the regressors. In a practical matter, it means that if the test
rejects the null hypothesis, we will prefer to use the random effect model. On the other hand, if we fail to reject the Haussmann test, it means that there is no reason to believe that there is a common trend and that FE and RE should be close enough to use them simultaneously (Wooldrifge, 2009, p. 493).
On the table below we can see the results from the Haussmann Test. We cannot reject the null Hypotheses that the coefficients are not systematically different from each other as Prob>chi2 is larger (0.45) than the significance level of 5% (or p>0.05). This means that we can conclude that both the fixed and random effects are similar. However, the standard errors of the random effect should be lower than the one for fixed effects. For this reason, we will consider random effect model more efficient.
As our two first hypothesis are testing the sign of a relationship we have to consider them as one-sided hypothesis. We will thus need to take the appropriate p-values for one tailed hypotheses in our statistical tests.
As already mentioned in the previous part, we have inversed the ranking to make the analyses easier to understand. This means that a rank of 100 indicates a maximum level of organizational citizenship. To make our result tables easier to understand we made summary of what the signs of our main coefficient β1, β2, π1 and π2 would indicate.
Table 3: Sign Summary
β1 When β1 is positive it means that an increase
of organizational citizenship within a company is positively correlated with an increase of homogeneity in terms of gender within the company
24 β2 When β2 is positive it means that an increase
of organizational citizenship within a company in positively correlated with an increase of homogeneity in terms of ethnicity within the company
π1 When π1 is positive it means that an homogenous groups composed mainly by woman have a larger impact on organizational citizenship than a homogenous group composed by mainly by men
π2 When π2 is positive it means that a homogenous groups composed mainly by minorities have a larger impact on organizational citizenship than a homogenous group composed mainly by non-minorities groups.
Gender Homogeneity and Organizational Citizenship
Based on our hypothesis (1) we expected organizational citizenship to be positively correlated with gender homogeneity in companies. We can see that the results are showing an opposite effect of what we would predict, as all the coefficient are negative (β1reis between -0.186 and -0.204 and β1FE
is between -0.0706 and -0.003 in the fixed effect model). However, by looking at the one tailed hypothesis test we can see that either in fixed or random effect the coefficients are not significant. Indeed, taking the fixed effect model or the random effect model doesn’t impact the significance level of our coefficients. The fact that we reject the coefficient β1RE despite it is a very large
coefficient, make us doubt about the power of our study. It could be that our coefficients are not significant because we are not comparing enough entities between each other. If this it is the case, it would mean that our insignificant results are due to a sample size issue and not to an effect size issue.
Ethnic Homogeneity and Organizational Citizenship
In this section we analyze the hypothesis (2) that stipulates that organizational citizenship is positively correlated with ethnic homogeneity. By analyzing the regressions with fixed effects, the regressors are first showing us what the theory predicts. Indeed, organizational citizenship seems to be positively correlated with ethnic homogeneity (β2RE is equal to between 0.24 and 0,191 while
25 β2FE is equal to between 0.0286 and 0.056 in the fixed effect model). Analyzing the significant
level, shows us that this results are significant only when we use the random effect and our results are becoming insignificant when using the fixed effect model. The fact that the random effect model has lower standard error than fixed effect model (SERE=0.130 and SEFE=0.259) and the
result for Hausmann Test leads us to believe that the covariance between our estimators and the time invariant specific intercept of each company is equal to zero. This would mean that companies do not have time invariant characteristics that are influencing their level of homogeneity in terms of gender or ethnicity. On a practical point of view, this conclusion is however not very plausible and make us thus doubt about the consistency of our random effects coefficients. Moreover, the difference of the results between the random effect and the fixed effect are surprising. Indeed, the Haussmann Test indicates that both the fixed and random effects should give similar outcome as the unique error terms are not correlated with the regressors. As we still find very different results we believe that it is because we are lacking control for the random effect while we don’t have this problem in the fixed effect model.
Homogenous Woman Group vs. Homogenous Men Group and Organizational Citizenship
This third hypothesis stipulates that homogenous organization composed mainly by women will have a higher organizational citizenship than a homogenous organization composed mainly by men. When dividing our data set between homogenous woman and men group and comparing the rank means of these two groups we cannot stipulate that the rankings are statistically different (t=-0.8714). From this first analysis we can thus not confirm what the theory would have predicted. After that we introduced an interaction variable between a dummy indicating if it there is a majority of women or not to the level of homogeneity. As shown in the results (4) and (5) of the table 3, the effect of homogenous women group seems to have an opposite effect that we would predicted. Indeed, the coefficient from the table 3 are negative (π1 respectively equal to -0.0360 and -0.251). However, these coefficients do not provide strong evidence to reject our hypothesis as none of the coefficients are significant.
Homogenous Minorities Group vs. Homogenous Majorities Group and Organizational Citizenship The fourth hypothesis stipulates that a homogenous organization composed mainly by minorities will have a higher organizational citizenship than a homogenous organization composed mainly by majorities. When dividing our data set between homogenous minorities and majorities group and
26 comparing the rank means of these two groups we can stipulate that the rankings are statistically different (t=-4.75). For this reason, we can confirm, that homogenous groups composed by minorities or majorities seem to have a different impact. This is however not confirmed by the result from table 3. Indeed, the coefficient are negative and not significant.
We believe that to get significant results for our two last hypotheses we would need more precision in defining exactly the type of homogeneity. Indeed, it might be possible that very homogenous male group vs. very homogenous female group have different effects that low homogenous male group vs low homogenous female group. If this is the case, our model is not the best fit as the relationship between the type of homogeneity and organizational citizenship is not linear.
27 Table 1: Hypothesis 1-4 testing regression results
(1) (2) (3) (4) (5)
VARIABLES Random Effect Random Effect Fixed Effect Random Effect with
interaction
Fixed Effect with interaction Organizational Citizenship Organizational Citizenship Organizational Citizenship Organizational Citizenship Organizational Citizenship GenderHomogeneity -0.186 -0.204 -0.0706 -0.191 -0.00327 (0.139) (0.171) (0.310) (0.170) (0.291) EthnicHomogeneity 0.247*** 0.190* 0.0286 0.191* 0.0554 (0.125) (0.130) (0.259) (0.131) (0.260) WomGenderHomogeneity -0.0360 -0.256 (0.187) (0.275) MalGenderHomogeneity -0.0855 -0.0484 (3.89e-05) (5.13e-05)
Employees -0.000121 -9.35e-05 -0.000121 -9.27e-05
(3.90e-05) (5.13e-05) (3.89e-05) (5.13e-05)
Electronics -19.17*** -17.85* (5.858) (9.287) Healthcare -8.645 -7.411 (8.599) (9.003) Media 1.137 2.201 (20.86) (21.95) Pharmaceutical -1.570 -1.198 (8.085) (8.039) Retail -2.840 -2.454 (5.521) (5.591) Transportation -10.84 -10.06 (23.53) (23.44) Engineering -7.280 -7.229 (7.228) (7.217) Entrainment 44.39*** 44.45*** (3.119) (3.077) Advertising 55.59*** 57.22*** (2.974) (4.810) Biotechnology -3.071 -2.735 (15.27) (15.32) Construction -1.186 -1.109 (6.808) (6.789) Financial 5.442 5.998 (5.919) (6.280)
28 Food -20.53*** -20.34*** (4.867) (4.781) Healthcare -5.347 -4.229 (4.634) (6.334) Information 6.991 7.143 (5.164) (5.160) Manufacturing 8.134 8.304 (11.77) (11.73) Professional Services -0.602 -0.159 (5.565) (5.625) 2007 -1.841 -1.829 (2.577) (2.589) 2008 -2.886 -2.886 (3.215) (3.199) 2009 -2.805 -2.781 (3.201) (3.193) 2010 -5.058 -4.916 (3.533) (3.509) 2011 -4.526 -4.424 (3.531) (3.532) 2012 -4.500 -4.423 (3.767) (3.756) 2013 -5.507 -5.483 (3.726) (3.669) 2015 -6.261* -6.305* (3.710) (3.712) 2016 -4.452 -4.848 (3.833) (3.784) (0.231) (0.235) Constant 42.27*** 45.74*** 56.22*** 45.39*** 56.32*** (4.305) (5.260) (9.387) (5.256) (9.210) Observations 994 994 994 994 994 R-squared 0.015 0.016 Number of company 234 234 234 234 234
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Haussmann Test
Test: Ho: difference in coefficients not systematic chi2(8) = 2.35
29
5. Mechanisms
A large number of research have been conducted concerning the connexion between ethnic and gender composition and how well a team is performing (Cohen, M. and Bailey, S., 1997; Steewart, G., 2006; Hoogendoorm, S. and van Praag, M., 2012; Hoogendoorn, S. et al., 2013. For example, Hoogendoorm and van Praag (212) are attributing the positive effect of heterogeneous groups to the fact that “truly ethnically diverse teams benefit from a more diverse pool of relevant knowledge facilitating (mutual) learning.” (p.26). On the other hand, many authors are also attributing high firm performance to high organizational citizenship (Brow, L. & Peterson, R., 1994; Judge,M. et al., 2001). In our study these different findings are misleading because it would mean that as homogeneity leads to high organizational citizenship, homogeneity should as a result lead to high performance. This assumption would however contradict Hoogendoorm, S. and van Praag, M. (2012) conclusions.
For this reason, we would like to study if evidence of a relationship between the performance of the companies and their demographic composition can be found. To study this, we introduce a new model with sales and profit as an outcome. This gives us following model:
(1) Company Salesit= 𝛽0+ 𝛽1GenderHomogeneityit+ 𝛽2EthnicHomogeneityit + ∝𝑖 + 𝑢𝑖𝑡
(2) 𝐶𝑜𝑚𝑝𝑎𝑛𝑦 𝑃𝑟𝑜𝑓𝑖𝑡= 𝛽0+ 𝛽1GenderHomogeneityit+ 𝛽2EthnicHomogeneityit + ∝𝑖 + 𝑢𝑖𝑡
In table below, we can see some evidence that could confirm the findings of previous authors. Indeed, we can see that all our coefficients are negative. This indicates that there is a negative relationship between homogeneity and firm performances. This is observed for both proxies of performance, sales and profit. However, as none of our coefficients are significant we would need further research with more control variable to see if we can get significant results. For this reason, in this study we cannot really make clear assumption on how organizational citizenship, group performance and group demographics are connected. A further study about how organizational citizenship is mediating the relationship between group performance and group homogeneity could nicely link all these contradictory findings.
30 Table 2: Homogeneity and Company Performance
(1) (2) (3) (4)
Fixed Random Fixed Random
VARIABLES Profit Profit Sales Sales
EthnicHomogeneity -320.6 -124.8 -273.7 -271.4 (367.1) (170.3) (590.6) (338.1) GenderHomogeneity -392.1 -46.18 -348.7 -317.6 (262.1) (205.3) (425.9) (341.1) Electronics -36,783*** -40,037*** (2,173) (5,504) Healthcare -33,563*** -43,041*** (2,430) (2,583) Media -26,625*** -13,590 (7,668) (15,116) Pharmaceutical -34,993*** -46,305*** (1,519) (2,539) Retail -37,217*** -41,886*** (3,516) (8,714) Transportation -32,358*** -31,912*** (3,635) (8,528) Engineering -28,106*** -31,711*** (4,585) (5,136) Entrainment -19,289*** -23,637** (6,829) (10,346) Advertising -31,660*** -36,419*** (3,094) (5,642) Biotechnology -31,725*** -39,525*** (4,392) (6,048) Constant 23,400*** 41,557*** 31,792*** 59,191*** (7,205) (2,982) (9,616) (4,915) Observations 169 169 168 168 R-squared 0.026 0.021 Number of company 48 48 47 47
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
31
6. Concluding Discussion
In this thesis we studied the correlation between different demographic variables within a company and its organizational citizenship level. Based on subjective theories, about social identity, organizational demography and past empirical researches, we formulated our four main hypotheses. While the two first hypotheses are reflecting the level of homogeneity, the two others are based on the type of homogeneity. The two first hypothesis are stipulating that there will be a higher organizational citizenship within companies with more homogenous groups. The latter are stating, that homogenous groups composed mainly by women or minorities, will have a higher organizational citizenship level than groups composed mainly by men or majorities. The results show, for the first hypothesis, an opposite pattern of what our hypothesis would have predicted. Indeed, the results lead us to believe that there is a negative relation between organizational citizenship and homogenous groups in terms of gender. However, the results for this hypothesis are not showing a statistical significance. The second hypothesis stipulating that homogenous groups in terms of ethnicity will lead to higher organizational citizenship are confirmed by the pattern and the significance level of our results. The homogeneity type results are contracting both of our hypothesis. Indeed, for the third hypothesis, concerning the type of homogeneity in terms of gender, we can see that male homogenous groups seem to have higher organizational citizenship than female homogenous groups. However, these results show a very small statistical significance. On the other hand, when looking at how the type of homogeneity in terms of ethnicity is influencing organizational citizenship, we cannot conclude that majorities’ homogenous groups seem to have higher organizational citizenship. Here, it is important to notice, that because of the different limitations of our study that we are going to present later, we cannot strictly affirm that our results are going to be observed in every company. In addition to that, we cannot conclude either that companies should only have homogenous groups. Indeed, as presented in the mechanism part, while in this work we presented the impact of homogeneity on organizational citizenship, there are many other organizational performances being negatively affected by homogeneity. For example, creativity or efficient group work. This point would constitute a very interesting point for further research. In fact, we could imagine the same kind of ranking but for other organizational performance factors and see how organizational citizenship mediates the relationship of homogeneity and to other performance indicators. We need to underline that taking fortune ranks as a proxy is linked to limitations. First of all, as the analysis has been done in the United Stated
32 we cannot certify that the results would have been similar in other countries or continents. This is why we can only say that our results are country specific for the United States. Moreover, the fortune ranking does not solely take the survey answers into account. Indeed, part of the ranking is also based on grades that a jury is giving to the companies work environment. Therefore, we cannot acknowledge with certainty that the ranking is only representing the organizational citizenship of a company.
We believe that our results are interesting so far as they confirm that some demographic set up influences organizational citizenship. This is a very interesting finding that confirms the importance for companies to take these demographic variables into consideration while implementing different company policies and that diversity management is a very important point for companies that want to reach a high organizational citizenship level.
33
7. Appendix
Appendix 1: Reversed Causality Check
To test it, we are going to build the following model:
(5) GenderHomogeneityit= α + β1 RANKit-1 + ... + βqRANKit-q + et
(6) EthnicHomogeneityit= α + β1 RANKit-1 + ... + βqRANKit-q + et
GenderHomogeneityit and EthnicHomogeneityit are the level of homogeneity in time t for a given
company i. RANKit-q represents the q lagged ranking a given company i. By looking at our
results on table 3, we can confirm that while the lagged value in time -1 is large and significant (β 1=-0.265) the lagged values from the year before seem to follow a random pattern with β that is going from positive to negative. Moreover, not all the coefficients are significant. From these results, we can conclude that the level of homogeneity in terms of gender doesn’t seem to be correlated with its past level of organizational citizenship and therefore we can say that there is no reversed causality between the level of gender homogeneity within a company and the past ranks. When looking at the results for model 6, it seems that we could come to the same
conclusion. Indeed, the coefficient seems to follow a random pattern as well and doesn’t appear to be significant.
34 Table 3: Reversed Causality Test
(1) (2) VARIABLES Gender Homogeneity Ethnic Homogeneity t-1 -0.265** 0.177 (0.0999) (0.109) t-2 -0.0195 -0.238 (0.182) (0.199) t-3 -0.0544 -0.123 (0.113) (0.123) t-4 0.124 -0.107 (0.106) (0.113) t-5 0 0 (0) (0) t-6 0.0736 0.0370 (0.108) (0.117) t-7 -0.0473 0.335** (0.125) (0.136) Constant 28.84*** 9.604** (4.026) (4.326) Observations 31 31 R-squared 0.405 0.340
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Appendix 2: Sample Selection Bias
We are measuring sample selection bias by evaluating the correlation between the mean ratio of woman (ratiowoman) and minorities (ratiominorities) within a company and the number of times the company is appearing in the ranking (NOT). We also decided to add the mean number of employees as control variable (Memployees). We made a simple OLS model out of it:
35 Following the result of table 4, we can see that the average homogeneity level in terms of gender doesn’t seem to have an impact on the selection of the sample. The coefficient is very small (𝛽1= -0.000403) and not significant (t= -0.04). When looking at the average level of homogeneity in terms of ethnicity we can see that the coefficient is higher (𝛽2=0.0387), however the coefficient is not significant (t=1.39). From this analysis, we can conclude that there is no sample selection bias.
Table 4: Sample Selection Bias
VARIABLES Number of time a company is appearing in
the ranking Mean ratio Woman -0.000403
(0.00937) Mean ratio Minorities 0.0387
(0.0279) Mean Employees 2.28e-05***
(3.73e-06)
Constant 6.145***
(0.128)
Observations 992
R-squared 0.041
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
36
8. Bibliography
Ancona, D.G., & Caldwell, D.F (1988). Beyond task and maintenance defining external functions in groups, Group & Organization Management, 13(4), 468-494
Ashforth, B.E. & Mael, F. (1989). Social Identity Theory and the Organization, Academy of Academy of Review, 14(1), 20-39
Baron, J.N. & Pfeffer, J. (1994). The social psychology of organizations and inequality, Social Psychology Quarterly, 190-209
Bender, K.A., Donohue, S.M. & Heywood, J.S. (2005). Job satisfaction and gender segregation, Oxford economic papers, 57(3), 479-496
Bernardi, R.A., Bosco, S.M. & Vassill, K.M. (2006). Does Female Representation on Boards of Directors Associate With Fortune's “100 Best Companies to Work For” List?, Business & Society, 45(2), 235-248
Chatman, J.A. & O'Reilly, C.A. (2004). Asymmetric Reactions to Work Group Sex Diversity among Men and Women., Academy of Management Journal, 193-208
Cox, T.H., Lobel, S.A. & McLeod, P.L. (1991). Effects of ethnic group cultural differences on cooperative and competitive behavior on a group task, Academy of Management journal, 34(4), 827-847
Feather, N.T. & Rauter, K.A. (2004). Organizational citizenship behaviours in relation to job status, job insecurity, organizational commitment and identification, job satisfaction and work values, Journal of occupational and organizational psychology, 77(1), 81-94
Fulmer, I.S., Gerhart, B. & Scott, K.S. (2003). Are the 100 best better? An empirical investigation of the relationship between being a “great place to work” and firm performance, Personnel Psychology, 56(4), 965-993
37 Godthelp, M. & Glunk, U. (2003). Turnover at the Top: Demographic Diversity as a Determinant of Executive Turnover in The Netherlands, European Management Journal, 21(5), 614-625
Hodson, R. & Sullivan, T.A. (1985). Totem or tyrant? Monopoly, regional, and local sector effects on worker commitment, Social Forces, 63(3), 716-731
Hoffman, C. & Hurst, N (1990). Gender stereotypes: Perception or rationalization?, Journal of personality and social psychology, 58(2), 197
Hogg, M.A. (2006). Social identity theory, Contemporary social psychological theories, 13, 111-1369
Hogg, M.A. & Turner, J.C. (1985). Interpersonal attraction, social identification and psychological group formation, European Journal of Social Psychology, 15(1), 51-66
Jackson, S.E., Brett, J.F., Sessa, V.I., Cooper, D.M., Julin, J.A. & Peyronnin, K. (1991). Some differences make a difference: Individual dissimilarity and group heterogeneity as correlates of recruitment, promotions, and turnover, Journal of applied psychology, 76(5), 675
Jehn, K.A., Northcraft, G.B. & Neale, M.A. (1999). Why differences make a difference: A field study of diversity, conflict and performance in workgroups, Administrative science quarterly, 44(4), 741-763
Jeong, H. & Baek, Y. (2016). The Effect of Diversity Sensitive Orientation on Job Satisfaction and Turnover Intention, World Academy of Science, Engineering and Technology, International Journal of Social, Behavioral, Educational, Economic, Business and Industrial Engineering, 10(1), 387-390
Kontoghiorghes, C. and Bryant, N. (2004). Exploring employee commitment in a service organization in the health care insurance industry, Organization Development Journal, 22(3), 59
Lee, C. & Farh, J.-L. (2004). Joint effects of group efficacy and gender diversity on group cohesion and performance, Applied Psychology, 53(1), 136-154
