Not leaving your unsatisfactory job: analyzing female, migrant, elderly and lower-educated employees

In recent years, research on why employees leave organizations peaked (Hom et al., 2017). Studying turnover behavior gained popularity since it was associated with negative implications for organizations, such as high replacement costs (O'Connell and Kung, 2007) and inferior service quality (Hausknecht et al., 2009). Yet, research showed that turnover can also be positive for individual employees, as job switches are strong predictors of career progress and wage growth (Cheramie et al., 2007; Stumpf, 2014). Likewise, not being able to switch jobs when being dissatisfied is harmful and constitutes unhealthy retention, as it increases the risk of reduced productivity (Hom et al., 2012), workplace deviance (Sheridan et al., 2019) and lower well-being (Stengård et al., 2016). Although establishing a wide range of antecedents (Fakunmoju et al., 2010; Griffeth et al., 2000), previous literature rarely addressed differences in turnover trajectories for employees who face challenges in their careers. With widespread acknowledgment of career challenges for female, migrant, elderly and lower-educated employees – e.g. wage gaps, lack of promotion opportunities and labor market discrimination (see, e.g. Ballenger, 2010; Blommaert et al., 2012; Kahanec and Zaiceva, 2009; Kunze, 2018; Moore, 2009; Yap and Konrad, 2009) – it appears as if not everyone is equally able and likely to switch jobs. Consequently, we study whether migrant, female, elderly and lower-educated employees are less inclined to act on job dissatisfaction and likely to remain stuck in unsatisfactory jobs.

We specifically focus on job searches as an indicator of turnover, as this allows us to study whether employees facing challenges in their careers refrain from engaging in turnover processes altogether. Switching jobs often consists of a sequential process in which employees move through several intermediate steps – e.g. considerations on quitting jobs and on whether to search for another job (Griffeth et al., 2000; Hartog et al., 1988; Mobley, 1977). Following a subjective evaluation of utility (SEU), employees assess the likelihood of success before embarking on job hunts (Mobley, 1977). Searching for a different job is typically the first stage after employees become dissatisfied with their jobs and marks the onset of turnover processes. Consequently, as we argue that limited career perspectives might negatively affect the success rate of job searches, this stage captures the earliest moment at which employees facing challenges in their careers can be discouraged to translate job dissatisfaction into job switch processes.

Moreover, for some groups of employees with challenges in their careers, existing knowledge poses a conundrum. For example, migrant and female employees are incidentally reported to search for jobs more often as result of experiencing higher levels of workplace exclusion and lacking promotion opportunities (Downes et al., 2014; Hofhuis, van der Zee and Otten, 2014; Keith and Williams, 2002; Mckay et al., 2007; van Hooft et al., 2004). Accordingly, employees with career challenges may be more often dissatisfied with their employment, which could explain observations of elevated job search rates. Focusing on differences in job search behavior when groups of employees have similar degrees of dissatisfaction overcomes this issue and allows us to study whether employees with career challenges still face obstacles in their job search trajectories despite on average searching more often for different jobs.

Furthermore, in a first attempt to discover which mechanisms explain that employees with limited career perspectives are discouraged in searching for different jobs, we scrutinize how they respond to their perceptions of labor market alternatives. Although challenges on the demand side of the labor market are well-documented -e.g. the existence of labor market discrimination by employers- (see, e.g. Bertrand and Mullainathan, 2004; Blommaert et al., 2012; Hersch, 2007; Moore, 2009), little research investigates responses and job searches of employees themselves (van Hooft et al., 2004; Pager, 2007). Until now, sociological contributions studying job searches centered around job search methods and strategies (see, e.g. Pager and Pedulla, 2015; Mau and Kopischke, 2001; Weber and Mahringer, 2008) -e.g. by showing differences between formal, informal, online and offline search methods or showing that migrant employees are likely to include lower standard jobs in their searches-. We extend research on responses to limited labor market alternatives by studying whether perceptions of labor market alternatives discourage employees with career challenges to translate job dissatisfaction into job searches.

We use data from the work context of employees in the Dutch Public Sector (WORKresearch 2019). This cross-sectional data were gathered by Statistics Netherlands (CBS) and commissioned by the Ministry of the Interior and Kingdom Relations of the Netherlands. With many types of functions represented in the Dutch public sector, this dataset allows us to study the full range of the labor market and to make several contributions to the literature. First, whereas previous research largely neglects turnover differences between groups of employees, our study assesses whether employees with career challenges are more likely to remain in unsatisfactory positions as result of hampered job mobility. Second, existing literature focuses largely on turnover in the private sector and in the USA (Rubenstein et al., 2018). As labor market inequalities continue to exist in the Netherlands (see, e.g. Blommaert et al., 2012), it is important to examine in the Dutch context to what extent employees are discouraged by their perceptions of labor market alternatives. Third, our study contributes to the vast body of research on turnover processes, as analyzing to what extent a widely documented predictor of turnover behavior -i.e. job dissatisfaction-might work differently for varying groups of employees adds to knowledge on differences in turnover trajectories.

Throughout the turnover literature, it is widely documented that job dissatisfaction predicts why employees search for different jobs and leave jobs and/or organizations (see, e.g. Delfgaauw, 2007; Hom et al., 2017; March and Simon, 1958; Li et al., 2016; Tett and Meyer, 1993). In this study, we expect to reproduce these findings before assessing differences between employees.

To understand whether employees facing challenges in their careers are discouraged to translate job dissatisfaction into job search behavior, we rely on the principles of SEU (Mobley, 1977). Mobley (1977) argued that employees will only engage in job search behavior when they evaluate the utility of their efforts to be fruitful. This means that before embarking on job hunts, employees “estimate (…) the chances of finding an alternative to working in the present job (…) and the costs of search” (Mobley, 1977, p. 237). In the sections below, we argue that several mechanisms affecting employees' career outcomes and career perspectives can also reduce the utility of searching for different jobs, as they (1) increase the costs of searching for different jobs and (2) reduce the likelihood of successfully finding alternative employment. While a wide range of different mechanisms could negatively influence career perspectives and job search utility, we first aim to establish that employees facing challenges in their careers are indeed less likely to translate job dissatisfaction into job searches. Then, additionally, we move to underlying mechanisms by investigating perceived labor market alternatives as concrete indicator of limited career perspectives that may reduce the success rate of job searches and could therefore account for why they are less likely to act on job dissatisfaction.

For female employees, a wide array of research shows they continue to face obstacles in career development. Among other things, they earn less than their male counterparts for similar work -i.e. gender wage gaps- (Blau and Kahn, 2017; Kunze, 2018), have fewer promotion opportunities – i.e. glass ceilings- (Ballenger, 2010; Chisholm-Burns et al., 2017) and face further reduction of wages when becoming mothers – i.e. motherhood penalties- (Correll et al., 2007; Musick et al., 2020). While not being exhaustive, several underlying mechanisms are extensively documented. First, human capital theory suggest that career disadvantages of women are due to periods of maternity leave (Cukrowska-Torzewska and Matysiak, 2020) and them working part-time (Bardasi and Gornick, 2008; Manning and Petrongolo, 2008). Second, women are reported to have smaller professional networks, which limit their opportunities for promotions and management positions (Timberlake, 2005; von Essen and Smith, 2023). Third, research shows that traditional gender roles push them to do the lion share of household labor (Birch et al., 2009; Lachance-Grzela and Bouchard, 2010) and to take care of children and family members (Bianchi et al., 2006; Bracke et al., 2008; Hooyman, 1990). We argue that limited career perspectives and family obligations simultaneously reduce female employees' utility and increase the costs to search for jobs. Moreover, with fewer network contacts and lower human capital, the likelihood of successfully finding another job is lower than for men. Given that returns on job searches are likely to be lower for female employees, we expect the limited utility of job searches to reduce their propensity to act on job dissatisfaction. Altogether, we formulate:

Migrant employees also continue to face limited career development, with – among other things-reports of them being less likely to have employment security and more often being in temporary employment (Dekker, 2017), lower wages (Kahanec and Zaiceva, 2009), lower quality jobs (Zwysen and Demireva, 2020) and limited promotion opportunities (Greenhaus et al., 1990; Yap and Konrad, 2009). Similar to female employees, both human capital and social capital mechanisms are often presented as explanations. First, migrants' human capital can be a mismatch with the required human capital in destination countries, as work experience and education obtained in foreign countries might be less transferable to destination countries and the education itself might be of lower quality (Chiswick and Miller, 2010; Li and Sweetman, 2014). Migrants are also reported to have fewer host-country specific skills, such as knowledge on the labor market and language skills (see e.g. Roshid and Chowdhury, 2013). Second, migrant employees are likely to have less social capital, as they likely have smaller professional networks (Li et al., 2008; Verhaeghe et al., 2015). Moreover, given strong ethnic segregation in social networks (DiPrete et al., 2011; Van Tubergen, 2014), migrant networks are frequently dominated by co-ethnic relations and therefore likely to contain fewer resources. Again, we expect these mechanisms to reduce job search utility as well, as migrant employees' limited human and social capital reduces the likelihood of finding another job and thus reduces their propensity to act on job dissatisfaction. We formulate:

For elderly employees, career stage theory could explain why they are likely to evaluate job searches with lower utility. Typically, employees move through four career stages in their working life (Griffin et al., 2014; Super, 1957), which include variation in peoples' attitudes towards employment (Super et al., 1981). In the first three stages employees find their way in their careers (Griffin et al., 2014), where they are mostly occupied with joining organizations, determining career paths, acquiring skills and landing promotions. In the late career stage, however, people have established careers, stable employment, longer tenure and are fully integrated in the organization. Elderly employees in late career stages are unlikely to sacrifice these positions, becomes less concerned with career development, more likely to slowly move towards retirement and more likely to shift attention towards family relations (Cohen, 1991). As such, we expect elderly employees to evaluate job searches with higher costs and lower utility than younger employees. Furthermore, human capital theory could also explain why elderly employees evaluate job searches with lower utility. Recent research shows a clear digital divide (van Deursen and van Dijk, 2011; Van Deursen and Van Dijk, 2014), with elderly employees being likely to have less digital and technological skills than younger employees. With employment being digitalized (Hildebrandt et al., 2019), elderly employees might be less able to navigate through digitalized job search processes, which further reduces its utility for elderly employees. Altogether, we expect elderly employees to evaluate the utility of job searches to be lower than younger employees, which reduces their propensity to act on job dissatisfaction. We formulate:

For lower-educated employees, both human capital theory and mechanisms of career orientation could also account for lower utility in job searches. Human capital theory suggests that higher education increases skills and productivity of individuals more than lower education (Becker, 1994; Pericles Rospigliosi et al., 2014) and that graduating from higher education serves as threshold to enter the graduate labor market (Anderson and Tomlinson, 2021). As “graduate jobs” are typically characterized as well paid, rewarding and higher quality employment (Tomlinson, 2012) and with lower-educated employees largely being excluded from the prospect of getting into these jobs, their likelihood of successfully finding a better job than current positions -and therefore the utility of searching-is reduced. Moreover, mechanisms of career orientation could also explain differences in the evaluation of job search utility. Over the past decades, higher-educated employees in developed countries more often delayed parenthood, whereas lower-educated individuals continue to become parents at earlier ages (Dion, 1995; Sobotka, 2010; Sobotka and Beaujouan, 2018). This results in differences in career orientations, as postponed parenthood is related to smaller family size (Leridon and Slama, 2008) and increased risks of childlessness (Te Velde et al., 2012). For lower-educated employees, engagement in job searches is thus more likely to interfere with family life. Likewise, postponed parenthood for higher educated and career minded individuals reduces work-family competition (Martin, 2000), which makes it easier for higher-educated employees to achieve career goals and move into satisfactory, stable and well-paid jobs. We formulate:

Whereas several mechanisms could hamper transitions from job dissatisfaction to job search for employees facing challenges in their careers, we spotlight the degree to which this is due to a lack of perceived alternatives in the labor market. For female, migrant and elderly employees, a perception of limited alternatives could result from the prevalence of taste-based discrimination and statistical discrimination in hiring decisions (see, e.g. Blommaert et al., 2012; Heath et al., 2008; Hersch, 2007; Moore, 2009). Taste-based discrimination originates in the preference of in-group members to not hire members of certain out-groups (Becker, 1971; Guryan and Charles, 2013). Both prejudice and similarity attraction mechanisms, where people have subtle preferences for those like themselves (McPherson et al., 2001), make employers favor in-group candidates. Since management positions continue to be largely for male, majority and middle-aged employees (Mor Barak, 2016), managers–or gatekeepers–are more likely to hire candidates like themselves. Moreover, statistical discrimination occurs when employers, in absence of comprehensive applicant information, perceive the suitability of candidates on group stereotypes (Phelps, 1972). Since negative stereotypes -like negative performance bias and perceived inflexibility–exist for female (González et al., 2019; Heilman, 2012), migrant (Midtbøen, 2014; Zschirnt and Ruedin, 2016) and elderly (Karpinska et al., 2013; Posthuma and Campion, 2009) employees, their chances of finding jobs are reduced. Additionally, as the labor market is segmented, job opportunities are the “function of workers' educational credentials, preferences and skills” (Johnson and Mortimer, 2002, p. 48). Logically, having fewer educational credentials limits alternatives of lower-educated employees and thus could also affect their perceptions of labor market positions.

In considering whether they should engage in job search behavior, employees are likely to include perceptions of their labor market position. With employers being less likely to hire employees facing challenges in their careers, job search efforts are less likely to result in successfully landing alternative employment. Additionally, when they do engage in job searches, the costs might be higher as it would also take longer to obtain alternative employment. Consequently, we expect that when employees facing challenges in their careers perceive their labor market alternatives to be limited, their evaluation of utility of job searches is also more negative, which makes them less likely to act on job dissatisfaction.

We use data from the WORKresearch 2019, commissioned by the Dutch Ministry of the Interior and Kingdom Relations and collected by Statistics Netherlands (CBS). This survey has the primary goal of studying work experiences in the Dutch public sector and contains information on employability, job mobility, organizational culture, integrity, work satisfaction, technological development and leadership. The target population of the survey is all employees working in the public sector, consisting of national government, local government, education and “independent governing bodies” – e.g. Statistics Netherlands. The only exceptions being the Police and the Ministry of Defense, which were not included in the data collection. The sample is stratified by subsector, realizing representative subsamples for thirteen subsectors of the Dutch public sector, such as “national government”, “local government”, “provinces”, “judicial bodies” and “educational institutions”. Additionally, a subset of private sector respondents was included for benchmarking purposes.

In total 94,965 public servants were approached via a written letter to participate in an online survey. After both three and six weeks, reminder letters were sent to non-responding employees. As incentive respondents could win an iPad, of which eighteen were issued. Most information was gathered through the online survey, but the survey data was enriched with CBS register data. The register data contains information on demographic characteristics, marital status, household composition, income and the occupation of employees. After the six-week period, a total of 39,640 respondents participated in the study. This resulted in a response rate of 41.7%, which is acceptable given that response in Dutch surveys is usually lower than in other countries (De Leeuw and de Heer, 2001). As the survey was relatively long, CBS collected the data using a split-run method. One subset of questions was asked to all respondents and two subsets of questions were asked to different groups of respondents.

Only a subset of the total sample qualifies for our study. First, questions were asked to incoming, sitting and departing personnel. We focus on the incoming and sitting personnel only, as departed personnel clearly passed the job search stage already. Incoming personnel is included, as this concerns employees that arrived in the year before 2019 and thus had sufficient time to develop job attitudes and to potentially search for new jobs already. With this step, we excluded 2,984 respondents. Second, none of the private sector employees answered our questions of interest by design, so we excluded 2,189 respondents. Third, we excluded respondents with missing values on our dependent variable. With this step, we excluded an additional 3,758 respondents. Our final N is 30,709 employees.

We measure job search by asking respondents whether they are currently looking for another job. Frequently, job search is included as an item in scale variables such as turnover intentions, as for example with the scales of Meyer et al. (1993) and of Walsh et al. (1985). We focus on the job search dimension only as it is the intermediate step in early-stage withdrawal processes where employees evaluate their utility of looking for job alternatives (Mobley, 1977). Thus, it is also the first step in which employees can be discouraged from engaging in turnover processes. Moreover, as we are specifically interested in whether employees who face challenges in their careers are willing to either search or not, -rather than an assessment of varying dimensions of job search, upon which Blau's (1993) widely used scaled is based-, we consider the use of this single item to be legitimized. In our sample, 17.1% of respondents answered “yes” and indicated to search for another job.

We measured job dissatisfaction by asking respondents how satisfied they were with their jobs when taking everything into account. Respondents could answer on a five-point Likert scale, ranging from (1) “very unsatisfied” to (5) “very satisfied”. The measurement was coded so that a higher score reflects more job dissatisfaction. Wanous et al. (1997) assessed measurement of job satisfaction using a single item and found it to be acceptable when compared to scale measurements with multiple items.

For gender, age and migrant status measurement was based on the CBS register data. For gender, we have information on respondents being either 0 “male” or 1 “female”. For migrant status, we have information on respondents without a migration background, respondents with a Western migration background and on respondents with a non-Western migration background. For age, our variable consists of ten categories with five-year intervals. The lowest category is below 25 and the highest category is 65+. For educational level, measurement was based on a survey question, where respondents were asked what their highest followed education is. Answer categories are based on the Dutch school system and were coded so that respondents were either 0 “higher educated” or 1 “lower educated”. Lower educated employees are considered to be all employees that did not follow education at either University or the University of applied sciences.

To measure the labor market alternatives of employees, scholars have consistently relied on their perceived labor market alternatives (Li et al., 2016). While Kirschenbaum and Mano-Negrin (1999) showed objective labor market alternatives to be better for predicting actual turnover, they also stated that perceptions of the labor market are more likely to drive job searches. Thus, we rely on measurements of perceived labor market alternatives, using items based on the perceived external employability scale of Rothwell and Arnold (2007). While they also measure internal employability in their scale, we use the items measuring employees' opportunities in external organizations–i.e. the labor market-. Respondents were asked to what extent they agreed with four items, measuring their view of external job opportunities, their likelihood of landing a similar or different job and their attractivity to other employers. An overview of items can be found in  Appendix 1. Answer categories ranged on a five-point Likert scale from (1) “strongly disagree” to (5) “strongly agree”, where higher scores reflect more perceived labor market alternatives. Our scale for is reliable, with a Cronbach's alpha of 0.78.

We control for some factors that might influence the willingness or necessity for job searches. First, we control for income, based on register data of Statistics Netherlands. No information on the specific income of respondents is available, so we rely on income divided in ten deciles. Second, we control for the number of paid work years. Categories consist of “1–4 years”, “5–9 years”, “10–19 years” and “20 years or more”. Third, we controlled for both tenure at the employer and tenure in the job. Both variables are continuous variables, measured in years. Fourth, again based on register data, we controlled for whether employees had temporary or permanent contracts. Fifth, we controlled for organizational commitment. Our measure of organizational commitment consists of four items based on the scale of Allen and Meyer (1990). Respondents were asked about their attachment to, value within and meaning of the organization. An overview of items can be found in  Appendix 1. Respondents answered on a five-point scale, where a higher score reflects more organizational commitment. We find support for the items to form one latent construct, with Cronbach's α = 0.88.

Additionally, we control for variables related to the household. First, marital status was measured based on register data. Our measure consists of “married”, “divorced”, “partner deceased” and “unmarried” status. Only being currently married was coded as “married”, with the rest coded as “unmarried”. Second, based on register data, we control for children in the household. As only 49 respondents had five or more children, this variable was coded as “no children”, “one child”, “two children”, “three children” and “four children or more”.

To test our hypotheses, we performed binary logistic path analyses. Logistic regression analyses give odds ratios, assessing the likelihood that respondents are currently looking for another job. A ratio of above one indicates higher odds of that an increase in the independent variable leads to an increase in job searches. An odds ratio of below one indicates that an increase in the independent variable leads to lowered odds of searching for another job. To evaluate the practical implications of odds ratios, we followed Liberman (2005) in computing probability ranges of the likelihood that employees search for another job. To provide a general representation of the probability ranges, we generated probability pairs based on the sample grand mean -i.e. the average probability of searching for another job is 0.17-. Odds smaller than one were inversed before probabilities were calculated. First, we assess the relation between job dissatisfaction and engagement in job searches. Second, we test whether female, migrant, elderly and lower-educated employees are less likely to translate job dissatisfaction into job searches. Third, based on the model for indirect moderation of Van Kollenburg and Croon (2022) [¹,²], we test whether employees facing challenges in their careers are less likely to translate job dissatisfaction into job searches as result of limited perceived alternatives in the labor market. For this model, standard errors were bootstrapped from 1,000 sampling distributions. As bootstrapping techniques are incompatible with latent variables as moderating variables, we used mean scores for our construct variables.

To account for missing values on our independent variables, we apply full information maximum likelihood (FIML) techniques. This method uses all available raw data to estimate parameters and standard errors (Allison, 2003). We can use FIML because our split-run data collection design is a planned missing design and therefore missing values are assumed to be missing completely at random (MCAR) (Little and Rhemtulla, 2013). However, a small portion of our data was not missing by design, and we found the probability of having a missing value on our measure of perceived alternatives to depend on age, with elderly employees being less likely to have a valid score. Additionally, we found that elderly migrant employees are less likely to have a score on educational level, because respondents were asked which education they followed in the Netherlands. To account for such violations of our MCAR assumption, we perform robustness checks for all our analyses in which we deal with our missing values by applying listwise deletion methods.

Descriptive statistics are given in Table 1 for the observed data; hence, differences in sample size between variables occur. Correlations between all variables can be found in Table A8 in the Supplementary Material. Since we apply numerical integration according to the Montecarlo method, general model fit statistics are unavailable. We report Akaike information criteria (AIC) to make comparisons between models. As AIC is merely useful for the comparison of the models, we are unable to make absolute claims on the goodness of fit of our models. For R², we report McKelvey and Zavoina (1975) pseudo-R squared for binary outcome variables.

Model 1 in Table 2 shows the results of the effects of all independent variables on dependent variable job search. Our Model has an AIC of 849,807.36, which we use as benchmark for our later models. Also, our model has an R² of 0.26, indicating that 26% of the variance in job searches is explained by our model. Expectations on the relation between job dissatisfaction and job search are confirmed, as employees who are more dissatisfied with their jobs are more likely to search for different jobs (OR = 2.24, p < 0.001). For every increase in job dissatisfaction, employees are up to fourteen percentage points more likely to be searching for another job. Additionally, we find differences in job search rates for several groups of employees with challenges in their careers. For migrant employees, we observe that non-Western migrant employees are more likely to search for different jobs than both employees with a western and without a migration background, with OR = 1.20 and p = 0.016. We observe no significant difference between employees with a western migration background and employees without a migration background. For female employees, we find no differences in general job searches as compared to male employees. For elderly employees, we find that age is negatively associated with job searches. For every increase in age categories -i.e. five-year increase-, the odds of searching for a different job are 0.86 times as big (p < 0.001). For lower-educated employees, we observe that they are less likely to search for different jobs than higher-educated employees, with their odds of searching for a job being 0.71 times as big (p < 0.001).

For our first hypothesis, we initially tested whether being advantaged in general -i.e. being a higher-educated male without a migration background and below 50 years old –, as compared to being part of employee groups that face more challenges in their careers, increases the likelihood of translating job dissatisfaction into job searches (not shown). This model reports an R² of 0.26 and an AIC of 899,861.72, which is slightly higher than the first model, meaning that the added interaction effects do not increase the explanatory power of the model. We do not find significant evidence for that higher-educated male employees without a migration background and below 50 years old are significantly more likely to translate job dissatisfaction into job searches than the other employees.

Model 2 in Table 2 shows the interaction effects of all groups of employees that face challenges in their careers separately. For this Model, we explain 26% of the variance in job search and report an AIC of 870,671.89, which is comparable to the other models. When we split our analyses for the four groups of employees, we do find that the effects of job dissatisfaction on job search are significantly different for all groups. For female employees, the relation between job dissatisfaction and job search is positively moderated, with OR = 1.11 and p = 0.01. Contrary to what we expected, women are up to two percentage points more likely than men to look for another job when they are dissatisfied with their job. We therefore reject H1a. This could also explain why we previously did not find evidence for an increased likelihood of translating job dissatisfaction into job searches when we bundle all four groups of employees facing challenges in their careers together in the analysis.

For the other three groups of employees with challenges in their careers, we find what we expected. For migrants, we find that employees with a Western migration background are up to three percentage points less likely to translate their job dissatisfaction into job searches than those without a migration background (OR = 0.81 and p = 0.002). Employees with non-Western migration background are up to four percentage points less likely to do so (OR = 0.76 and p = 0.002). For age, we find that for every five-year increase the likelihood to act on job dissatisfaction reduces with up to one percentage point (OR = 0.96 and p < 0.001). For education, we find that lower-educated employees are up to two percentages points less likely than higher-educated employees to translate their job dissatisfaction into job searches, with OR = 0.88 and p = 0.016. We confirm H1b, H1c and H1d.

Figure 1 presents the results after including perceived alternatives in the labor market as indirect moderation effect. An overview of all coefficients can be found in  Appendix 2. We report an R² of 0.29 and an AIC of 983,594.85 for this model. This means that adding the indirect moderation effect does not increase the explanatory value of the model. While employees with a western migration background (β = −0.02, p = 0.01), elderly employees (β = −0.20, p < 0.001) and lower-educated employees (β = −0.06, p < 0.001) do significantly perceive their alternatives in the labor market as more limited -and we also find a significant moderating effect of perceived alternatives on the labor market on the relationship between job dissatisfaction and job search (OR = 0.96, p < 0.001)-, the moderation effects for employees with challenges in their careers on the relation between job dissatisfaction and job search remain unaffected. Hence, these are not indirectly moderated through perceived alternatives in the labor market, and we reject H2.

To evaluate the robustness of our findings, we assessed whether applying FIML to our missing data affected our results. We replicated all our models with probit regression analyses and weighted least square mean and variance adjusted (WLSMV) estimators. Missing data was dealt with by listwise deletion, resulting in a final sample size of 20,719 respondents. All tables and coefficients can be found in  Appendix 3. For our first hypothesis, we found no differences in effects for the moderations of gender, age and education on the relation between job dissatisfaction and job search. Yet, we were unable to reproduce the effects for employees with a western and non-Western migration background. Conclusions based on these findings should be interpreted with caution. For our second hypothesis, we do not find an indirect moderating effect of perceived alternatives on the relation between job dissatisfaction and job searches, and our conclusions remain unaltered.

Additionally, we assessed whether the operationalization of age affected the results. When using a binary operationalization where employees in the late career stage -i.e. above 50- are considered elderly employees and all others considered non-elderly employees, we find that elderly employees are up to two percentage points less likely to act on their job dissatisfaction, with OR = 0.86, p < 0.001. The moderating effect was again not indirectly moderated by perceptions of having limited alternatives on the labor market and conclusions remained unaltered.

Differences in turnover trajectories of female, migrant, elderly and lower-educated employees as compared to male, non-migrant, young and higher-educated employees are often neglected. As these employees face challenges in their careers – e.g. wage gaps, job insecurity and limited promotion opportunities-they could simultaneously face challenges in switching jobs. We argued that they are likely to evaluate the utility of job searches to be lower than other employees, which reduces their propensity to engage in job hunts when they are dissatisfied with their jobs. Such hampered job mobility would increase the risk of them staying reluctantly in unsatisfactory positions, leading to limited career progress, workplace deviance, absenteeism and reduced productivity. We specifically focused on job search behavior, as this is the first step in turnover processes in which employees evaluate the likelihood of success of job hunts and thus the first instance in which employees with challenges in their careers can be discouraged from engaging in job switch processes. Additionally, in an attempt to discover underlying mechanisms, we tested whether this was due to their perceptions of labor market alternatives. We used data from the WORKresearch 2019 to analyze 30,709 Dutch public servants with binary logistic path analyses.

We conclude that migrant, elderly and lower-educated employees are indeed less likely to translate their job dissatisfaction into job search than majority, younger and higher-educated employees. While several mechanisms -i.e. lower human capital and social capital, work centrality and family obligations-could underlie negative evaluations of utility to search for different jobs, we find that migrant, elderly and lower-educated employees are similar in experiencing hampered job mobility. This finding is essential and contributes to existing knowledge for two reasons. First, we discovered previously neglected but persistent inequalities in job mobility. In addition to persisting labor market inequalities (see, e.g. Blommaert et al., 2012) and other career inequalities such as wage gaps and job insecurity, already disadvantaged employees in the Netherlands are simultaneously discouraged to switch jobs. This makes them more likely to stay in employment they are not satisfied with, which is associated with reduced career progress, wage growth (Cheramie et al., 2007; Stumpf, 2014), productivity (Hom et al., 2012) and overall well-being (Stengård et al., 2016). Second, this finding extends the turnover literature because group differences in turnover processes are often neglected (Lee, 2012; Peltokorpi et al., 2015). By showing that differences occur even in one of the most important antecedents of turnover, we show that we should not rely solely on overarching theories encompassing turnover antecedents for all employees (Lee, 2012). Rather, when assessing job switch processes, scholars should focus on contextual factors and differences between employees.

In contrast, women appear to be more flexible and to experience fewer obstacles in their responses to job dissatisfaction than male employees. Women more actively translate job dissatisfaction into job searches, which might be explained by traditional gendered employment patters in Dutch society. With women in the Netherlands working part time on average (van der Lippe et al., 2006), men remain more likely to be the breadwinners of the household. As men are more likely to be the most important providers for the family, they are less flexible and job search efforts come with higher costs and therefore lower utility. Women might be more able than men to act on job dissatisfaction, which makes them more likely to search for satisfactory positions. Moreover, -although still facing challenges for acquiring management positions-trends of improved labor market positions for women (see, e.g. Cipollone et al., 2014; Kiss, 2020) could explain why we find that women are more likely to act on job dissatisfaction. With increased labor market participation of women in decent quality jobs, it could be that especially employees in lower job segments -i.e. migrant and lower-educated employees-suffer from hampered job mobility.

Furthermore, we conclude that obstacles in job search behavior of migrant, elderly and lower-educated employees are not the product of their perceived alternatives on the labor market. Instead, despite perceiving their labor market alternatives as more limited, they are not discouraged in engaging in job hunts. This is in line with previous research of Pager and Pedulla (2015), who showed that minority employees adapt to unfavorable labor markets by casting a broader range of job searches than majority employees. Like this study, our findings indicate that employees that face challenges in their careers are resilient to perceiving their labor market alternatives as limited. Even though they are likely to have fewer opportunities due to discrimination in hiring decisions (e.g. Blommaert et al., 2012; Hersch, 2007; Moore, 2009), this does not discourage them in translating job dissatisfaction into job searches. These outcomes –at least partially – might also account for why some employees with challenges in their careers might report high turnover rates, while simultaneously facing limited labor market opportunities. Alternatively, an explanation could be that these groups of employees might not see the labor market as unfavorable enough to be discouraged in job searches. Despite perceiving that they have fewer opportunities, they might still be convinced that it has utility to search for different jobs.

As we were unable to detect underlying mechanisms of differences in the likelihood to translate job dissatisfaction into job searches, we urge scholars to continue investigations into underlying mechanisms and propose two starting points for future research. First, embeddedness in family relations and kinship obligations can influence job switch processes (Lee et al., 2008; Ramesh and Gelfand, 2010). The more employees are involved in taking care of their families, the less important their individual preferences become in job switch processes. Given that migrant, elderly and lower-educated employees are more dependent on family members, do more household labor and are more family oriented (see, e.g. Cohen, 1991; Nauck and Settles, 2001; Pinquart and Sörensen, 2005), it could be that family priorities drive them to accept lower standards of satisfaction. Second, as this study focuses on data from the public sector, it could be that the attractiveness of working in the public sector accounts for our findings. Public sector organizations are characterized by favorable working conditions such as job security, healthy work-life balances and permanent contracts (see, e.g. Allen, 1988; Munnell and Fraenkel, 2013; Van Raaij et al., 2002). In exchange for keeping these benefits and in combination with having lower chances of finding alternative employment, it could be that migrant, elderly and lower-educated employees accept lower standards of job satisfaction. Investigating mechanisms of both family obligations and working in the public sector could be important to uncover why employees with challenges in their careers are less likely to act on job dissatisfaction.

Regarding organizational practice, our findings have important implications. While not because of perceived labor market alternatives, employees with challenges in their careers are at risk to stay reluctantly in unsatisfactory positions, which leads to workplace deviance, reduced productivity and limited career progress (Griffeth et al., 2000; Hom et al., 2012; Sheridan et al., 2019; Stumpf, 2014). Organizations must be aware of these processes and might therefore want to enhance their focus on improving workplace conditions and target job dissatisfaction of these employees. This is important for addressing the positions of reluctant stayers and therefore for increasing the healthy retention of employees facing challenges in their careers (Mor Barak et al., 2006).

The findings and conclusions of this study ought to be interpreted considering its limitations. First, our investigation of potential underlying mechanisms for hampered relations between job dissatisfaction and job search of employees facing challenges in their careers remains narrow. We were only able to include the perceptions of labor market alternatives, which leave us to speculate over the potential explanations for hampered job mobility effects. To tackle this, future research should identify what underlies differences in engaging in job search efforts between groups of employees.

Second, our measure of limited alternatives in the labor market is based on perceptions of employees. Although subjective measures of labor market alternatives are suitable for predicting job searches, it is less suited for predicting actual turnover (Kirschenbaum and Mano-Negrin, 1999). It would have been better if we had been able to include both subjective and objective measures of labor market alternatives. Moreover, since actual turnover clearly constitutes actual retention, we suggest that future research focuses on objective measures of labor market alternatives and relates these to actual turnover of employees who face career challenges.

Third, to test our indirect moderations we relied on cross sectional data. As we essentially test a mediation mechanism for our moderation effects and mediations are processes that take place over time (Maxwell and Cole, 2007), we ideally require longitudinal data to test for mediated moderation effects. Yet, although our cross-sectional design does not allow us to make causal inferences, it does give us a first insight into how labor market alternatives are related to job search behavior of employees with career challenges.

In this study, we discovered inequalities in job mobility of employees with career challenges. Migrant, elderly and lower-educated employees are less likely to translate their job dissatisfaction into job search behavior than majority, younger and higher-educated employees. As such, these groups of employees are more likely to stay reluctantly in unsatisfactory positions, making them at risk of suffering from reduced productivity, limited career progress, increased workplace deviance and overall well-being. Additionally, we find that women are more likely to translate job dissatisfaction into job searches than men, possibly due to traditional gendered employment patterns in Dutch society. With men often remaining the most important breadwinners of the household, this could explain men's limited flexibility in job mobility. Simultaneously, employees with challenges in their careers do not seem to be discouraged by their perceptions of labor market alternatives. Whereas they do perceive their alternatives in the labor market as limited, they remain resilient to labor market circumstances when engaging in job hunts. In response to our findings, organizations must be aware that migrant, elderly and lower-educated employees might be more likely to stay in unsatisfactory positions. Organizations ought to address their job dissatisfaction to increase healthy employee retention.

The supplementary material for this article can be found online.