Time for women, money for men?
The correlation between promotion opportunities and job satisfaction for
the sexes
Luna Bakker
Universiteit van Amsterdam
10993010
Economie en Bedrijfskunde
Financiering en Organisatie
Supervisor: Ross Gardner
29-01-2018
Abstract
Since job satisfaction is subjective, not many economists have used it as the primary topic of their research. In this thesis, the correlation between the perception of promotion
opportunities and job satisfaction for men and women has been tested, since prior papers on this topic appear to be out of date. From these papers it appears that men value
promotion opportunities more than women. Using an extensive dataset from surveys taken in 2015 on European workers’ attitudes, a generalised ordered logit regression has been performed. It appears that the correlation between job satisfaction and promotion
opportunities is significant for both men and women, without a statistically significant difference between the sexes. However, no conclusions can be drawn on causal relationships, only on correlations.
Statement of originality
This document was written by Luna Bakker, 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.
1. Introduction
A lot of research has been carried out about the effect of promotions on job performance. However, job satisfaction is a relatively new topic in economics. The effect of promotions on job satisfaction has been looked at, as well as the gender difference in job satisfaction. An investigation to see whether the relationship between promotions and job satisfaction differs between women and men is a new combination. This different relationship regarding gender has never been the sole focus of investigation. For example, Clark (1997) identified the determinants of job satisfaction for women and men, but his research only briefly covered the subject of promotions. Additionally, this particular research appears to be out of date, given the overall changes in social hierarchy and gender perceptions over the past 20 years. Finally, Clark (1997) focussed his research on US workers. Workers from other cultures may different attitudes toward social hierarchy and gender than US workers (Hofstede, 2003, p. 11; Trompenaars & Hampden-Turner, 2011, para. 1).
Job satisfaction is a relevant sign, both for employees and employers. Job satisfaction makes it less likely employees will be absent or quit their job (Clark, 1997, pp. 343-344). It might even make them put in more effort, which will in turn increase firm performances (Clark, 1997, p. 344). In that case, the activities of the employees are better aligned with the wishes of the employer. Being more satisfied with the job, one’s intrinsic motivation tends to be higher and as Gneezy and Rustichini (2000, p. 802) show, intrinsic motivation leads to better performance. Their research showed that there may be a potential influence of an individual's perception of promotion opportunities on job performance and that this should be taken into account by employers. Another consideration may be the influence of an individual's perception of promotion opportunities on job performance that may differ between gender. Gneezy, Niederle and Rustichini (2003, p. 1070) seem to have provided evidence for this claim by conducting another experiment, where it appears that men improve performance in a competitive environment, whereas women do not. An additional justification for the relevance of job satisfaction is that an employee’s performance is generally difficult to measure (A. Souza-Poza & A.A. Souza-Poza 2000, p. 136). Since one’s satisfaction and performance are correlated, measuring job satisfaction could be a proper replacement for performance.
To make use of a large and extensive 2015 dataset of European workers' attitudes towards job satisfaction, the main research question of this thesis will be:
Does the perception of promotion opportunities have a different correlation to the job satisfaction of men and women in Europe in 2015?
This thesis will start by reviewing previous literature. The subject will be defined and the literature will be used to explain the variables used in this study. The hypothesis will conclude directly from that. Thirdly, the research method will be explained, after which the results will be revealed. Finally, conclusions will be drawn.
2. Literature review
Freeman (1978) was one of the first economists to write about job satisfaction. The purpose of his paper was to show how to use econometrics to investigate the subjective variable job satisfaction. An analysis of subjective variables is more complex than of common objective variables, as job satisfaction does not solely depend on the unbiased situation, but also on personal experiences and aspirations (Freeman, 1978, p. 139). The perception of promotion opportunities leads to increased diligence and consequently actual promotion might occur. In this thesis, promotion is defined as career advancement, so a higher position in the firm hierarchy. Promotion opportunities have both a direct effect on job satisfaction and an indirect effect, since potential career advancement leads to organisational commitment, which in turn positively affects job satisfaction (Quarles, 1994, p. 191). A lack of promotion opportunities is shown to decrease job satisfaction, regardless of gender (Clark, 1997, p. 364).
The above introduced the correlation between one of the independent variables, promotion opportunities, and the dependent variable, job satisfaction. In the subchapter ‘values and aspirations’, determinants of job satisfaction will be discussed, as well as their different impacts to men and women. These variables will be included as control variables in the regression and gender will be included as the other independent variable. In the following section, ‘competition’, the difference in effect of promotion opportunities to men and women will be depicted more extensively. This chapter will concluded by the formulation of a hypothesis based on the literature.
2.1 Values and aspirations
Job satisfaction generally tends to be higher among women, which Hakim (1991, p. 101) calls a paradox, since women relatively have bad jobs compared to men. She devotes this paradox to the difference in values that men and women possess (1991, p. 113). Men find wage or promotion the most important aspect of their job, whereas women think the amount of working hours and relationships with colleagues are most important (Clark, 1997, p. 351). Besides, on average, women do not prioritise their job as much as men do and therefore prefer a job that allows them to do outside-of-work activities, like leisure, childcare, etc. Jobs that fulfil these wants are easier to find, which indicates that more female workers have a job that satisfies their aspirations than men do (Hakim, 1991, p. 107). Conversely, some research has also shown that women value wage and promotions slightly more than men and men tend to care a bit more about the amount of working hours, although men and women’s values appear similar (Rowe & Snizek, 1995, p. 222). Despite this discrepancy, some values seem to have an effect on the job satisfaction, regardless of the sex. Therefore, wage, the relationship with colleagues and some measure for work-life balance will be included in the regression. Additional notice regarding the variable wage can be found below.
Another explanation for the higher job satisfaction among women, is that women were found to have more intense emotional perceptions, which tends to result in a more positive expression of job satisfaction (A. Sousa-Poza, & A.A. Sousa-Poza, 2000, p. 147). To control for a potential higher overall satisfaction with life of women, life satisfaction will be included in the regression as well. Additionally, this variable should be included, because one’s subjective well-being is determined by recent changes instead of one’s general satisfaction (Wunder, 2012, p. 752). As current changes in the respondents’ mood are captured in the control variable life satisfaction, these will likely not influence job satisfaction.
On the contrary, women’s job satisfaction, relative to that of men, has been shown to only be higher for those that are lower-educated, not working among mainly men or working in managerial positions and for women whose mother did not have a complex job (Clark, 1997, p. 360). This provides evidence for the claim that job satisfaction depends on women’s expectations of their profession and the jobs their peers have. For example, education has been shown to be negatively correlated with job satisfaction (Clark and Oswald, 1996, p. 361). Education creates more ambitious aspirations, which, if not fulfilled, reduce job satisfaction, since the gap between diligence and reality is enlarged (Clark & Oswald, 1996, p. 361). Besides, a better education improves the chance of employment at
another job and, as mentioned by Green (2010, p. 902), outside opportunities influence one’s job satisfaction. In another study it has been shown that in most countries job satisfaction is equal for men and women (A. Sousa-Poza and A.A. Sousa-Poza, 2000, p. 147). This diminishing gender gap in job satisfaction might be the result of more women being educated, which leads to more desire among women to have a salient profession. Another explanation for the correlation between education and job satisfaction is that higher educated employees have more autonomy, resulting in a greater job satisfaction (Kaufman & Fetters, 1980, p. 252). All things considered, education seems to influence job satisfaction and therefore a measure of education will be included as control variable.
Furthermore, age will be accounted for in the regression. A study conducted by Clark (1997, p. 36) indicates that general life satisfaction of older women is significantly higher than that of younger women. In his research it remains unclear whether this difference is due to the difference in age or generation. Semykina and Linz (2013, p. 614) claim that, for younger women, job satisfaction increases as promotion opportunities for women within the company grow. For older women, however, this does not seem to be the case. Nowadays women tend to see themselves equal to men, whereas older women are assumed to care less about inequality due to their more traditional mind-sets.
It has been shown that in work environments where assertiveness is allowed or even stimulated, job satisfaction is higher (Freeman, 1978, p. 140). In these settings, employees will express their discontent instead of leaving the job in order to realise their desire elsewhere. However, Card, Mas, Moretti and Saez proved quite the opposite. They set up an experiment where some of the employees at the University of California received information about their colleagues’ salaries, while others did not (2012, p. 2986). Subsequently, both groups where asked to fill in a survey about one’s job satisfaction (2012, p. 2987). Card et al. (2012, p. 2993) proved that employees who were informed about their peers’ wages appeared to be more dissatisfied than the uninformed workers. Besides, they learned that women who know they earn less than their peers would rather search for another job than demand a wage increase (2012, p. 2996). For this reason, the dissatisfied women leave their job, whereas the dissatisfied men stay. A variable that accounts for whether the respondent can express its emotions or if these should be hidden is included as control variable as well.
This experiment also provides a justification for including subjective wage rather than objective wage. Subjective wage is a relative measure, so taking into account what one feels he or she should earn according to his or her activities and responsibilities, the salary of peers and one’s previous wage. Job satisfaction is related to the change in
income rather than the absolute income (Clark, 1999, p. 197) and a change in income is compared to a change in income of peers (Ferrer-i-Carbonell, 2005, p. 1015). However, although relative income has an effect on one’s contentment, it is stressed that absolute income is significantly correlated as well (Caporale, Georgellis, Tsitsianis, & Ping Yin, 2009, p. 47). The higher the income, the more one’s life satisfaction increases from a rise in wage (Caporale et al., 2009, p. 47). Conversely, a study indicates that reference income is irrelevant, as no significant effect was found after including educational background in the definition of a peer group (Drichoutis, Nayga, & Lazaridis, 2010, pp. 483-485). Due to the controversy regarding what wage measure to include, a test will be conducted whether absolute or relative income fits the data best.
2.2 Competition
Promotion opportunities lead to competition if only one or several employees receive the promotion, as is usually the case. Competitiveness is discoursed in, as phrased by Hakim, being ‘committed’, by working late, in the weekends and overtime (1991, p. 107). According to her, this is traditionally male behaviour. Furthermore, women, on average, tend to be more averse to competitive environments than men (Niederle & Versterlund, 2007, p. 1097). In the experiment conducted by Niederle and Versterlund (2007, p. 1073), participants first performed a task under both a piece-rate scheme and a tournament scheme and subsequently had the choice under which scheme to be compensated for the next task. Although both men and women performed better during the tournament than under the piece-rate scheme, men significantly chose the tournament more frequently than women (2007, pp. 1078-1080). This difference in preferences has several explanations.
Firstly, having the chance of being promoted is a risk. As described by Croson and Gneezy (2009, p. 453), women tend to perceive risk as a threat, whereas men encounter risk as a challenge. It appears that women experience negative emotions more intensely than men do (A. Souza-Poza & A.A. Souza-Poza, 2009, p. 542). As a consequence, the already more negative attitude of women towards risk will become even more dejected. Men, on the other hand, seem to have more desire to show that they are better than their peers and might therefore value winning more than women (Sarin & Wieland, 2012, p. 157). Garratt, Weinberger and Johnson (2013, p. 811) confirmed this preference of competing by men, adding that the wish to compete depends on age and performance as well. They collected data of a highly competitive race, to which athletes were invited if they performed above a certain threshold (2013, p. 807). Athletes above the age of 40 also had
the possibility to enter a different race, one younger athletes were excluded from. All athletes were asked to choose whether to join the contest, and, if applicable, which one they preferred to join. As Garratt et al. (2013, p. 811) found, young boys nearly always enter the contest. Older men and younger women who evaluate themselves to have an above average chance of winning, take part in the competition as well. However, among the older women who were expected to perform well, only 38% joined (Garratt, Weinberger, & Johnson, 2013, p. 812). These results again indicate that men compete for pleasure, in contrast to women, who only wish to enter a competition if they deem it probable to win. They were, again, not able to show whether these differences are the result of age characteristics or that the mind-set of the new generation of women has changed.
Secondly, although research suggests that overconfidence appears among both genders, men seem to be significantly more self-assured than women (Croson & Gneezy, 2009, p. 542). Moreover, women seemed to expect themselves to be less competent than men (Gneezy, Niederle, & Rustichini, 2003, p. 1058). Solely that belief decreases performance when competing against men just as much as when women actually have a worse ability. Besides, as men appear to anticipate winning more frequently, they might enjoy competing more as well (Gneezy et al., 2003, p. 1058). Following on from this, Günther, Ekinci, Schwieren and Strobel (2010, p. 396) discuss the stereotype threat, which means that a certain group will perform worse as a consequence of a bias implying that the performance of that specific group is lower. This happens if women compete against men, but especially if women compete against men in a typically male task (Günther et al., 2010, p. 399).
All in all, men appear to enjoy competing better and they seem to rate themselves higher than women do. This might explain why men are more satisfied if promotions are within sight, or reversely explained, why women’s satisfaction does not increase that much due to promotion opportunities.
2.3 Hypothesis
The perception of promotion opportunities has shown to be an important aspect of one’s job satisfaction. It gives a motive to work towards a target and increases organisational commitment. However, career advancement does not often lead to more leisure, which has been considered an important factor in the worker’s rate of job satisfaction (Rowe & Snizek, 1995, p. 222; Clark, 1997, p. 351). As mentioned earlier, women appear to value having time off more than men. Smith, Smith and Brower (2011, p.
382) agree, but stress that men find a work-life balance important as well. Millennials, regardless of the gender, reported finding a healthy work-life balance of great value (Smith, 2010, p. 444). This indicates that the gender gap regarding this matter is closing, as valuing leisure more could lead to a weaker effect of promotion opportunities on job satisfaction. Regardless of these current changes, women are hypothesised to still value spending time with their children more than men do. Furthermore men tend to be more competitive and especially enjoy competition more than women. Women, in general, might even dislike competitive environments. If the potential promotion is at the expense of someone else’s promotion, this could result in rivalry. Hence, men will probably have a higher job satisfaction if they perceive promotion opportunities than women.
The statistical hypothesis is stated as: and
where is the coefficient of the interaction variable of promotion opportunities among male workers and is the coefficient among female workers.
3. Methodology
In this chapter the method that is used to answer the research question will be outlined. It starts with an explanation of the regression model. The relevance of the variables will be discussed in the literature review. Subsequently the dataset will be described. The method that is used to statistically test the regression model will be covered in the final section of this chapter.
3.1 Model
The regression model used to answer the research question is described below:
In the survey that is used, job satisfaction was measured on a scale from 1 to 4: 1 being very satisfied to 4 being very dissatisfied. Since job satisfaction is the dependent variable, this ordinal data will be dealt with using a probability model. The main explanatory variable is the perception of promotion opportunities. The perceptions of
promotion opportunities are measured instead of the actual promotion opportunities, because perceptions determine one’s satisfaction, since one usually does not know the exact probability of promotion. Luckily, perceptions can be asked in surveys, whereas actual opportunities are difficult or even impossible to measure. The extent to which one has prospects to career advancements were rated from one to five; 1 being the highest possible answer and 5 the lowest. Female is a dummy variable for the gender of the respondent, where 0 displays a male respondent and 1 is a female respondent.
Furthermore, a measure of wage will be included as control variable. Due to the controversy regarding whether absolute or subjective wage is more influential to job satisfaction, a test has been performed. Adding both will lead to multicollinearity. Including solely subjective wage to the regression model leads to a higher adjusted R-squared and has therefore been chosen as independent variable. The results of the regression including subjective wage can be found in table 2, whereas the regression including absolute wage can be found in the appendix. As education is claimed to be an influential determinant for job satisfaction, this must be also be controlled for when measuring the relationship between job satisfaction and promotion opportunities. The respondent was asked to state the highest level of education or training he or she successfully completed. The answers ranged from no education to an advanced level of tertiary education. To control for the importance workers attach to working hours and a healthy work-life balance, a measure for work-life balance has been included. The respondents were asked to rate how well the working hours fit in with his or her family or social commitments outside work. The satisfaction with co-workers is measured subjectively, like promotions. The respondents’ opinion about this topic is more relevant than the objective number. The respondent was asked whether he or she feels as if he or she in general gets along well with their colleagues; 1 indicating strong agreement to 5 indicating strong disagreement. Furthermore, since Freeman (1978) and Card et al. (2012) claim that the facility of assertiveness is important, this will be controlled for in the regression. The question whether the worker’s job requires that he or she hides his or her feelings accounts for this. Responds range from always to never. Moreover, general life satisfaction might differ among men and women and must therefore be included as a control variable. In the interview, five questions were asked about the psychological well-being over the last two weeks. The average of these questions indicate the overall life satisfaction of the respondent, again 1 being the highest and 6 being the lowest. Finally, the error term captures all omitted variables. Possible variables that could cause noise are the hazards the employee is exposed to during work and the relationship with one’s supervisor. These
features are not expected to have a significantly different effect on men and women and have therefore not been included as separate variables.
Table 1: summary descriptives
Variable Obs Mean Std. Dev. Min Max ———————————————————————————————————————————————————————————————— female 35,368 .519 .500 0 1 age 35,368 45.490 57.508 15 999 balance 35,240 1.915 .746 1 4 assertiveness 34,590 3.206 1.416 1 5 satisfaction 35,226 1.939 .694 1 4 wage 35,068 2.765 1.301 1 5 promotion 33,933 3.066 1.359 1 5 colleagues 34,128 1.555 .742 1 5 abswage 2,542 6.635 3.304 1 12 education 21,137 4.789 1.517 1 7 life 35,324 2.571 1.001 1 6 3.2 Data
On the 19th of November, the data was downloaded from the UK Data Archive. The data was constructed for the European Working Conditions Survey Integrated Data File, 1991-2015. This is a survey (EWCS), distributed by the European Foundation for the Improvement of Living and Working Conditions, to assess working conditions across European countries. The research was repeated periodically, with a new survey conducted about every five years. In 1991, the first survey was completed by 12,500 workers across 12 countries. During the years, the investigation expanded, resulting in 44,000 respondents across 35 European countries in 2015. The survey was taken face-to-face. Only workers, both employees and self-employed, were invited to participate. To reflect the working population in EU countries, people of several occupations and ages, yet at least aged 15, were called. Furthermore, questions were added every new period. Hence, the data from 2015 was not only the most recent data available, but compared to previous years also appears to be the most detailed and complete dataset available.
All self-employed respondents were taken out of the dataset, as their current opinion about promotion opportunities is irrelevant and may manipulate the outcome. Therefore, the number of observations is 35,368 for the questions with the biggest response rate and 21,137 for the question used that got the least response. All questions
that were either left blank, were not answered or were reported as ‘do not know’ were adjusted as missing values to let STATA deal with it.
3.3 Data analysis
Considering that the dependent variable in this model is ordinal, using the Likert scale from one to four, multiple regression might be an inappropriate tool to analyse the data. Although commonly used for ordinal variables, there has been some controversy on this practice. On the one hand, it is claimed that multiple regression is only valid for continuous variables (Winship & Mare, 1984, p. 512). On the other hand, some assert that multiple regression can be used for non-continuous variables, provided that some specific conditions are met (Glass, Peckham, & Sanders, 1972, p. 238). For instance, the data must be normally distributed (Glass et al.,1972, p. 238). Although the number of observations of the EWCS data is large (n>30), from Cameron & Trivedi’s Information Matrix test it appears that there is skewness and kurtosis, which implies non-normality (see appendix). Accordingly, conclusions drawn from the multiple regression analysis would be unreliable and can only be used as an indicator. Therefore, multiple regression analysis is only used to refine the model equation and further analysis will be done using other methods.
For all tests and regressions done for this research, the statistical software STATA has been used. Table 2 shows the results of the multiple regression. The adjusted R-squared is higher for the model including subjective wage rather than absolute wage, which indicates that subjective wage is a better predictor for job satisfaction (Stock & Watson, 2007, p. 194). The regression results with absolute wage are included in the appendix. Age is insignificant, but since excluding age from the regression does not yield a higher adjusted R-squared, age will be maintained as independent variable.
Freeman presented probability functions, like probit and logit, as a method to deal with an ordinal dependent variable (Freeman, 1978, p. 7). However, to use an ordinal probit or logit model, the parallel regression assumption should not be violated (Long & Freese, 2006, p. 151). The Brant test implies that this assumption is indeed infringed using the EWCS data. The coefficient of education and gender is not equal for every extent of job satisfaction. Consequently, a regression analysis merging all categories of job satisfaction may distort the outcome.
Table 2: multiple regression ———————————————————————————————————— (1) ————————————————————————————————————— promotion 0.080*** (0.00) female -0.047* (0.02) promotionxfemale -0.007 (0.01) age -0.000 (0.00) wage 0.129*** (0.00) education -0.007* (0.00) balance 0.179*** (0.01) colleagues 0.093*** (0.01) assertiveness -0.024*** (0.00) life 0.131*** (0.00) constant 0.674*** (0.03) ————————————————————————————————————— R-squared 0.3364 adj. R-squared 0.3361 df 19077 ————————————————————————————————————— * p<0.05, ** p<0.01, *** p<0.001
The values in brackets represent the standard errors
Table 3: Brant test
Variable chi2 p>chi2 df ——————————————————————————————————————— All 286.20 .000 20 ——————————————————————————————————————— promotion*female 1.85 .396 2 education .04 .979 2 colleagues 46.25 .000 2 promotion 17.37 .000 2 wage 115.56 .000 2 assertiveness 11.87 .003 2 balance 14.36 .001 2 female 3.39 .183 2 age 11.19 .004 2 life 1.16 .559 2
A substitution of the ordinal probit or logit model, is the generalised ordered logit regression. This model tests the different categories of job satisfaction separately and therefore disentangles the data from the parallel regression assumption (Williams, 2006, p. 12). The highest satisfaction level will be compared to all lower levels. Subsequently, the two highest levels will be compared to all lower levels and finally the lowest level will be compared to all higher levels.
4. Results
In this chapter the results of the performed regressions will be presented. A subsequent analysis will follow. To start with, the difference in job satisfaction between men and women has been tested. Table 4 shows the proportions of the level of job satisfaction among men and women, not taking into account any other independent variable. The results of a generalised ordered logit regression are presented in the first regression of table 6. The coefficient of -0.088 implies that women have a 8.8% bigger chance than men to report to be very satisfied compared to reporting any of the other levels of satisfaction. The results of table 4 and 6 indicate that women report a high level of job satisfaction significantly more often than men do (p=0.00). However, mentions of moderate and low levels of satisfaction were not significantly more frequent among men. The coefficient of being highly and moderately satisfied compared to being highly and
moderately dissatisfied is significant with a significance level of 5% as well, but table 4 indicates that men report a moderate satisfaction relatively more often than women. Therefore it can be concluded that the coefficient of both satisfaction levels compared to both dissatisfaction levels is solely significant due to the large significance of high satisfaction. These results are in line with the expectation that women are more satisfied and express their emotions more intensely.
Table 4: proportions of gender and satisfaction
satisfaction Men Women Total ——————————————————————————————————————————————————— Very satisfied 4,017 4,616 8,633 23.66% 25.29% 24.51%
Satisfied 10,237 10,868 21,105
60.30% 59.55% 59.91% Not very satisfied 2,226 2,280 4,506 13.11% 12.49% 12.79% Not at all satisfied 496 486 982
2.92% 2.66% 2.79%
Total 16,976 18,250 35,226
100.00% 100.00% 100.00% ———————————————————————————————————————————————————
The upper numbers are absolute values; the lower numbers are relative
In table 5 the relationship between promotion and job satisfaction is depicted. Since these variables jointly satisfy the parallel regression assumption, an ordinal logit regression has been performed. The results indicate a significant, positive relationship between promotion opportunities and job satisfaction (p=0.00). Among the respondents, prospects of career advancement led to a higher reported level of job satisfaction, not taking into account any other characteristic of the respondent. Controlling for other characteristics, the relationship appears to be slightly weaker, although still significant with a significance level of 1%. The latter can be derived from the second regression of table 6. Since all independent variables are included in this regression, the parallel regression assumption is violated. Hence a generalised ordered logit regression has been executed.
Table 5: ordinal logit regression satisfaction (1) ———————————————————————————————————— promotion 0.574*** (0.01) ———————————————————————————————————— / cut1 0.448*** (0.03) cut2 3.610*** (0.04) cut3 5.566*** (0.05) ———————————————————————————————————— * p<0.05, ** p<0.01, *** p<0.001
The values in brackets represent the standard errors Cut1 differentiates ‘very satisfied’ from all lower levels of satisfaction, cut2 differentiates the two levels of satisfaction from the two levels of
dissatisfaction, cut3 differentiates ‘very dissatisfied’ from all upper levels.
Table 6: generalised ordered logit regression
(1) (2) (3) (4) ————————————————————————————————————————————————————————————————— Very satisfied female -0.088*** -0.199*** -0.126 -0.202* (0.02) (0.04) (0.09) (0.09) promotion 0.230*** 0.243*** 0.377*** (0.02) (0.02) (0.02) age 0.001 0.001 0.000 (0.00) (0.00) (0.00) wage 0.352*** 0.351*** (0.02) (0.02) education -0.022 -0.022 -0.038** (0.01) (0.01) (0.01) balance 0.743*** 0.743*** (0.03) (0.03) colleagues 0.553*** 0.553*** (0.04) (0.04) assertiveness -0.056*** -0.056*** -0.112***
(0.01) (0.01) (0.01) life 0.479*** 0.479*** 0.743*** (0.03) (0.03) (0.02) promotionxfemale -0.027 -0.039 (0.03) (0.03) constant 1.171*** -3.195*** -3.229*** -0.902*** (0.02) (0.13) (0.14) (0.11) ————————————————————————————————————————————————————————————————— Satisfied female -0.067* -0.306*** -0.277 -0.270 (0.03) (0.05) (0.15) (0.14) promotion 0.360*** 0.365*** 0.597*** (0.02) (0.03) (0.02) age -0.001 -0.001 -0.001 (0.00) (0.00) (0.00) wage 0.616*** 0.616*** (0.02) (0.02) education -0.027 -0.027 -0.043** (0.02) (0.02) (0.01) balance 0.612*** 0.612*** (0.03) (0.03) colleagues 0.302*** 0.302*** (0.03) (0.03) assertiveness -0.103*** -0.103*** -0.170*** (0.02) (0.02) (0.02) life 0.474*** 0.474*** 0.688*** (0.02) (0.02) (0.02) promotionxfemale -0.009 -0.028 (0.04) (0.04) constant -1.656*** -7.434*** -7.447*** -4.714*** (0.02) (0.17) (0.18) (0.14) —————————————————————————————————————————————————————————————————
Not very satisfied
female -0.095 -0.419*** -0.942* -0.893* (0.06) (0.09) (0.44) (0.44) promotion 0.488*** 0.441*** 0.802*** (0.05) (0.06) (0.06) age -0.004 -0.004 -0.007* (0.00) (0.00) (0.00) subwage 0.791*** 0.793*** (0.05) (0.05) education -0.026 -0.025 -0.055 (0.03) (0.03) (0.03)
balance 0.584*** 0.585*** (0.06) (0.06) colleagues 0.206*** 0.206*** (0.05) (0.05) assertiveness -0.172*** -0.171*** -0.243*** (0.03) (0.03) (0.03) life 0.435*** 0.434*** 0.681*** (0.04) (0.04) (0.04) promotionxfemale 0.118 0.098 (0.10) (0.10) constant -3.503*** -10.355*** -10.171*** -7.146*** (0.05) (0.39) (0.41) (0.35) ————————————————————————————————————————————————————————————————— * p<0.05, ** p<0.01, *** p<0.001
The values in brackets represent the standard errors
To test if the correlation between promotion opportunities and job satisfaction differs between men and women, the interaction term promotion*female was added. The same method to produce the results shown in the second regression of table 6 was used, but for the results in the third regression the interaction term was included. A positive coefficient would indicate that the relationship between promotion and job satisfaction is stronger for women, whereas the opposite is indicated with a negative coefficient. As the interaction term is insignificant for all levels of job satisfaction, it appears that there is no significant difference in correlation (p=0.37; p=0.82; p=0.23). This contradicts the hypothesis.
Two explanations for the lack of evidence of a difference will be considered in the following sections. Firstly, it can be argued that when one has the perception of promotion opportunities, he or she is more satisfied with the job and therefore reports a higher satisfaction of other job related values as well, like wage and colleagues. Since job values may differ between men and women and these values have been included as control variables, no significant difference in correlation appears, because this is already captured by the control variables. An ordinal logit regression may be used to test a difference in the importance of job values for men and women, as neither wage, colleagues, assertiveness and balance violate the parallel regression assumption. Table 7 shows the results for men and women separately. The confidence intervals enable to check if the coefficients differ significantly. All four values seem significant for both men and women, although assertiveness is the only value that does not significantly differ among both genders at a significance level of 5%. It appears that wage counts more in the determination of job
satisfaction of men than of women, whereas balance and colleagues are more important for women. This is conform to the prior literature. Moreover, the lack of difference between the gender’s correlation of promotion and satisfaction might be due to the different values of both sexes. A subsequent regression, excluding wage, balance and colleagues, has been performed to test this, of which the fourth regression of table 6 presents the results. The coefficients of the interaction term are again insignificant for all levels of job satisfaction (p=0.17; p=0.43; p=0.31). Therefore, this explanation may be invalidated.
Table 7: ordinal logit regression
satisfaction men women
————————————————————————————————————————————————————— promotion 0.286*** 0.285*** (0.02) (0.02) age 0.000 -0.000 (0.00) (0.00) subwage 0.528*** 0.449*** (0.02) (0.02) education -0.035* -0.015 (0.01) (0.02) balance 0.596*** 0.781*** (0.03) (0.03) colleagues 0.310*** 0.404*** (0.03) (0.03) assertiveness -0.071*** -0.102*** (0.02) (0.02) life 0.508*** 0.459*** (0.02) (0.02) ————————————————————————————————————————————————————— / cut1 3.060*** 3.430*** (0.13) (0.14) cut2 7.017*** 7.405*** (0.15) (0.17) cut3 9.346*** 9.828*** (0.17) (0.19) ————————————————————————————————————————————————————— * p<0.05, ** p<0.01, *** p<0.001
The values in brackets represent the standard errors Cut1 differentiates ‘very satisfied’ from all lower levels of satisfaction, cut2 differentiates the two levels of satisfaction from the two levels of
dissatisfaction, cut3 differentiates ‘very dissatisfied’ from all upper levels.
Secondly, women’s values may have changed over the years. As previously mentioned by Semykina and Linz (2013), younger, as opposed to older women, appear to report a higher level of job satisfaction if promotion opportunities for women within the company grow. If young women outweigh the older women, an explanation for the lack of difference in the EWCS data could be provided. From table 4 it appears that age is irrelevant for one’s job satisfaction. To check if this result is representative for both sexes if promotion opportunities arise, a generalised ordinal logit regression including the interaction variable promotion*age has been performed for men and women separately. The results, shown in table 9 and 10, indicate that the correlation between promotion opportunities and job satisfaction does not differ across ages, with the exception of men reporting to be very satisfied and satisfied (p=0.00; p=0.04). Following the argumentation in the first paragraph, the significant coefficient of being satisfied compared to being dissatisfied is significant due to the large significance of being very satisfied. Note that the higher the number of the ordinal variables, the more dissatisfied the respondent is, whereas age follows a numerical order. Therefore the coefficients of age, promotion and the interaction term suggest that age has a negative effect on very satisfied men’s job satisfaction, whereas the perception of promotion opportunities has a positive effect. This means that a male worker who initially reported to be very satisfied, will become less satisfied as he grows older and becomes more satisfied as promotion opportunities arise. However, no conclusions on causal relationships can be drawn, only correlations have been tested. A possible explanation on the significant correlation is that young men who have good growing opportunities within the company appear more satisfied. On the contrary, it can be argued that young men who are highly satisfied have a bigger change of being promoted.
The results are inconsistent with the claim of Semykina and Linz (2013) that older women seem more satisfied than men and younger women. However, they use data from some non-western countries over 1998 to 2008, whereas the EWCS data was generated with information from European countries in 2015. There may be more gender equality in European countries compared to non-western countries. In the non-western countries, the switch in women’s mind-set may have only evolved among the younger women, while in European countries most women, regardless of their age, have a modern mentality
already. If this argument is valid, traditionally men and women differed in their attitude towards promotion opportunities, but due to current changes in gender equality, their interest in promotion may have become more uniform.
Table 8: generalised ordered logit regression
men women ————————————————————————————————————————————————————— Very satisfied promotion 0.368*** 0.240*** (0.04) (0.03) promotionxage -0.003*** -0.000 (0.00) (0.00) age 0.012*** 0.001 (0.00) (0.00) subwage 0.378*** 0.327*** (0.03) (0.02) education -0.024 -0.020 (0.02) (0.02) balance 0.665*** 0.830*** (0.04) (0.05) colleagues 0.522*** 0.576*** (0.05) (0.05) assertiveness -0.034 -0.078*** (0.02) (0.02) life 0.510*** 0.457*** (0.04) (0.03) constant -3.659*** -3.408*** (0.22) (0.19) ————————————————————————————————————————————————————— Satisfied promotion 0.405*** 0.392*** (0.04) (0.04) promotionxage -0.001* -0.000 (0.00) (0.00) age 0.003 0.001 (0.00) (0.00) subwage 0.639*** 0.589*** (0.03) (0.03) education -0.051* 0.002 (0.02) (0.02) balance 0.516*** 0.725*** (0.04) (0.05)
colleagues 0.264*** 0.344*** (0.04) (0.04) assertiveness -0.094*** -0.112*** (0.02) (0.02) life 0.508*** 0.442*** (0.03) (0.03) constant -7.375*** -8.097*** (0.24) (0.27) —————————————————————————————————————————————————————
Not very satisfied
promotion 0.427** 0.365 (0.15) (0.26) promotionxage 0.000 0.006 (0.00) (0.01) age -0.003 -0.038 (0.01) (0.03) subwage 0.833*** 0.736*** (0.07) (0.08) education -0.032 -0.017 (0.04) (0.05) balance 0.546*** 0.618*** (0.07) (0.08) colleagues 0.134* 0.287*** (0.06) (0.07) assertiveness -0.142*** -0.215*** (0.04) (0.05) life 0.430*** 0.449*** (0.06) (0.06) constant -10.163*** -9.838*** (0.69) (1.25) ————————————————————————————————————————————————————— * p<0.05, ** p<0.01, *** p<0.001
The values in brackets represent the standard errors
A supplementary result from the second regression of table 6 is that education does not appear to be of high-impact for one’s job satisfaction. The coefficient education is insignificant for all levels of job satisfaction (p=0.301; p=0.920; p=0.716).
5. Conclusion
This final chapter starts by providing an answer to the research question and accordingly assesses the hypothesis. Implications of the outcomes will be mentioned. Furthermore, in this chapter the main results of the conducted research will be presented. Finally, the validity and reliability of the paper will be discussed, which results in recommendations for prospective research.
The results of the performed generalised ordered logit regression do not provide sufficient evidence to reject the statistical null hypothesis. Therefore, the theoretical hypothesis, stating that there is a difference in correlation between job satisfaction and promotion opportunities for men and women, appears unproven for this dataset. The perception of promotion opportunities seems to have a significant, positive effect on job satisfaction, regardless of gender. From several research papers it appears that women dislike competition more than men (Niederle & Vesterlund, 2007; Croson & Gneezy, 2009). Hence, men would report a higher job satisfaction as a consequence of promotion opportunities compared to women. Considering the regression results in this paper, it could be argued that women do not consider promotion opportunities within the company as a competition and therefore their job satisfaction increases to the same extent as that of men. Furthermore, women who observe and mention some possibility of career advancement may be confident to have a considerable chance of actually being promoted. The feeling of competing against peers may accordingly not have a negative impact on job satisfaction. Men, on the other hand, are assumed to be more confident and might therefore report perceptions of promotion opportunities more often than women (Croson & Gneezy, 2009). As appears from prior research, men enjoy competition, so the feeling of competing due to promotion opportunities does not have a negative impact on their job satisfaction either. A possible explanation for the lack of difference is that the key references date from the 1990s (Hakim and Clark) and women’s attitudes may have changed since. Traditional gender roles might have faded, which may have caused an increase in the average willingness of women to work. The perception of career advancement may have become increasingly valuable for women over the years. Therefore, the importance of promotion opportunities to job satisfaction does not seem to be significantly different for men and women. The results suggest that it would be beneficial for the employer to provide his or her employees with sufficient perspective of possible career advancement. Based on research by Clark (1997) and Gneezy and
Rustichini (2000), it is assumed that job satisfaction leads to improved performance. As women appear to value promotion equally to men, employees should focus on providing women with as many opportunities for career development as men.
Hakim (1991) mentions the gender paradox, which describes why women on average report a higher level of satisfaction than men, although more women are in lower-level professions than men. The regressions from the EWCS data indicate that women indeed report to be very satisfied more frequently than men. According to previous literature, women appear to perceive emotions more intensely than men (A. Sousa-Poza, & A.A. Sousa-Poza, 2000, p. 147). Because no effect has been found on the frequent reporting of women being dissatisfied, it may only be confirmed that women perceive positive emotions more intensely than men. Furthermore, as initially proposed, a sufficient relative wage, a good relationship with colleagues and a healthy work-life balance appear to be significantly relevant to an employee’s job satisfaction. One’s education, age and the openness to be assertive do not seem to be so, contrary to what is claimed in prior literature.
This research made use of subjective variables, because those are more determining for an employee’s satisfaction than absolute values. However, including subjective variables may also lead to circular reasoning. For example, an employee might get along with his or her colleagues very well and therefore appreciate his or her job. Consequently, the employee might also highly rate other aspects of the position, such as the opportunities for career advancement. In this case, no conclusions can be drawn regarding the influence of promotion opportunities to job satisfaction; only suggestions on the correlation between the two can be made. Therefore, no statements on internal validity can be made, since this type of validity is only relevant for causal relationships (Roe & Just, 2009, p. 1266). Causal relationships can be explored, although not by using absolute values, as the subjective subjects can hardly be captured by absolute values. Alternatively, plural surveys at different moments can be a solution. If surveys can be taken before and after someone has the feeling that growing within the company is possible, it is measurable more precisely what the actual influence of promotion opportunities on job satisfaction is. However, the spurious correlation has been limited. Spurious correlation occurs if two uncorrelated variables seem correlated because they change similarly due to a common cause (Curwin, Slater, & Eadson, 2013, p. 438). From prior research it appeared that promotion is determining for an employees job satisfaction (Rowe & Snizek, 1995; Clark, 1997). Therefore it is assumed that there exists a causal relationship between
promotion opportunities and job satisfaction and accordingly the hypothesis was formulated.
Furthermore, the adjusted R-squared of the main model is quite low. Therefore the statistical conclusion validity of the research could be questioned, because only about 33% of the data can be explained by the model. Since the variables that seemed most influential to job satisfaction from literature were included in the model, it might be suggested that the literature omits variables that are important, although to a lesser extent. Possibly some prominent variables or many little influential control variables are missing in the model. For instance, the type of job the employee has might have an impact on one’s job satisfaction. The data regarding one’s profession are absent, so this dataset made it impossible to take that into account. It would be recommended to include a variable on job type in future research. However, it can be assumed that a selection bias is limited to absent. The surveys were taken by the European Foundation for the Improvement of Living and Working Conditions, which has access to nearly all European citizens. Surveys were taken face-to-face instead of online, which does not exclude citizens without a computer. Furthermore, the number of observations is about 35 thousand, which enables to draw more general conclusions about European employees. Since the dataset solely includes European citizens, the research question is focussed on European workers and the external validity only extents that far. No conclusions about non-European employees can be made. Finally, the reliability of this paper must be discussed. Surveys result in subjective answers, which is subject to ones attitude instead of reality. However, job satisfaction is always subjective and the research question is posed in terms of perception of promotion opportunities instead of actual promotional opportunities, which enlarges reliability. Additionally, by including a variable of general life satisfaction, the respondents’ attitudes that may vary daily have been neutralised, which again increases reliability.
References
Caporale, G. M., Georgellis, Y., Tsitsianis, N., & Yin, Y. P. (2009). Income and happiness across Europe: Do reference values matter?. Journal of Economic Psychology,
30(1), 42-51.
Card, D., Mas, A., Moretti, E., & Saez, E. (2012). Inequality at work: the effect of peer salaries on job satisfaction. American Economic Review, 102(6), 2981-3003. Clark, A. E. (1997). Job satisfaction and gender: why are women so happy at
work?. Labour Economics, 4(4), 341-372.
Clark, A. E. (1999). Are wages habit-forming? Evidence from micro data. Journal of
Economic Behavior & Organization, 39(2), 179-200.
Clark, A. E., & Oswald, A. J. (1996). Satisfaction and comparison income. Journal of Public
Economics, 61(3), 359-381.
Croson, R., & Gneezy, U. (2009). Gender differences in preferences. Journal of Economic
Literature, 47(2), 448-474.
Curwin, J., Slater, R., & Eadson, D. (2013). Quantitative Methods for Business Decisions (7e ed.). Andover, United Kingdom: Cengage Learning EMEA.
Drichoutis, A. C., Nayga, R. M., & Lazaridis, P. (2010). Do reference values matter? Some notes and extensions on “income and happiness across Europe”. Journal of
Economic Psychology, 31(4), 479-486.
Ferrer-i-Carbonell, A. (2005). Income and well-being: an empirical analysis of the comparison income effect. Journal of Public Economics, 89(5), 997-1019. Freeman, R. B. (1978). Job satisfaction as an economic variable. American Economic
Review, 68(2), 135-141.
Garratt, R. J., Weinberger, C., & Johnson, N. (2013). The state street mile: Age and gender differences in competition aversion in the field. Economic Inquiry,
51(1), 806-815.
Glass, G. V., Peckham, P. D., & Sanders, J. R. (1972). Consequences of failure to meet assumptions underlying the fixed effects analyses of variance and covariance.
Review of educational research, 42(3), 237-288.
Gneezy, U., & Rustichini, A. (2000). Pay enough or don’t pay at all. The Quarterly Journal
of Economics, 115(3), 791-810.
Gneezy, U., Niederle, M., & Rustichini, A. (2003). Performance in competitive environments: Gender differences. The Quarterly Journal of Economics,
118(3), 1049-1074.
Green, F. (2010). Well-being, job satisfaction and labour mobility. Labour Economics,
17(6), 897-903.
Günther, C., Ekinci, N. A., Schwieren, C., & Strobel, M. (2010). Women can’t jump? —An experiment on competitive attitudes and stereotype threat. Journal of
Economic Behavior & Organization, 75(3), 395-401.
Hakim, C. (1991). Grateful slaves and self-made women: fact and fantasy in women's work orientations. European Sociological Review, 7(2), 101-121. Hofstede, G. (2001). Culture's consequences: Comparing values, behaviors,
institutions and organizations across nations (2nd edition). Beverly Hills, CA, United
States: Sage publications.
Kaufman, D., & Fetters, M. L. (1980). Work motivation and job values among professional men and women: A new accounting. Journal of Vocational Behavior, 17(3),
251-262.
Long, J. S., & Freese, J. (2006). Regression models for categorical dependent variables
using Stata (2nd edition). College Station, TX, United States: Stata Press.
Niederle, M., & Vesterlund, L. (2007). Do women shy away from competition? Do men compete too much?. The Quarterly Journal of Economics, 122(3), 1067-1101.
Quarles, R. (1994). An examination of promotion opportunities and evaluation criteria as mechanisms for affecting internal auditor commitment, job
satisfaction and turnover intentions. Journal of Managerial Issues, 6(2), 176-194. Roe, B. E., & Just, D. R. (2009). Internal and external validity in economics research:
Tradeoffs between experiments, field experiments, natural experiments, and field data. American Journal of Agricultural Economics, 91(5), 1266-1271.
Rowe, R., & Snizek, W. E. (1995). Gender differences in work values: Perpetuating the myth. Work and Occupations, 22(2), 215-229.
Semykina, A., & Linz, S. J. (2013). Job satisfaction and perceived gender equality in advanced promotion opportunities: An empirical investigation. Kyklos,
66(4), 591-619.
Smith, K. T. (2010). Work-life balance perspectives of marketing professionals in generation Y. Services Marketing Quarterly, 31(4), 434-447.
Smith, K. T., Smith, L. M., & Brower, T. R. (2011). An examination of work-life balance perspectives of accountants. International Journal of Critical Accounting, 3(4), 367-383.
Sousa-Poza, A., & Sousa-Poza, A. A. (2000). Taking another look at the gender/ job-satisfaction paradox. Kyklos, 53(2), 135-152.
Stock, J. H., & Watson, M. W. (2007). Introduction to econometrics (3rd edition). Boston, United States: Pearson Education.
Trompenaars, F., & Hampden-Turner, C. (2011). Riding the waves of culture:
Understanding diversity in global business (3rd edition). London, United Kingdom:
Nicholas Brealey Publishing.
Wieland, A., & Sarin, R. (2012). Domain specificity of sex differences in competition.
Journal of Economic Behavior & Organization, 83(1), 151-157.
Williams, R. (2006). Generalized ordered logit/partial proportional odds models for ordinal dependent variables. Stata Journal, 6(1), 58.
Winship, C., & Mare, R. D. (1984). Regression models with ordinal variables. American
sociological review, 512-525.
Wunder, C. (2012). Does subjective well-being dynamically adjust to circumstances?. Economics Letters, 117(3), 750-752.
Appendix
Table 9: Cameron & Trivedi's decomposition of IM-test
for skewness and kurtosis
Source chi2 df p —————————————————————————————————————————————————— Heteroscedasticity 1336.36 61 0.0000 Skewness 211.30 10 0.0000 Kurtosis 115.61 1 0.0000 —————————————————————————————————————————————————— Total 1663.26 72 0.0000
Table 10: multiple regression
(1) (2) ———————————————————————————————————————————————————— promotion 0.108*** 0.080*** (0.02) (0.00) female -0.023 -0.047* (0.07) (0.02) promotionxfemale -0.017 -0.007 (0.02) (0.01) age 0.000 (0.00) abswage -0.004 (0.00) education -0.025* -0.007* (0.01) (0.00) balance 0.215*** 0.179*** (0.02) (0.01) colleagues 0.101*** 0.093*** (0.02) (0.01) assertiveness -0.030** -0.024*** (0.01) (0.00) life 0.146*** 0.131*** (0.02) (0.00) wage 0.129*** (0.00) constant 0.953*** 0.672*** (0.10) (0.03)
———————————————————————————————————————————————————— R-sqr 0.2811 0.3365 adj. R-squared 0.2758 0.3361 degrees 1367 19078 ———————————————————————————————————————————————————— * p<0.05, ** p<0.01, *** p<0.001
The values in brackets represent the standard errors
The first regression depicts the regression including absolute wage. Note that the adjusted R-squared of this regression is lower than for the regression including subjective wage (R=0.3361). The second regression describes the regression excluding age to show that the adjusted R-squared remains the same regardless of whether age is included or not.
