Job satisfaction, search and labor mobility
A paper studying the relation between job satisfaction and search and mobility behavior of Dutch healthcare employeesBSc thesis Economics Karlijn Kersten (6077390) Supervisor: Prof. H. Oosterbeek
Acknowledgements
I would like to thank Hessel Oosterbeek from the UvA for his supervision in writing this bachelor thesis. In addition, I would like to thank Marcel Spijkerman, Olivier Tanis and Aloys Kersten for their useful comments and suggestions.
Index
1. Introduction 4
2. Literature review 5
2.1 Labor mobility theory 5
2.2 Job satisfaction as a determinant of labor mobility 6 2.3 Job satisfaction as a determinant of search behavior 8
3. Hypotheses 9
3.1 Hypothesis mobility behavior 9
3.2 Hypothesis search behavior 9
4. Data and methodology 9
4.1 Data 10
4.1.1 Job satisfaction variables 10
4.1.2 Dependent variables: search and mobility 11
4.1.3 Control variables 11
4.1.4 Consequences of partial non response 12
4.2 Methodology 14
4.2.1 Model specification 14
4.2.2 Endogeneity problems 15
5. Results 16
5.1 Descriptive analysis 16
5.2 Results search behavior 17
5.3 Results mobility behavior 18
5.4 Endogeneity 21
5.5 Interpretation results 21
6. Conclusion and discussion 22
6.1 Conclusion 22 6.2 Discussion 23 References 25 Appendix I 27 Appendix II 29 Appendix III 31 Appendix IV 32 Appendix V 35
1 Introduction
This paper studies the relation between the experienced job satisfaction of Dutch healthcare
employees (e.g. employees with a care function) and their search and mobility behavior. By matching a survey, which includes different aspects of job satisfaction and the search behavior, with the employment status in the years following the survey this paper attempts to answer the following questions:
1. What is the effect of different aspects of job satisfaction on labor mobility and search behavior of care workers in the Dutch healthcare sector?
2. Is there a difference in the impact of job satisfaction on labor and search behavior?
The Dutch Ministry of Health, Welfare and Sport foresees a discrepancy between the supply and demand of employment qualified for care functions in the future (Spijkerman, 2012). The demand for employees qualified for care functions increases more than the supply for these functions. In
expectation there is a shortage of 26 thousand employees in the Dutch healthcare sector in 2015. The main reason that drives this shortage is the growing demand for healthcare by the Dutch population. An ageing population, increasing wealth, and expanding technologies contribute to this increasing demand.
According to the Dutch Ministry of Health, Welfare and Sport there are three solutions that solve the problem of excess demand (VWS, 2011). The first solution is an increase in productivity, which means that there is less need for employment given the same demand. The second solution states that employers should recruit more employment. The final solution, and the focus of this paper, is to keep a higher ratio of the existing employment working in the healthcare sector. To maintain a larger share of the existing employment employers need to understand the motives behind mobility behavior. In addition, this question is also relevant for policymakers since the Dutch Ministry of Health, Welfare and Sport is responsible for healthcare in The Netherlands.
Different studies conclude that subjective aspects of a job influence the decision to change jobs (e.g. Clark, 2001 & Kristensen and Westergaard-Nielsen, 2004). These studies show that satisfaction with different subjective aspects of a job (e.g. satisfaction with the work itself,
supervisors, and the degree of autonomy) and the overall level of job satisfaction determine labor mobility. Economists, however, do not use subjective variables very often for the reason that they ‘measure what people say’ rather than ‘measure what people do’ (Freeman, 1978). Nevertheless, neglecting subjective variables in the evaluation of an employees’ mobility behavior can give wrong results, since they contain useful information about how satisfied an employee is with his or her job.
Before an employee decides to move to another job, the employee first decides to start searching. These two decisions, the mobility and search decision, should be distinguished from each other (Van Ophem, 1991). However, since both decisions stem from dissatisfaction with the current job they are likely based on the same motives. The questions is whether overall job dissatisfaction and dissatisfaction with different aspects of a job have the same impact on search and mobility behavior. A study of Kristensen and Westergard-Nielsen (2004) concludes that this is not the case. According to them job satisfaction is a better predictor of search behavior than it is in predicting mobility behavior.
Most studies that take the relation between job satisfaction and search and mobility behavior in account, use data representing all employed persons instead of looking at a specific group of employees. In addition, no research has been done using Dutch data on all healthcare workers. Tummers et al. (2013) study the relation between satisfaction with different aspects of a job and the intention to leave the job using nurses working in The Netherlands and conclude that job satisfaction significantly influence the intention to leave. This study, however, does not study whether or not the nurses actually switch jobs and does not include other care functions. To help fill this gap, this paper analyzes the relation between job satisfaction and mobility and search behavior for employees in the Dutch healthcare sector who have a care functions (e.g. nurses and social workers).
This paper concludes that the overall job satisfaction level and satisfaction with different aspects of a job significantly influences the search decision. The mobility decision is also influenced by different satisfaction variables, although in a more restricted way.
The remaining of this paper has the following structure. Section 2 provides an overview of the literature related to labor mobility and the relation between job satisfaction and search and mobility behavior. Section 3 gives the hypotheses . Chapter 4 describes the data and methodology, which is followed by the results in section 5. Finally, section 6 gives a conclusion and discussion of the results.
2 Literature review
This chapter gives an overview of the existing literature on labor mobility and search behavior. The first part will provide an overview of the theories on labor mobility and search behavior. The second part will elaborate more on the relation between subjective measurements of job satisfaction and search behavior. The final part discusses the relation between job satisfaction and labor mobility..
2.1 Labor mobility theory
The neoclassical theory of labor assumes a labor market with perfect information. Each individual knows the characteristics and wages of all job offers. However, Cahuc and Zylberberg (2004) state
this view is too simplistic. Therefore, they use the theory of job search to evaluate the behavior of individuals who have imperfect information about the labor market. According to this theory an individual keeps searching for a job as long as he or she thinks an improved job is available. Most search theories assume only unemployed individuals who look for a job. Cahuc and Zylberberg (2004) state that an employed individual still has the option to look for a job that improves his or her
position as long as the search costs are lower than its future increase in job characteristics or earnings. Burdett (1978) shares this view since in his search model he assumes that an employee selects two wages. The first one, X, is the reservation wage and the second one, Y, is the wage the employee quits searching. As long as the current wage is lower than Y the employee continues to search for a better alternative.
The search decision, whether or not an individual starts to search, should thus be
distinguished from the mobility decision, whether or not an individual accepts another job offer (e.g. Hartog et al. (1988), and Van Ophem (1991)). When deciding to search, an employee can discover that his or her current position is superior to the alternatives available and rejects all offers (Gesthuizen & Dagevos, 2005). The other possibility is that the employee discovers that an
alternative is better than the current job and accepts this alternative. Although search and mobility are two distinct decisions, both decisions are likely based on the same motives since both actions are a consequence of dissatisfaction with the current job situation.
2.2 Job satisfaction as a determinant of labor mobility
The existing empirical research about the effect of job satisfaction on labor mobility uses different methodologies. One stream of literature uses only the overall satisfaction level to measure job satisfaction (e.g. Freeman, 1978 and lévy-Garboua et al., 2007). Other studies use besides the overall level of job satisfaction, satisfaction with different aspect of a job (e.g. Bartel, Clark, Kristensen, Westergaard-Nielsen). Examples of aspects are promotion prospects and the work itself. Both type of empirical research, however, suggest a negative relation between job satisfaction and labor mobility.
With respect to the second stream of research the following results were found. The approach taken by Clark (2001) is hedonic in the sense that he uses satisfaction with different aspects of a job that are best in predicting quits, instead of using the aspects that are reported to be most important for an employee. The seven characteristics of job satisfaction that his study takes into account are promotion prospects, total pay, relations with supervisors, job security, ability to work on their own initiative, the actual work itself, and hours of work. Without the use of control variables (e.g. age, tenure, household, gender, and so on) he concludes that all of the domains have a significant effect on mobility. The strongest correlation is with overall job satisfaction, closely
satisfaction predicts labor mobility for both males and females. Clark (2001) concludes that out of the seven satisfaction variables job security has the largest effect, while satisfaction with pay, the work itself and the use of initiative have a similar but lesser effect on labor mobility. His study finds no empirical evidence that satisfaction with supervisors and promotion prospect have a significant effect on labor mobility.
Clark (2001) uses British data for his analysis, Kristensen and Westergard-Nielsen (2004) on the other hand apply the same methodology to Danish data. Although they include slightly different subjective satisfaction variables; instead of promotion prospects, relation with supervisors and ability to use their own initiative they use work time, work environment and distance to work. In
accordance with the results of Clark (2001) they find that overall job satisfaction has a highly significant negative relation with labor mobility. With respect to different aspects of a job, they conclude that the aspect which best predicts mobility behavior is dissatisfaction with the type of work. In addition, earnings are found to be the second most important predictor of labor mobility, while job security and distance to work are found to be insignificant. These results are different than the result of Clark (2001). Differences between countries can be an explanation for this. Kristensen and Westergard-Nielsen (2004) conclude that all results also hold for their gender-, age-, and education subgroups.
Bartel (1982) was one of the first who studied the relation between satisfaction with different job characteristics and labor mobility. For her study she used young and mature employed males from the United States. She regressed wages and different aspect of a job on labor mobility. The nonwage characteristics of a job she used were whether or not a job required repetitive functions (e.g. performance of the same task), the performing of a variety of duties and control over a variety of functions, and working under stress. Her results show that the performing of repetitive functions has a negative effect on labor mobility of young males, while the other characteristics are not significant. For mature males on the other hand this study concludes that besides the performing of repetitive functions also a job involving a variety of duties and control over an activity have a negative relation with labor mobility. Working under stress was not found to have a significant effect on labor mobility for mature males.
Bockerman and Ilmakunnas (2009) also studied the relation between satisfaction with different job aspects and actual switches. They conclude that feeling neglected by the supervisor and not having promotion prospects are not significantly influencing actual switches. These results are not similar to the result of Clark (2001), since he concludes that promotion prospect and the relation with the supervisor do have an impact on labor mobility.
The study of Cottini et al. (2011) includes repetitive work, whether or not an employee can influence his or her work, if the employee is informed about changes in the workplace, and if the
employee has participated in a training during the year before. All of these variables are dummy variables, so no scaling is applied here. This study concludes that receiving help from supervisors is negatively related to labor mobility, while help from other colleagues has no significant impact. Furthermore, the results of Cottini et al. (2011) show no evidence that workplace conflicts affects labor mobility. Furthermore, this paper shows that being able to influence your working decisions decreases the probability of labor mobility. The sharing of new information about the workplace and participating in trainings were found to be not significantly influence labor mobility.
Above researches use employees covering all sectors of the labor market to study the relation between job satisfaction and labor mobility. This paper, however, is especially interested in the relation between job satisfaction and the labor mobility of employees with a healthcare function. A study by Tummers et al. (2013) uses Dutch nurses to study the effect of job satisfaction on their behavior. However, as a dependent variable they use the intention to leave rather than looking if job dissatisfaction actual leads to labor mobility. In their study they use the satisfaction with the amount of working pressure, good leadership, the degree of autonomy, career opportunities, organizational vision, and working atmosphere as indicators of job satisfaction. They conclude that all of these variables, except for satisfaction with the degree of autonomy, significantly influence the intention to leave the organization. Since they do not observe actual labor mobility, it is doubtful whether or not the relation also holds for actual switches.
2.3 Job satisfaction as a determinant of search behavior
In contract to labor mobility little researches study the relation between job satisfaction and search behavior. A study using the variables adverse working conditions (e.g. harm, hazard, physical work), feeling neglected and no promotion prospects considers the effect on on-the-job search and actual job switches (Bockerman and Ilmakunnas, 2009). The empirical evidence suggests that employees dissatisfied with their working conditions have a higher probability to search than
employees who are satisfied. In addition, employees feeling neglected and who have little promotion prospects have a higher probability of searching. Furthermore, adverse working conditions have a larger effect on job search than it has on actual switches. Although, not surprisingly on-the-job search and actual switches are positively related. In addition, Kristensen and Westergard-Nielsen (2004) come to the same conclusion and state that job search is an even better predictor of quit behavior than job dissatisfaction. This study does not show what the impact of different job satisfaction aspects is on search behavior.
3 Hypotheses
This chapter presents the hypotheses for both search and mobility behaviour.
3.1 Hypothesis mobility behaviour
According to the literature the overall job satisfaction influences the labor mobility decision of an employee (e.g. Kristensen & Westergaard-Nielsen, 2004 and Clark, 2001). Therefore, this paper expects to find the same result. The more satisfied an employee is with his or her current job the less likely it is that he or she decides to move to another job. With respect to the satisfaction with
different aspects of a job, several studies conclude that satisfaction with earnings, the degree of autonomy, and the work itself have a negative relation with labor mobility(Clark, 2001; Kristensen & Westergaard-Nielsen, 2004; and Cottini et al., 2011). This paper, therefore, expects to find the same result. While the results of Clark (2001) and Cottini et al. (2011) shows a negative relation between satisfaction with supervisors and colleagues and labor mobility, the study of Bockerman and Ilmakunnas (2009) find no significant relation. This paper expects to find a negative relation in accordance with the study of Clark (2001) and Cottini et al. (2011). Satisfaction with stress related work is found to be negatively related with labor mobility, and thus this paper is likely to find the same result. There is no research done about the relation between satisfaction with the amount of workload and labor mobility. This paper, however, expects to see a negative relation.
3.2 Hypothesis search behaviour
Since deciding to search is related to mobility behaviour, this paper expects that job satisfaction is negatively related to search behaviour as well. The less satisfied an employee is with its current job the more likely it is that the employee starts searching. However, since searching is a phase that comes before the change of jobs, this paper expects that job satisfaction is even more related to search behaviour than it is to mobility behaviour. Which is in accordance with the study of Kristensen and Westergard-Nielsen (2004).
4 Data and methodology
This section provides a detailed description of both the data and methodology. In the first part the data used in the analysis is described. The second part will give a description of the methodology to estimate the effect of various job satisfaction aspects on labor mobility and the search decision of employees employed in the Dutch healthcare sector.
4.1 Data
The data used in this paper are derived from two sources. The first is a survey conducted by the Dutch Ministry of Health, Welfare and Sport in may 2009. This survey includes rich information of employees working in the healthcare sector, besides characteristics of the person and his or her job the survey also includes job satisfaction and search behavior. Since this paper studies only employees who have a care function, the respondents who do not have a care function are withdrawn from the sample.This means that 8,041 respondents are relevant for the analysis. The second source is a microeconomic dataset of the Dutch Central Bureau for Statistics (CBS), which contains information about the employment status of the whole Dutch population for the years 2008, 2009 and 2010. This second source is used to distinguish which of the 8,041 relevant respondents were mobile at the end of 2010, so one and a half year after the survey.
4.1.1 Job satisfaction variables
The survey includes different questions about how the employee experience his or her job. Multiple variables are created that reflect the employees’ satisfaction with different aspects of the job. Appendix II shows the descriptive statistics for all job satisfaction variables.
First, the survey includes questions about the overall job satisfaction and the overall
satisfaction with the organization the employee is currently employed. The variable ‘job satisfaction’ is reflecting the overall job satisfaction with a four point scale division1. This variable is made into three dummy variable, with all three very satisfied as the reference point. The variable ‘organization satisfaction’ reflects the overall satisfaction with the organization using the same five point scale division. These two variables, however, have a correlation of more than 0.6 with each other and thus cannot be used in the same regression. Therefore, the variable ‘organization satisfaction´ is not used in the analysis.
Second, the survey includes a wide range of questions with respect to the satisfaction with different aspects of the job (e.g. supervisor, working times, the work itself, and wage). If all these question are included in the analysis the number of variables increases with more than 25. In
addition, some questions are very similar in the sense that they refer to the same aspect of a job, and thus very related. Therefore, a principal component analysis (PCA) is applied. PCA is a variable
reduction method whereby different variables are grouped according to their correlation. The variables which highly correlate with each other are converted into the same component. Six components are distinguished. The six satisfaction components are the work itself; relation with supervisors and other colleagues; being able to influence work and working times; work related
stress; earnings; and work related pressure respectively. A number of employees did not respond to all of the questions included in the principal component analysis. To keep the number of
observations as high as possible this paper approximated the answers for employees who did not respond to four or less questions.
4.1.2 Dependent variables: search and mobility
The dataset of the CBS contains information about the employment status of the Dutch population. Appendix III shows the descriptive statistics for both the dependent variables search and labor mobility.
Every employer in The Netherlands has a number which represents the firm or institution, and thus where the employee is working at the end of each year. By matching the survey of the Dutch Ministry of Health, Welfare, and Sport with the CBS dataset the employment status of the respondents are identified. Comparing the employment status at the end of 2008, 2009 and 2010 indicates whether an employee has changed job during the period 2008-20102. The variable
‘mobility’ is used as the dependent variable in the analysis and equal to one if the employee did not change jobs, equal to two if the employee changed jobs within the same branch, equal to three if the employee changed jobs to another branch in the healthcare sector, equal to four if the employee changed jobs outside the healthcare sector, and equal to five if the employee moved to non working. Since only the identification numbers are included in the CBS dataset this paper cannot distinguish between an employee who voluntarily changed job or who involuntary changed jobs. Moreover, this paper can also not distinguish between employees who do not have a job on 31 December 2010 voluntarily or involuntarily. Which means that it is possible that employees moved to another job or out of the labor force not because of dissatisfaction but because of retirement or an ending contract.
The other dependent variable used in the analysis is the search behaviour of an employee. Employees were asked, in the survey of May 2009, whether or not they were currently searching for another job and if so the degree of search intensity (e.g. looking for vacancies, applying for a vacancy, etc.). The variable ‘search’ is equal to one if the employee is currently not looking for another job. If the employee is looking for another job the variable ‘search’ can range from two to five reflecting the search intensity (where 2 is the lowest degree of search intensity).
4.1.3 Control variables
This paper controls for relatively a lot of factors and characteristics that may influence the labor mobility and search decision by including a set of control variables. The control variables are
presented in table 4.1, appendix I shows a further explanation of the control variables and appendix IV shows the descriptive statistics for all of the control variables.
Table 4.1: Control variables
Gender Contract
Age Hours of work
Age Branch
Tenure Wage
Tenure Job connection
Education Physical heavy work
Houshould Mental heavy work
Unemployment Harassments on the job
Firm size
To control for regional unemployment a variable is included that is equal to the average unemployment percentage of the years 2009 en 2010 in the specific region.
A modification is made with respect to the variable wage. A significant number of employees did not answer the question reflecting their wage. In addition, some employees indicated to have a wage which is not possible, either because it is too high or too low. Instead of taken these
observations out of the sample a wage equation proximate the wage for these observations to keep the number of observations as high as possible. The wage equation included gender, age, the square of age, tenure, the square of tenure, branch, and education of the respondents that gave a plausible answer.
Employees are asked whether they thought their knowledge and skills are sufficient enough to meet the requirements set for the job. In addition, they were asked if they were facing problems exercising the job. The variable ‘job connection’ measures whether this is the case. Furthermore, employees are asked if they experience adverse working conditions (e.g extremely low or high temperatures, mentally, harassments of patient/clients or colleagues). For the same reasons as for the job satisfaction variables, principal component analysis reduces the number of variables to three components of adverse working conditions. The three components are: mentally heavy work; physically heavy work; and work concerning unwanted intimacies.
4.1.4 Consequences of partial non response
There are several sources of partial non response. The following section discusses these sources. First, the CBS matches the survey with the employment status according to a number that was given to each person. This number identifies each person according to gender, birth date, and address code. In total 1,686 respondents were not found in the CBS dataset, and therefore cannot be analysed since labor mobility is not identified. In order to have unbiased results these 1,686
respondents should be randomly drawn from the sample. Different tests compared the answers given in the survey of both groups (e.g. respondents that are matched and respondents that are not matched with the CBS dataset). For almost all questions the tests do not reject the H0 hypothesis (the two groups are the same). The latter is not true for the variable age, in the group that is matched the mean age is 44.08 years and in the group that is not matched the mean age is 44.77 years. From an economic perspective, however, the age difference between the two groups is not very large. Furthermore, given the large sample size a test rejects the h0 hypothesis with only small differences. For these two reason, this paper assume that the results will not be biased significantly by this source of partial non response.
Second, a number of respondents did not answered the whole survey, and thus have ‘missings’ for some of the questions. An employee is included in the analysis when he or she has no missings, with at least one missing the employee is taken out of the analysis. This means that another 1,487 respondents on the survey are excluded from the analysis. To check if these employees who are not in the analysis explain labor mobility or search behaviour this paper regressed mobility and search behaviour on whether or not an employee is in the analysis. According to the results whether or not an employee is in the analysis significantly influence labor mobility. Concerning the difference in percentage of the different directions in mobility with each group, there are indeed some small differences. The employees who are not in the analysis can thus potentially bias the results of this paper. However, for the reason that the behaviour of these employees, who are not in the analysis, is not monitored this paper cannot say anything about it. It is possible that the employees who are not in the analysis have the same behaviour as employees who are in the analysis, and the results will not be biased. With respect to search behaviour this is not significantly influenced by whether or not an employee is in the analysis. Therefore, the results will not be biased because of this.
Third, the survey took place in may 2009, the dataset of the CBS however, includes the employment status at the end of 2008, 2009 and 2010. The employment status used in the survey and in the CBS dataset are formulated differently and thus not comparable. The employment status of 31 December 2010 is therefore compared with the employment status of 21 December 2008, four months before the survey took place. Employees who started their job between 31 December 2008 and May 2009 are withdrawn from the sample, since the answers these respondents gave in the survey were not based on the job held on 31 December 2008. Since this concerns only 169 employees, this paper assumes that it does not significantly influence the results.
In summary from the original 8,041 care workers in the survey 3,341 are excluded from the analysis. This means that 4,700 respondents remain. There are no reasons to assume that non responses influences the results in this paper.
4.2 Methodology
4.2.1 Model specification
To study the effect of job satisfaction on the labor mobility decision the analysis uses a multinomial logistic regression. Since the dependent variable search included an order from low to high search intensity this paper uses an ordered logistic regression to study the relation between job satisfaction and the search decision. This method is very similar to other papers that study a comparable relation (e.g. Clark, and Kristensen & Westergard-Nielsen, ), using a logistic regression. The difference between a logistic and a multinomial logistic regression or ordered logistic regression is that in the latter two it is possible to have a dependent variable with more than two categories. A multinomial or ordered logistic regression compares the effect of each independent variable on the probability of each outcome. In this study on the probability of the different directions labor mobility and the probability on different intensities of search behaviour. The analysis controls for different
characteristics of an employee, which allows this paper to study the probability change of each job satisfaction variable on the direction of labor mobility and search behaviour.
The model of labor mobility used in this study is based on equation (1) and (2).
Pr (labor mobility= X) = α₁ + β₁ job satisfaction + δ₁ satisfaction domain i + υ₁ (1)
Pr (labor mobility= X) = α₂ + β₂ job satisfaction + δ₂ satisfaction domain i + ε₁ C + υ₂ (2)
Where the X can take on the values 1 to 5, representing the labor mobility direction. α₁ and α₂ are constants. Job satisfaction represents the effect of overall job satisfaction on the probability of the direction of labor mobility. Satisfaction domain i represents the effect of the different aspects of job satisfaction on the probability of each direction of labor mobility, where i = 1,2,3,4,5,6 (the six different job satisfaction variables). Equation (1) studies the relation between job satisfaction and labor mobility without any control variables. Equation (2) studies the same relation including all control variables specified in the previous section (see table 4.1: control variables), where C represents all control variables.
The model of search behaviour in this studies is also based on 2 equations: (3) and (4).
Pr (search = X) = α₃ + β₃ job satisfaction + δ₃ satisfaction domain i + υ₃ (3)
Where X can take on the values 1 to 5 representing the different intensities of searching in an ordered way. α₃ and α₄ are constants. The other variables have the same meaning as in equation (1) and (2). The difference between equation (3) and (4) is that the last one includes all control variables specified in the previous section, while equation (3) includes no control variables.
In general, an important assumption of the multinomial logistic regression is independence of irrelevant alternatives (Kwak and Clayton-Matthews, 2002). Which means that if a employee can choose between option A and option B, and an option C also comes available, the relative probability between option A and option B should not change. For the dependent variable labor mobility this assumption holds since there exist no extra option, all of the labor mobility directions are included in the analysis. This paper thus assumes that in theory an employee should have the opportunity to move to whatever other job he or she is capable of doing as long as he or she gives up some job characteristics (e.g. wage, status). For the analysis of the other dependent variable search this paper uses an ordered logistic regression for which the independence of irrelevant alternatives
assumptions does not have to hold.
4.2.2 Endogeneity problems
The analysis uses different variables to control for personal and other characteristics. However, it is possible that certain characteristics, not included in the analysis, have an impact on either search or mobility behaviour, and thus included in the error term, correlate with one of the independent variables. This means that it is possible that some of the independent variables are endogenous and therefore the result will be biased. A characteristic which is not included in the analysis and might be in the error term is the degree of outside opportunities for each employee (Kristensen and
Westergard-Nielsen, 2004). An employee who has less outside opportunities than others might be more satisfied than an employee who has better outside opportunities, holding everything else constant. In addition, the analysis does not include the time the employee has to travel to work. The time to travel to work, however, can have an impact on job satisfaction. Employees who have a relatively long travel time might be less satisfied than employees with a shorter travel time to work. The travel time, however, can also correlate with the search and mobility decision. Another variable that might be in the error term and correlates with job satisfaction is the productivity of an employee (Bockerman and Ilmakunnas, 2009). A more productive person is probably more satisfied than a person who is less productive. The reason for this is that in general a person is more satisfied getting more done compared to less. A productive person obtains with a higher probability another job compared to a less productive person. In general, there might be more factors that this paper does not observe which are in the error term and related to either overall job satisfaction or with one of the job satisfaction aspect variables. Given the lack of useful instruments it is not possible to
overcome the potential bias by using instrumental variables (Kristensen and Westergard-Nielsen, 2004). The study of Kristensen and Westergaard-Nielsen (2004), however, uses different
specifications of their model to analyse to what extent endogeinity is a problem. In this paper, the same procedure is followed. The use of different specifications of a model is very similar to a robustness check. This paper thus checks the robustness and endogeinity by modelling different specifications of the model.
5 Results
This chapter presents the results of the analysis. The first part will present the first results for both mobility and search behaviour.. In the second part the results with respect to search behaviour are shown. The third part describes the results for mobility behaviour. The final part gives the
interpretation of the results for both mobility behaviour and search behaviour.
5.1 Descriptive analysis
Before discussion the results of labor mobility and the search decision this paper provides some first general results. Table 5.1 includes the correlations between searching and labor mobility.
Table 5.1: Correlation search and mobility behaviour , in % Not searching Search intensity 2 Search intensity 3 Search intensity 4 Search intensity 5 Total No mobility 89.8 81.2 81.3 56.1 48.3 86.4
Mobile within the
same branch 2.3 4.9 4.8 12.3 13.0 3.2
Mobile to another
branch in healthcare 3.0 6.6 7.0 15.6 24.5 4.6
Mobile out of the
healthcare sector 2.0 4.3 4.2 8.0 9.7 2.7
Mobile out of the
workforce 2.8 3.1 2.8 8.0 4.5 3.1
Table 5.1 shows 14 percent of all employees was mobile. More than 10% of the employees who did not search is still mobile. The employees with a higher search intensity are much more mobile, up to more than 50%. The results show that there is a firm correlation between search behaviour and mobility, but this correlation is not perfect. Which indicates, in accordance with the expectations of this paper, that the mobility and search decision are not the same.
5.2 Results search behaviour
In the previous section this paper concludes that the different search intensities have a different impact on the probability to become mobile and therefore an ordered logit model is indeed the correct model. However, comparing the results of the ordered logit regression with the results of a normal logit regression model, which included the dependent variable search as a dummy variable (searching or not searching), shows very similar results. Since the results are very similar this paper prefers to use this normal logit regression, which is an simpler mode and therefore easier to interpret . Moreover, this model change does not affect the conclusions of this paper. Table 5.2 presents the results of the relation between overall job satisfaction and the satisfaction with the six different aspects of a job and whether or not an employee search. The results are a comparison with employees who do not search.
Table 5.2: Results search behaviour
Without control variables Robust standard error With control variables Robust standard error Exp. (B)
Satisfaction with the work itself -.46*** .05 -.42*** .05 .66
Satisfaction with supervisors and colleagues -.33*** .04 -.36*** .05 .70
Satisfaction with the degree of autonomy -.02 .04 -.06 .05 ---
Satisfaction with the amount of workload -.08* .04 -.13*** .05 .88
Satisfaction with earnings -.09* .04 -.08 .05 ---
Satisfaction with work related stress -.11** .04 -.05 .06 ---
Overall job satisfaction 1 (dissatisfied) +3.20*** .37 +3.37*** .38 29.08 Overall job satisfaction 2 (partly satisfied) +2.07*** .21 +2.06*** .21 7.85 Overall job satisfaction 3 (satisfied) +1.02*** .19 +.96*** .19 2.61
Number of observations 4700 --- 4700 --- ---
Wald chi2 625.76 (9) --- 734.25 (40) --- ---
Prob > chi2 0.00 --- 0.00 --- ---
Pseudo R2 0.1833 --- 0.2277 --- ---
The results without including the control variables show that all job satisfaction variables significantly influence whether or not an employee search. The only exception is satisfaction with the degree of autonomy. The negative coefficient of the satisfaction with the six different aspects of the job means that employees who are more satisfied with these aspects have a lower probability to search compared to employees who are less satisfied with the six aspects. Satisfaction with the job in general, on the other hand, has a positive sign because employees who are more satisfied with the job in general have a higher probability to search compared with employees who are very satisfied with their job. With the inclusion of the control variables the results slightly change. Satisfaction with earnings and work related stress no longer significantly influence whether or not an employee search. Furthermore, the impact of overall job satisfaction increases when employees become more dissatisfied.
With regard to the control variables, the results show that age, an employee with a contract on a call basis, an employee experiencing skills and knowledge to be higher than the job requires, and physical heavy work influence the search intensity in a significant way. The other included control variables do not influence the search decision significantly. Older employees are less likely to search relative to younger employees, which is in accordance with the literature. The other
significant control variables also influence the search intensity in a way that is expected.
In summary, the overall results show that the overall job satisfaction level and satisfaction with the work itself, supervisors and colleagues, and the amount of workload significantly influence the search decision.
5.3 Results mobility behaviour
Table 5.3 presents the results of the relation between overall job satisfaction and the satisfaction with the six different aspects of a job and mobility behaviour. The first column shows the results without control variables, the third column with all control variables described in section 4.1. All results are compared to employees who did not change jobs, the baseline result.
Table 5.3: Results labor mobility behaviour
Without control variables Robust standard errors With control variables Robust standard errors Exp (B)
Mobile within the branch
Satisfaction with the work itself -.23*** .10 -.29*** .11 .75
Satisfaction with supervisors and colleagues -.23*** .09 -.27*** .10 .76 Satisfaction with the degree of autonomy +.07 .09 +.07 .10 --- Satisfaction with the amount of workload -.07 .09 -.17* .10 .84
Table 5.3: Results labor mobility behaviour (continued) Without control variables Robust standard errors With control variables Robust standard errors Exp (B)
Satisfaction with work related stress +.13 .09 +.04 .13 --- Overall job satisfaction 1 (dissatisfied) +.53 .60 +.54 .64 --- Overall job satisfaction 2 (partly satisfied) +.15 .34 +.21 .35 ---
Overall job satisfaction 3 (satisfied) -.28 .25 -.24 .25 ---
Mobile to an other branch in healthcare
Satisfaction with the work itself -.31*** .08 -.35*** .09 .70
Satisfaction with supervisors and colleagues -.11* .08 -.18*** .08 .84 Satisfaction with the degree of autonomy -.08 .07 -.09 .08 --- Satisfaction with the amount of workload +.18*** .07 +.07 .08 ---
Satisfaction with earnings +.14** .08 +.07 .08 ---
Satisfaction with work related stress -.09 .08 -.00 .10 --- Overall job satisfaction 1 (dissatisfied) +1.12*** .43 +1.37*** .47 3.94 Overall job satisfaction 2 (partly satisfied) +.19 .29 +.28 .30 ---
Overall job satisfaction 3 (satisfied) -.06 .21 -.06 .22 ---
Mobile outside the healthcare sector
Satisfaction with the work itself -.09 .11 +.00 .12 ---
Satisfaction with supervisors and colleagues -.19** .10 -.24** .12 .79 Satisfaction with the degree of autonomy +.14 .10 -.01 .11 --- Satisfaction with the amount of workload +.04 .09 -.03 .10 ---
Satisfaction with earnings -.02 .09 -.22** .10 .80
Satisfaction with work related stress +.21** .09 +.01 .13 --- Overall job satisfaction 1 (dissatisfied) +.94 .73 +.63 .67 --- Overall job satisfaction 2 (partly satisfied) +.25 .39 +.18 .40 ---
Overall job satisfaction 3 (satisfied) +.03 .26 +.10 .27 ---
Mobile outside the labor force
Satisfaction with the work itself -.18** .10 -.14 .10 ---
Satisfaction with supervisors and colleagues -.09 .09 -.08 .10
Satisfaction with the degree of autonomy +.04 .09 +.02 .10 --- Satisfaction with the amount of workload -.07 .09 -.19* .10 .83
Satisfaction with earnings +.15* .09 +.12 .10 ---
Satisfaction with work related stress +.10 .09 -.11 .12 ---
Overall job satisfaction 1 (dissatisfied) +.93 .65 +.90 .67 --- Overall job satisfaction 2 (partly satisfied) +.71** .34 +.70** .35 2.01
Overall job satisfaction 3 (satisfied) +.02 .26 +.07 .28 ---
Number of observations 4700 --- 4700 --- --- Wald chi2 189,11 (28) --- 829.75 (164) --- --- Prob > chi2 0.0000 --- 0.0000 --- --- Pseudo R2 0.02 --- 0.15 --- ---
With respect to employees who moved to a job within the same branch, satisfaction with supervisors and other colleagues and the work itself have a significant negative effect on the probability of mobility. The higher an employee is satisfied with these aspects the lower the probability he or she moves to another job within the same branch. Satisfaction with earnings also significantly influence labor mobility in this direction. Although the coefficient is positive, which means that the more satisfied an employee is with his/her earnings the more likely this employee moves to a job within the same branch. This is not a result which is expected. When the control variables are included the results show that satisfaction with the work itself (now at a one percent significance level) and supervisors and colleagues still significantly influence an employee who moves within the same branch, and also satisfaction with the amount of workload has a significant effect. Earnings is not significant anymore.
Concerning employees who moved to a job in another branch within the healthcare sector, with the exception of satisfaction with the degree of autonomy and work related stress all
satisfaction variables significantly influence this direction of labor mobility. Furthermore, employees who are dissatisfied with their job in general have a higher probability to search compared to employees who are very satisfied. Satisfaction with the amount of workload and earnings, however, have a positive effect on the probability of mobility. Which is a result that is not in correspondence with the hypothesis. Including the control variables shows that only employees who are dissatisfied with overall job and satisfaction with the work itself, supervisors and colleagues have a significant impact on labor mobility.
Looking at employees who moved to a job outside the healthcare sector, the results show that satisfaction with supervisors and colleagues and work related stress significantly influence of labor mobility. However, satisfaction with work related stress has a positive coefficient. Again, not a result that is expected. However, this variables is no longer significant when the control variables are included in the model. In that model only satisfaction with supervisors and colleagues and earnings are significant and have the expected negative sign.
Finally, the results concerning employees who moved out of the labor force. Satisfaction with the work itself and earnings have a significant negative impact on the probability to work.
Satisfaction with earnings has a positive effect on the mobility decision, which means that employees who are more satisfied with their earnings more likely to quit working. Moreover, employees who are partly satisfied with the overall job have a higher probability to move out of the workforce compared to employees who are very satisfied. Employees who are dissatisfied, however, do not have a higher probability to move outside of the workforce. When the control variables are added to the analysis satisfaction with earnings is no longer significant and satisfaction with the amount of workload is significant.
With respect to the control variables, the results show that a temporary contract, a contract on a call basis and the amount of hours worked per week do significantly impact labor mobility. Employees who have a temporary contract or a contract on a call basis have a higher probability to change jobs, compared to employees who have a fixed contract. Which is a result that is expected. Working conditions (e.g. mentally and physical heavy work, harassments), earnings, tenure and the extent to which an employees’ knowledge and skills match the job show mixed results for the four directions of labor. Physical heavy work influences labor mobility in a positive way for employees who move to another branch within the healthcare sector. A control variable which, according to the literature, should have a negative impact on labor mobility is age. However, age in this paper does not have a significant impact on all four directions of labor mobility.
In summary, this paper concludes that especially satisfaction with the work itself and with supervisors and colleagues significantly influence labor mobility. Although, they are not significant for all four directions. The other satisfaction variables do not seem to have a significant impact on all of the four directions of labor mobility. Comparing this with the results of search behaviour shows that job satisfaction has a greater impact on searching than it has on mobility. With the inclusion of control variables the results show that, in addition to satisfaction with the work itself and supervisors and colleagues, satisfaction with the amount of workload and the overall job satisfaction significantly influence search behaviour. Whereas, with respect to mobility behaviour this is not the case.
5.4 Endogeneity
Using different specifications of both the search and mobility model outlined in the previous chapter, this paper concludes that the coefficients of the six different satisfaction aspects as well as the coefficient of the overall job satisfaction do not change significantly. The coefficients that are significant remain significant in multiple specifications of the model. Furthermore, the beta’s that do change, change only slightly. This paper therefore concludes that according to these results it seems that endogeneity is not a very big issue evaluating the results. In addition, this paper concludes that the results show that the coefficients are robust.
5.5 Interpretation of the results
In the previous section this paper discussed the results with respect to the job satisfaction variables that are significant. This section gives an interpretation of these results.
The last column of both table 5.2 and 5.3 shows the exponent of the beta’s, which represent the probability change on the mobility or search behaviour given a certain independent variable. Since this represents the actual change in the probability, the exponent of the beta’s are easier to interpreted than the beta’s itself. For example, table 5.3 shows that every unit that an employee is
more satisfied with the work itself the probability to move to another job within the same branch is smaller with a factor 0.75 compared to employees who are not mobile. Concerning, the exponent of the beta’s of search behaviour the impact of the job satisfaction variables is even larger. The
exponent of the beta of employees who are dissatisfied with the overall job is 29.08. Which means that an employee who is dissatisfied with the overall job his or her probability to search is larger with a factor 29.08 compared to employees who are very satisfied with the overall job. Overall, the impact of job satisfaction variables on search behaviour is relatively large.
The results clearly indicate that there is a difference in the impact of the job satisfaction variables on search and mobility behaviour. Some satisfaction variables (especially the work itself and relation with supervisors and colleagues) have an impact on the mobility behaviour, while relatively more satisfaction variables have an impact on search behaviour. Moreover, the results (see table 5.2 and 5.3 for the exponent of the beta’s) show that the impact of the job satisfaction variables that are significant is larger for search behaviour than it is for mobility behaviour.
A possible explanation for this difference in search and mobility behaviour is that employees who are dissatisfied with their current job can decide to start searching, and move to a better job. However, not all dissatisfied employees who search find a job that is better than the current one and therefore do not move. In addition, employees with a care function have specific human capital which they cannot use outside the healthcare sector, as a consequence their mobility possibilities are limited. Furthermore, there may be employees who are not dissatisfied but do move to another job. With respect to the results of this paper this means that it is possible that dissatisfaction with job characteristics have less impact on mobility compared to search behaviour.
The job satisfaction variables that significantly influence labor and mobility behaviour are satisfaction with the work itself and supervisors and colleagues. If employers want to maintain a larger share of their existing employees these are the aspects that they should be most concerned about. Employers should be less concerned with the earnings of employees, since the results showed that satisfaction with earnings does not have a significant impact on both labor mobility and
searching for another job.
6 Conclusion and discussion
In this final chapter the summary and conclusion of this paper are first presented. After that, the limitations of this study and suggestions for future research are covered.
6.1 Conclusion
This paper studies the relation between job satisfaction and search and mobility behaviour of employees working in the Dutch healthcare sector. Using different subjective measurements of job
satisfaction (e.g. overall satisfaction and satisfaction with the work itself, supervisors and colleagues, the amount of workload, earnings, the degree of autonomy, and work related stress) this paper analyzed the impact of satisfaction with these job characteristics on both search and mobility behaviour using a multinomial logistic regression.
This paper concludes that satisfaction with the work itself and supervisors and colleagues have a negative effect on mobility behaviour, but not for all four mobility directions. The other job satisfaction variables, in general, do not significantly influence labor mobility. Concerning search behaviour, this paper concludes that overall job satisfaction and satisfaction with the work itself, the amount of workload and supervisors and colleagues have a significant negative impact. In addition, compared to the results of mobility behaviour, most of the other satisfaction variables influence search behaviour significantly These results are not the same compared to other research studying the same relation. Especially satisfaction with earnings has a negative impact on labor mobility in existing literature. This paper, however, did not find such a relation.
Second, this paper concludes that satisfaction with different job characteristics has not the same impact on mobility and search behaviour. Search behaviour is more influenced by satisfaction with different aspects of a job than it influences labor mobility. Which is in accordance with the existing literature. In addition, since probably not all employees who search move to another job the result of this paper corresponds to this observation.
6.2 Discussion
The results of this paper show that some job characteristics have a significant impact on search and mobility behaviour. However, it is important to be aware of the limitations in this study. This section discusses the main limitations of this paper and gives suggestions for future research.
First, the results might be biased for the reason that this paper does not distinguish between employees who voluntarily or involuntarily change jobs. If employees changed job because they had a temporary contract, and not because of job dissatisfaction this leads to a bias in the results. In addition, endogeneity may also cause a bias in the results. Furthermore, the analysis showed that whether or not an employee is in the analysis influences labor mobility, which can also cause a bias in the results. When, however, these employees have the same behaviour as the employees who are in the analysis it is not a problem. Unfortunately, the behaviour of these employees is unknown.
Second, there are some problems with subjective questioning which can influence the results. According to Bertrand and Mullainathan (2001) respondents are sensitive to the order of the questions and to the way a question is stated. Another problem is the scaling effect. Respondents sometimes change their answer, when another scaling is applied. A final problem is that it is possible that the respondents does not have an opinion about the subject. Therefore, he or she gives an
answer that is not based on anything. These measurements errors might influence our results. Bertrand and Mullainathan (2001), however, concluded that subjective variables are useful as explanatory variables.
Finally, a suggestion for future research is to model search and mobility behaviour in one framework by regressing a series of joint equations. Since the search decision and the mobility decision are dependent of each other they cannot be modelled in one equation.
References
Bartel, A.P, (1982) “Wages, Nonwage Job Characteristics, and Labor Mobility”, Industrial and Labor Relations Review, 35(4): 578-589.
Bertrand, M., and Mullainathan, S., (2001) “Do People Mean What They Say? Implications for Subjective Survey Data”, The American Economic Review, 91(2): 67-72.
Bockerman, P. and Ilmakunnas, P., (2009) “Job Disamenities, Job Satisfaction, Quit Intention, and Actual Separations: Putting the Pieces Together”, Industrial Relations, 48(1): 72-95.
Burdett, K., (1978) “A Theory of Employee Job Search and Quit Rates”, The American Economic Review, 68(1): 212-220.
Cahuc, P., and Zylberberg, A., (2004) “labor Economics”, The MIT Press, Cambridge, Massachusetts.
Clark, A.E., (2001) “What really matters in a job? Hedonic measurement using quit data”, Labour Economics, 8(2): 223-242.
Cottini, E., Kato, T., and Westergaard-Nielsen, N., (2011) “Adverse workplace conditions, high-involvement work practice and labor turnover: Evidence from Danish linked employer-employee data”, Labour Economics, 18(6): 872-880.
Freeman, R.B., (1978) “Job Satisfaction as an Economic Variable”, The American Ecnomic Review, 68(2): 135-141.
Gesthuizen, M., and Dagevos, J., (2005) “Arbeidsmobiliteit in goede banen: Oorzaken en gevolgen van baan- en functiewisselingen en gevolgen voor de kenmerken van het werk”, Sociaal en Cultureel Planbureau, Den Haag.
Hartog, J., Mekkerlholt, E., and Van Ophem, H., (1988) “Testing the relevance of job search for job mobility”, Economic Letters, 27: 299-303
Jovanovic, B., (1979) “Job Matching and the Theory of Turnover”, Journal of Political Economy, 87(5): 972-990.
Kristensen, N., and Westergard-nielsen, N., (2004) “Does Low Job Satisfaction Lead to Job Mobility?”, IZA Discussion Paper No. 1026.
Kwak, C., and Clayton-Matthews, A., (2002) “Multinomial Logistic Regression”, Nursing Research, 51(6): 404-410.
Lévy-Garboua, L., Montmarquette, C., and Simonnet, V., (2007) "Job satisfaction and quits", Labour Economics, 14(2) : 251-268.
Van Ophem, H. (1991) “Wages, Nonwage Job Characteristics and the Search behaviour of Employees”, The Review of Economics and Statistics, 73(1): 145-151.
Spijkerman, M., (2012) “Arbeidsmarktprognose van VOV personeel in Zorg en Welzijn 2011-2015”, Panteia, Zoetermeer.
Tummers, L., G., Groeneveld, S., M., and Lankhaar, M. (2013) “Why do nurses inted to leave their organization? A large scale analysis in long term care”, Journal of Advanced Nursing, 2014
forthcoming.
VWS, (2011) Brief van de minister en staatsecretaris van Volksgezondheid, Welzijn en Sport aan de Tweede Kamer, TK 29282 nr. 128, Den Haag.
Appendix I
Table A1: Description control variables Control variables Definition
Gender = 1 if the worker is male, and 0 otherwise.
Age Age in years.
Age² Square of age
Ethnic234 = 1 if the employee is a Westerner immigrant, and 0 otherwise. Ethnic3 = 1 if the employee is a non Westerner immigrant, and 0 otherwise. Tenure Years of working at the current firm.
Tenure² Square of tenure.
Education25 = 1 if the employees' highest education is VMBO-TL or VMBO-GL, and 0 otherwise.
Education3 =1 if the employees’ highest education is MBO, HAVO, or VWO, and 0 otherwise.
Education4 =1 if the employees’ highest education is HBO, and 0 otherwise. Education5 =1 if the employees’ highest education is WO, and 0 othersie.
Household26 = 1 if the employee has no spouse and has children living at home under the age of 18, and 0 otherwise.
Household3 =1 if the employee has spouse and no children living at home under the age of 18, and 0 otherwise.
Household4 =1 if the employee has spouse and children living at home under the age of 18, and 0 otherwise.
Unemployment (region)7
A vector of 12 unemployment percentage (according to region) dummy variable.
Firm size The number of employees in the organization.
Contract28 = 1 if the employee has a temporary contract, and 0 otherwise. Contract3 =1 if the employee has a contract on a call basis, and 0 otherwise. Hours Hours worked as a percentage of full time hours, were full time is equal
to 36 hours per week.
Branch2910 =1 if the employee has a function in the nursery or home care, and 0 otherwise.
Bracnh3 =1 if the employee has a function in the handicapped care, and 0 otherwise.
3
If the ethnicity given in the survey is not equal to the ethnicity in the CBS dataset, the ethnicity is equal to the one given in the CBS dataset.
4 The baseline for both ethnic variables is if an employee is a Dutch native. 5
The baseline for all education variables is if an employees’ highest education is VMBO-BB or VMBO-KB. 6
The baseline for all household variables is if an employee has no spouse and no children living at home under the age of eighteen.
7 If the address code given in the survey is not equal to the address code in the CBS dataset, the address code is equal to the one that is given in the survey.
8 The baseline for the variables contract2 and contract3 is if an employee has a indefinitely contract. 9 The baseline for all seven branches is if an employee has a function in the hospital.
10 The eight branches that are distinguished in the Dutch healthcare sector are: hospital care, mental care, nursery and home care, handicapped care, youth care, social care, day care, and other care.
Branch4 =1 if the employee has a function in the mental care, and 0 otherwise. Branch5 =1 if the employee has a function in the social care, and 0 otherwise. Branch6 =1 if the employee has a function in the youth care, and 0 otherwise. Branch7 =1 if the employee has a function in the day care, and 0 otherwise. Branch8 =1 if the employee has a function in care other distinghuised, and 0
otherwise.
Wage Natural logarithm of hourly wage.
Job connection211 = 1 if the employee experience his/her knowledge and skills are lower than the requirement level, and 0 otherwise.
Jon connection3 =1 if the employee experience his/her knowledge and skills to be higher than the requirement level, and 0 otherwise.
Physical heavy work Principal component score for the employee. Mentally heavy work Principal component score for the employee. Harassments on the job Principal component score for the employee.
11
The baseline for both job connection variables is if an employee experience their knowledge and skills are equal to the requirement level of the job.
Appendix II
The tables below show the descriptive statistics for the different job satisfaction variables. The higher the principal component score the more satisfied the employee is with that specific aspect of the job.
Job satisfaction variable: satisfaction with the work itself
Range for the principal component score Number of observations
<-3 41 -3 -2 106 -2 -1 381 -1 0 1323 0 1 2136 >1 713 Total 4700
Job satisfaction variable: satisfaction with supervisors and colleagues Range for the principal component score Number of observations
<-3 25 -3 -2 159 -2 -1 587 -1 0 1261 0 1 1951 >1 717 Total 4700
Job satisfaction variable: satisfaction with the degree of autonomy
Range for the principal component score Number of observations
<-3 13 -3 -2 149 -2 -1 725 -1 0 1533 0 1 1685 1 2 582 >2 13 Total 4700
Job satisfaction variable: satisfaction with the amount of workload
Range for the principal component score Number of observations
<-3 49 -3 -2 725 -2 -1 1589 -1 0 1524 0 1 724 >1 89 Total 4700
Job satisfaction variable: satisfaction with earnings
Range for the principal component score Number of observations
<-3 89 -3 -2 801 -2 -1 1443 -1 0 1668 0 1 621 >1 78 Total 4700
Job satisfaction variable: satisfaction with work related stress
Range for the principal component score Number of observations
<-3 8 -3 -2 166 -2 -1 769 -1 0 1560 0 1 1554 1 2 612 >2 31 Total 4700
Job satisfaction variable: overall job satisfaction
Frequency Percent Valid Percent Cumulative
Percent Dissatisfied Partly satisfied Satisfied Very satisfied Total 85 810 2834 971 4700 1.8 17.2 60.3 20.7 100.0 1.8 17.2 60.3 20.7 100.0 1.8 19.0 79.3 100.0
Appendix III
The tables below show the descriptive statistics for the dependent variables search behaviour and mobility behaviour.
Dependent variable: search behaviour
Frequency Percent Valid Percent Cumulative
Percent Not searching Search intensity 2 Search intensity 3 Search intensity 4 Search intensity 5 Total 3849 287 256 153 155 4700 81.9 6.1 5.4 3.3 3.3 100.0 81.9 6.1 5.4 3.3 3.3 100.0 81.9 88.0 93.4 96.7 100.0
Dependent variable: mobility behaviour
Frequency Percent Valid Percent
Cumulative Percent Not mobile
Mobile to the same branch
Mobile to another branch within the healthcare sector
Mobile outside the healthcare sector Mobile outside the labor force Total 4060 152 216 128 144 4700 86.4 3.2 4.6 2.7 3.1 100.0 86.4 3.2 4.6 2.7 3.1 100.0 86.4 89.6 94.2 96.9 100.0
Appendix IV
The tables below show the descriptive statistics for all of the control variables used in the analysis.
Control variable gender:
Number of males 598
Number of females 4102
Control variable age:
N Minimum Maximum Mean Std. Deviation
Age
Valid N (list wise)
4700 4700
19.00 66.00 42.3715 11.29245
Control variable ethnic:
Frequency Percent Valid percent Cumulative Percent Dutch native
Westerner immigrant Non Westerner immigrant Total 4325 264 111 4700 92.0 5.6 2.4 100.0 92.0 5.6 2.4 100.0 92.0 97.6 100.0
Control variable Tenure:
N Minimum Maximum Mean Std. Deviation
Tenure
Valid N (list wise)
4700 4700
1.00 44.00 10.2609 8.78225
Control variable education:
Frequency Percent Valid Percent
Cumulative Percent Highest education is VMBO-BB or VMBO-KB
Highest education is VMBO-GL or VMBO-TL Highest education is MBO, HAVO, or VWO Highest education is HBO
Highest education is WO Total 65 478 1845 1905 407 4700 1.4 10.2 39.3 40.5 8.7 100.0 1.4 10.2 39.3 40.5 8.7 100.0 1.4 11.6 50.8 91.3 100.0
Control variable household:
Frequency Percent Valid Percent
Cumulative Percent No spouse and no children under 18 living at home
No spouse and children living at home under 18 With spouse,no children under 18 living at home With spouse and children at home under 18 Total 795 187 1810 1908 4700 16.9 4.0 38.5 40.6 100.0 16.9 4.0 38.5 40.6 100.0 16.9 20.9 59.4 100.0
Control variable unemployment rate (given as a percentage):
Frequency Percent Valid Percent Cumulative
Percent 3.75 4.4 4.9 4.95 5.3 5.35 5.65 6.00 6.15 6.65 Total 145 1020 769 563 339 832 142 492 160 238 4700 3.1 21.7 16.4 12.0 7.2 17.7 3.0 10.5 3.4 5.1 100.0 3.1 21.7 16.4 12.0 7.2 17.7 3.0 10.5 3.4 5.1 100.0 3.1 24.9 41.3 53.3 60.6 78.3 91.4 91.9 95.3 100.0
Control varibale firm size:
N Minimum Maximum Mean Std. Deviation
Firm size
Valid N (list wise)
4700 4700
1 9999 1432.37 2033.345
Control variable contract:
Frequency Percent Valid Percent Cumulative
Percent Indefinitely contract
Temporary contract Contract on a call basis Total 3959 436 305 4700 84.2 9.3 6.5 100.0 84.2 9.3 6.5 100.0 84.2 93.5 100.0
Control variable Hours:
N Minimum Maximum Mean Std. Deviation
Hours worked as a percentage of fulltime hours
Valid N (list wise)
4700
4700
0.00 1.00 0.7123 0.22984
Control variable branch:
Frequency Percent Valid Percent Cumulative Percent Hospital
Nursery and home care Handicapped care Mental care Social care Youth care Day care Other care Total 966 1193 437 449 439 507 312 397 4700 20.6 25.4 9.3 9.6 9.3 10.8 6.6 8.4 100.0 20.6 25.4 9.3 9.6 9.3 10.8 6.6 8.4 100.0 20.6 45.9 55.2 64.8 74.1 84.9 91.6 100.0
Control variable wage (given in the natural log form):
N Minimum Maximum Mean Std. Deviation
Wage
Valid N (list wise)
4700 4700
1.63 4.55 2.5234 0.28796
Control variable job connection:
Frequency Percent Valid Percent Cumulative
Percent Job connection 1 2 3 Total 2627 484 1589 4700 55.9 10.3 33.8 100.0 55.9 10.3 33.8 100.0 55.9 66.2 100.0
Control variable Physical heavy work:
N Minimum Maximum Mean Std. Deviation
Physical heavy work Valid N (list wise)
4700 4700
-1.76986 2.78425 -0.2141257 0.896133
Control variable mentally heavy work:
N Minimum Maximum Mean Std. Deviation
Mentally heavy work Valid N (list wise)
4700 4700
-3.28101 1.99659 -0.0838557 1.00431089
Control variable harassments on the job:
N Minimum Maximum Mean Std.
Deviation Harassments on the job
Valid N (list wise)
4700 4700
