1 Introduction

Although the existence of a mutual relationship between ‘objective’ characteristics related to the labour market position and ‘subjective’ attitudes and well-being is largely undisputed, in empirical studies scholars’ focus has been overwhelmingly directed towards the impact of the former on the latter (e.g., Khattab and Fenton 2009; Drobnic et al. 2010). That literature has convincingly demonstrated that employment provides more than the economic means necessary for making one’s living. Being engaged in paid work also fulfils a number of latent functions (Jahoda 1982) and contributes to subjective well-being by fostering self-esteem, a sense of belonging, status, etc. As a source of meaning (Glorieux 1995) paid work not only affects opinions with regard to the private sphere—such as self-esteem, happiness, and life satisfaction—but also contributes to the sense of taking up a meaningful position into society (Elchardus et al. 1996). This is one of the reasons why the expansion of the welfare state and social insurance do not mitigate the significance of employment for citizens’ well-being (e.g., Gallie et al. 1993).

The alternative hypothesis whereby attitudes and opinions cause labour market behaviour and career choices has attracted less attention from scholars (but see Marks and Fleming 1999; Boehm and Lyobomirsky 2008; Roberts et al. 2003; Samuel et al. 2013; Wright and Cropanzano 2000). Both on theoretical (e.g., Houtman 2003) and empirical grounds (e.g., Marks and Fleming 1999; Boehm and Lyobomirsky 2008; Roberts et al. 2003) this should be considered a gap. Theoretically, the strong focus on how ‘objective’ job characteristics influence ‘subjective’ opinions and attitudes often neglects people’s interpreting (or meaning making) capacity by implicitly assuming that experiences are interpreted similarly by all individuals. The importance of people’s interpretations of their own position is stressed, in particular, in the literature on the consequences of unemployment. The unemployed are thought to be at risk to get caught in a vicious circle in which their weak labour market position evolves in attitudes that hinder the improvement of their position (Charlesworth 1999). There is no reason to assume that such reciprocal effects only apply to unemployment situations. In this paper we focus on the employed and assess the relationship between the occupational status and feelings of relative deprivation.

Empirically, scholars often have to rely on cross-sectional data. Despite obvious advantages in terms of ease of data collection, (the direction of) causality needs to be assumed while working with this type of data. The existence of (strong) reciprocal effects would render that practice problematic. This is even more so given that the few studies that use panel data, in fact, all show effects in both directions (Marks and Flemings 1999; Boehm and Lyobomirsky 2008). Achieving a better understanding of this issue will clarify the scene against which cross-sectional analyses should be interpreted.

Against that background in this paper we used six-year panel data (age 23–29) to analyse the mutual relationship between “objective” occupational status and “subjective” feelings of relative deprivation among young adults. We consider the occupational status, as measured by the socio-economic status of the occupation (Ganzeboom et al. 1992) (operationalized as the average educational and economic resources of the occupation’s occupants, see also the Methods section) to be one of the most crucial indicators of the more general (multidimensional) socioeconomic status of people. The feelings of relative deprivation are “perceived undeserved collective disadvantage and associated emotions” (Pettigrew 2001, p. 352; more on this concept in Sect. 2.2). Because our data also contained a detailed description of the social background and educational career of the young adults, for which we can control in the analyses, they allowed us to scrutinize the mutual relationship between occupational status and attitudes in more detail than has been done so far.

This paper adds to the literature in five ways. Firstly, when presenting the analytical features of feelings of relative deprivation we explain why previous research on the impact of opinions and subjective well-being on indicators of occupational status has found mixed results. We argue that the group identification in feelings of relative deprivation is of crucial importance in this matter. Secondly, we extend previous research that demonstrated (1) the existence of such feelings of relative deprivation among pupils in secondary education, in particular in lower tracks of secondary education (Pelleriaux 2001; Demanet and Van Houtte 2011), thus establishing a link between these feelings and what could be considered as one of the main indicators of a pupil’s social position, i.e. track position [1], and (2) a link between these feelings of deprivation and school misconduct (e.g. being late for school, smoking on school grounds, …: Van Houtte and Stevens 2008). Our research tackles the evident follow-up research question concerning the relevance of these feelings for the early steps in youngsters’ professional careers. This research, thirdly, also contributes to a better understanding of the school-to-work transition. Our results show, among other things, that even after taking into account a detailed description of respondents’ personal characteristics and previous job experiences, youngsters’ social background still has a direct effect on their occupational status. Fourthly, whereas in general our results support the idea of reciprocal effects between relative deprivation and occupational status they also suggest them to vary over the course of the professional career. Finally, feelings of relative deprivation have proved (1) to be widespread among people performing low-status job and (2) to be politically highly consequential (for an overview see Derks 2006). Finding a net causal relationship between these feelings and the occupational status, indicates that the underpinning mechanism relating both is more complex than has been assumed thus far.

2 Theoretical Background and Expectations

In this paper we assessed the existence and strength of reciprocal effects between people’s occupational status and feelings of relative deprivation. Two reasons made us focus on the working population and their occupational status rather than the more often studied consequences and determinants of unemployment. First, we align with authors who consider unemployment to be an extreme position on a gradient ranging from low to high status positions: “the experience of frustration is, [however], not limited to the unemployed. A sizeable minority of those with jobs are frustrated too, albeit for different reasons and notwithstanding the fact that a large majority declare themselves satisfied with their jobs. Their frustration is due not to the emptiness and lack of experience that accompany being out of work but to the negative quality of their experiences in employment” (Jahoda 1982, p. 86; see also Burchell 1994, Graetz 1993). Focussing on the employed allowed us to study a larger target population. Secondly, and related to the first, an extensive literature demonstrates the relevance of the qualitative characteristics of the job (e.g., Bartley 1999; Janlert and Hammarström 2009) as well as perceptions of job insecurity or prestige (e.g., De Witte 1999; De Witte and Vets 2009; Glavin 2013) for health and well-being.

In the following paragraphs we first briefly review the general literature that assessed the impact of occupational characteristics on opinions and attitudes (Sect. 2.1). This literature suggests a clear variation in the extent to which different types of opinions and attitudes are affected by occupational status. Against that background we subsequently describe the analytical properties of feelings of relative deprivation and explain why direct effects of feelings of relative deprivation on occupational status are plausible (Sect. 2.2).

2.1 The Spill-Over Effects of Occupational Status

Although the empirical effects of different ‘objective’ qualities of people’s occupation on their ‘subjective’ job satisfaction are well documented (Spector 1997), scholars have always been interested in spill-over effects from job characteristics and job satisfaction to more general indicators of (1) subjective well-being—such as happiness (e.g., Boehm and Lyobomirsky 2008) or life satisfaction (e.g., Judge and Watanabe 1993; Marks and Fleming 1999), (2) mental health problems—like anxiety, burnout and depression (e.g., Faragher et al. 2005) or (3) more societal views—such as generalized trust (e.g., De Witte and Vets 2009).

It has regularly been observed that such spill-over effects are smaller when compared to the effects of job characteristics on different measures of job satisfaction. De Witte and Vets (2009), for example, compared the consequences of the perception of job instability on generalized trust and job satisfaction. Job insecurity increased both social distrust and job dissatisfaction. The association with job satisfaction, however, was significantly stronger than with social trust. Similarly, panel studies have shown that the spill-over effects from job to life satisfaction are rather small when measured as within-subjects changes (Judge and Watanabe 1993; Rode and Near 2005). Khattab and Fenton (2009) offer an explanation for that observation by showing that the influence of employment characteristics on satisfaction with life is mediated by sense of control. They only affect life satisfaction when working conditions deteriorate people’s sense of control.

The general conclusion that spill-over effects from employment characteristics to more general indicators of subjective well-being and societal views are limited should not misguide us. First, no matter how modest, spill-over effects underscore the general impact of people’s engagement in paid work. Secondly, research into spill-over effects is also an attempt to demonstrate the lasting effects of that labour market engagement, i.e. outcomes that gain a certain independence from the originating source. The latter is of course a prerequisite for the existence of reciprocal effects between occupational status and opinions.

Our choice in this paper to focus on feelings of relative deprivation is informed by both elements. First, it is well-known that feelings of relative deprivation are (1) more prevalent among people working in low-status jobs (Elchardus and Spruyt 2012) and (2) they prove to be fertile ground for radical political attitudes and behaviour (for a review see Smith et al. 2012). To the extent that societies aim to combat such feelings and behaviour a better understanding of their causal origins is of crucial importance. Secondly, research among pupils in secondary education suggests that feelings of relative deprivation (sometimes called “sense of futility”) tend to become self-fulfilling prophecies. Feelings of relative deprivation go together with behaviour, such as antisocial behaviour (Demanet and Van Houtte 2011; Van Houtte and Stevens 2008) and a lower study attachment (Van Houtte and Stevens 2010), which contradicts with what is expected at school. This aligns with the idea in the literature on labour market behaviour, of people being caught in a vicious circle. Indeed, interest in the relationship between the labour market position and attitudes partially stems from concerns that people in a weak labour market position—a qualitatively bad job or unemployment—develop feelings of fatalism which reduce their labour market attachment (Charlesworth 1999).

Although theoretically plausible, the empirical evidence for the idea of people being caught in a vicious circle, is mixed to say the least (e.g., Burchell 1993; Elchardus et al. 1996). Below we argue that feelings of relative deprivation may possibly solve this puzzle.

Studies that investigate the causal link between, for example, unemployment and indicators of specific aspects of life and general life satisfaction from a panel or career perspective, regularly find that dissatisfaction largely disappears from the moment people get back to work (for reviews of the literature, see De Witte 1993, 2003; Warr 1987; Verkleij 1988; Karsten and Klaus 2009). That finding suggests that when measured in terms of opinions strictly related to one’s personal life, the long-term negative effects of unemployment on well-being are rather modest (O’Brien 1996). Murphy and Athanasou’s review (1999) of longitudinal studies looking into the well-being effects of paid work, arrives at the same conclusion when they state that the effects of finding a job on subjective well-being are stronger than the effects of job loss. At the same time, however, a recent study covering a very long period (15 years) did find that job satisfaction influences the probability of retirement (Labriola et al. 2009). The long time span in that case may capture a process whereby enduring job dissatisfaction evolves into feelings of fatalism. It is plausible that in that particular case rather than job satisfaction itself, it is a derivate (fatalism, lack of sense of control: see Khattab and Fenton 2009) that is the real causal antecedent of the observed changes in labour market behaviour. Interestingly in this context, a panel study among different groups of school leavers in Australia (Feather and O’Brien 1986) found no impact of unemployment on a range of measures of self-concept, on stress, values or life satisfaction, however it did find an impact on attributions for youth unemployment. A shift from employment to unemployment was accompanied with an increased tendency to blame unemployment to socioeconomic causes, and a decreased tendency to see lack of motivation and other personal inadequacies as important causes of unemployment. Thus, the transition from employment to unemployment (and vice versa) affected elements that are related to an individual’s sense of control and feeling of self-efficacy. The latter are considered crucial elements of a coping strategy, which, as we argue in Sect. 2.2, might be the crucial factor in the emergence of a reciprocal relationship with occupational status.

Taken together the previous findings lead to two conclusions. First, they illustrate that the impact of employment may differ according to the type of outcome. That particular observation dovetails with public opinion research which shows a fundamental difference between what could be labelled as opinions about the “private” and the “public” sphere, that are, opinions strictly related to one’s personal life (e.g., satisfaction with life of job satisfaction) and opinions that include a diagnosis of society (Elchardus 2011; Funk 2000; Mutz 2006[1998]; Schnabel 2004; SCP 2003). Feelings of relative deprivation clearly include such a diagnosis as they present one’s own vulnerability and weak position as a consequence of injustice, as a form of discrimination against “people like us”, who in this society “never get what they deserve” (Derks 2006; Hogg et al. 2010). Secondly, the research findings suggest that when searching for reciprocal effects between ‘objective’ occupational status and ‘subjective’ opinions, opinions that include a societal diagnosis may be the most interesting. Too close a correspondence between employment characteristics and their subjective evaluation may undermine the necessary degree of independence of the latter in order to yield long-lasting effects on the former. That job satisfaction may not be the most relevant indicator in this context may also be inferred from studies which demonstrate its limited impact on work behaviour (e.g., Walsh and Tseng 1998 in their analysis of “work effort”; Morgan et al. 2013 who find that the intention to stay with an employer and job satisfaction in healthcare jobs were explained by different job characteristics). In the next section we discuss the concept of relative deprivation in more detail.

2.2 Feelings of Relative Deprivation

The concept of relative deprivation was introduced by Stouffer et al. (1949) and later refined (Runciman 1966; Vanneman and Pettigrew 1972; Grant and Brown 1995; Pettigrew 2001). Most of the time it is used for cases in which the groups or individuals that are being compared are very specific. From the onset of the theoretical work in this area, however, the specificity of the object of comparison was not a necessary condition for feelings of relative deprivation. The group with whom one compares the own group can be a generalized other, as is expressed in “people like us” (Runciman 1966, p. 12; Pettigrew et al. 2008, p. 387). Pettigrew et al. compared the explanatory power of group relative deprivation—tapped into with items with ‘people like yourselves’ as the subject—with individual relative deprivation with respect to ethnic prejudice and found that only group relative deprivation predicted ethnic prejudice.

Feelings of relative deprivation imply a sense of fatalism. Fatalism should be distinguished from low life satisfaction, low self-esteem or mere negative expectations for the future. Fatalism only becomes fatalism as soon as the negative perspective about one’s future becomes linked with the feeling of not being able to change one’s destiny (i.e., lack of control and low self-efficacy) (Meltzer and Musolf 2002, pp. 243-4). These two aspects may become linked in several ways. One way is by situating the cause of the personal situation outside the individual. This kind of linking is central in feelings of relative deprivation: the personal situation is presented as a consequence of a social situation in which some groups are favoured and others (“people like us”) are being disadvantaged (Smith et al. 2012).

Feelings of relative deprivation, sometimes called ressentiment, are “directed against lasting circumstances that are deemed injurious but beyond one’s control” (Meltzer and Musolf 2002, p. 244). In this way, feelings of relative deprivation are considered a consequence of ‘strain’ (Demanet and Van Houtte 2011). They are thought to result from (1) enduring goal blockage or (2) the inability to escape aversive situations (Agnew 2012). Obviously, a low-status job may yield both sources of strain.

Besides a feeling of lack of control, feelings of relative deprivation also entail a, albeit vague, group identification (Pettigrew 2001). The referencing to a group suggests that feelings of relative deprivation are to be viewed as a mechanism of coping, a strategy of interpretation that makes it possible to cope with a vulnerable situation (Mummendey et al. 1999). There are good reasons why the coping mechanism we describe as feelings of relative deprivation attracts vulnerable people. Common to all experiences of vulnerability is that they need to be coped with. Coping behaviour is not a process free from culture, nor is it a purely psychological matter. It takes place within and is shaped by the cultural context that challenges vulnerable people to explain their vulnerability discursively and to justify their reaction to it. Often this coping takes the form of a story for the simple reason that it is a very familiar and convincing cognitive structure (Bonilla-Silva and Embrick 2004; Tilly 2006). The most important function of these stories is to depersonalize experiences (e.g., Abrams and Hogg 2004; Hogg 2000; Hogg et al. 2008; Smith et al. 1994). Stories represent the personal vulnerability in such a way that the responsibility is put outside of the individual (Rydgren 2003, p. 53).

One of the easiest ways to achieve this is to represent the personal situation as part or even as a consequence of the opposition between groups in which one group disfavours the other group (Mummendey et al. 1999). When ‘my’ problems are represented as ‘our’ problems, ‘my’ responsibility in the vulnerability experienced decreases automatically. It is this aspect that stimulates those feeling vulnerable to group identification: “uncertainty […] ‘drives’ people to join groups” (Hogg 2000, p. 224). The further specification of that group does not need to be exact (Runciman 1966, p. 12). “People like us,” for instance, serves as a rhetorical means to put oneself under the safe umbrella of a group. Viewed in this way, feelings of relative deprivation are more than the superlative of an internalized personal default, as they are often proposed: at least partly, they can be interpreted as an attempt to protect one’s self-image. Murphy and Hampton (1988, p. 56) summarize this aptly when they state “ressentiment is ‘a protest’ against the demeaning action but also a defence against the action’s attack on one’s self-esteem”. It is the group identification in feelings of relative deprivation that enables these feelings to be easily shared by people and as such provides a common ground for the development of alternative status hierarchies (sometimes termed “counter cultures”). It is in this transformation process that feelings of relative deprivation will be easily self-confirming and once adopted become relatively independent of their originating source. As mentioned before, past research has shown that feelings of relative deprivation are already present among young people in secondary education (cf. supra). They direct pupils to tastes and behaviour that contradict with what is expected at school and so become self-fulfilling prophecies (see also Willis 1977). Investigating feelings of relative deprivation in the early labour market career is an evident follow-up research of this existing research. If feelings of relative deprivation are already being acquired in secondary education, and if these hinder the successful course of the educational career, do these attitudes also have consequences at the entry into the labour market?

Although this article’s objective is confined to establishing whether there exists an effect of feelings of relative deprivation on occupational status (in addition to the reverse effect), we should comment on how we think that such effect may come about. We expect the basic mechanism to concern one in which feelings of relative deprivation are part of a coping strategy in which people’s attention is withdrawn from labour market engagement to alternative (and often opposed) status hierarchies. “Social creativity” is the process by which people that experienced a negative social identity direct their efforts and emotional attachment to other status hierarchies (Jackson et al. 1996). Both ethnographers (see research into ‘counter cultures’, Charlesworth 1999; Willis 1977) and social psychologists (e.g., Mummendey et al. 1999) have provided ample evidence about the subtle ways in which such processes operate. Translated to occupational status at least two obvious pathways can be distinguished. First, strong feelings of relative deprivation are likely to reduce job performance and the associated chances on promotion (see Judge et al. 2001). Secondly, it seems also plausible that people who hold a fatalistic view for the future will not engage in strategies to ‘make’ that future by looking for a better job (Vansteenkiste et al. 2004).

3 Data and Measurements

The data for this study were drawn from a panel study among young people in Flanders, the Dutch-speaking part of Belgium. As part of the SONAR (acronym for ‘Study Group from Education to the Labour Market) research programme, a random sample was drawn from the National Register of young people living in Flanders who were born in 1976. In 1999, interviews were conducted face-to-face with 3,015 respondents from this birth cohort of 1976, at that time aged 23, about their educational and labour market careers (response rate of 66.4 %). A comparison of the background characteristics of the participants in this SONAR research with register data, has demonstrated a good representation by gender and region, but an underrepresentation of those with lower qualifications.

Two follow-up surveys were administered to the same people, in 2002 and 2005, at age 26 and 29. Retention rates were 68 and 80 % respectively, resulting in a final sample of 1,657. As a consequence of selective attrition [2], the final sample is somewhat more homogeneous than a single-wave survey would be. This implies most likely an underestimation of effects, strengthening our conclusions, should any effects be found.

The three waves took place in a period with moderately low unemployment rates in Flanders (respectively 5.1, 4.7 and 5.3 % for the age category of 20–64 years, according to Labour Force Surveys). In our analyses, we selected respondents who had a paid job at each of the three waves (and no missing data on the central variables), resulting in a final sample of 1,067.

3.1 Variables

The analyses focussed on the mutual relationship between occupational status and feelings of relative deprivation, each of which was measured in an identical way at each of the three waves. To estimate the relationship between feelings of relative deprivation and the occupational status accurately, we also included several control variables: gender, the highest educational level achieved, characteristics of the educational career and the educational background of the father.

The occupational status was measured by the International Socio-Economic Index of occupational status (ISEI) (Ganzeboom et al. 1992). Ganzeboom et al. (1992) define socioeconomic status scores as measures of “the positions of occupations in the stratification system” (p. 2) and explain it as follows “Our preferred way to think about SEI is that it measures the attributes of occupations that convert a person’s main resource (education) into a person’s main reward (income)” (p. 8–9). We derived this index from the Dutch classification of occupations (for details see Bakker et al. 1997) that was coded on the basis of respondents’ extensive description of the duties and tasks implied in their job. We divided the original scale (15–86) by 8 in order to better match the scales of the other variables in our analysis.

Feelings of relative deprivation were measured with six statements that respondents rated on a scale of 1 (totally not agree) to 4 (full agreement): (a) people like my parents and I do not need to expect a bright future, (b) my parents need to work hard and do not get compensated for that. That will most likely be the case for me as well, (c) whichever way you look at it, we are the kind of people who always have to do the heavy and dirty work, (d) we don’t know influential people and thus will always have to slave, (e) my parents have had a hard live, I will have a hard life and most likely my children too, (f) a nice career, that is not for our kind of people. Scale reliabilities are high (Cronbach’s alpha’s 0.87, 0.86 and 0.89 at age 23, 26, 29, respectively).

Principal components analysis confirmed the one-dimensional character of the set of items (the first component explained 65, 63 and 67 % of the variance in all of the items). For the analysis first a measurement model was tested (confirmatory factor analysis, with equality restrictions to assess measurement invariance over time) which is explained in more detail in Sect. 3.3.

Among the control variables were gender with men as reference category. We also controlled for the highest educational level that the respondent had attained with the categories of (a) no qualification of secondary education, (b) a diploma or certificate of (higher) secondary education, which implies having successfully finished 6 years of secondary education, (c) a diploma of higher education obtained at a college of higher education, and (d) a master’s degree obtained at a university. We used the qualification of secondary education as reference category as it was the largest category (about 42 %). We also included a measure for grade retention during secondary education: more than 29 % of the respondents have repeated a year, at least once, in secondary school. Finally, father’s educational level was included as a measure for parents’ social background [3], with the categories of (a) no diploma (almost 26%) which we used as the base category, (b) having successfully finished lower secondary education, normally until the age of 15, (c) higher secondary education, and (d) higher education at a college of higher education or a university.

3.2 Statistical Procedures

With the panel analyses we measured the occupational status and feelings of relative deprivation at the three points in time. We used a three wave cross-lagged effects panel model (Finkel 1995) which is an autoregressive model in which each dependent variable at time T was regressed on the same variable at (T-1), implying the prediction of the change in the dependent variable between T and (T-1). Furthermore, reciprocal effects were tested with cross-lagged effects. We were primarily interested in the extent that the occupational status and feelings of relative deprivation at an earlier moment in time affected relative deprivation and occupational status respectively at the following moment in time (see Fig. 1). To test all these postulated effects simultaneously, we used structural equation modelling (SEM), using the SEM-library for R-project (Fox 2006).

Fig. 1
figure 1

Conceptual model of the relationship between feelings of relative deprivation and occupational status among youth in Flanders

Full size image

This entailed a very strict test in which the causal effects over time were controlled for the synchronous relations between the two characteristics. We also performed the same models controlling in addition for background characteristics.

We used latent variable modelling for the measurement of feelings of relative deprivation, in particular to address two common issues in longitudinal analyses: correlated measurement errors and measurement invariance. Firstly, measurement errors were correlated over time. If not corrected for, effects may be biased considerably because the covariance at two time points is not separated into error covariance and covariance between the latent variables (Finkel 1995). Secondly, a comparison of constructs over time in panel analyses requires measurement equivalence (Brown 2006). Changes in measurement model parameters could indicate a change of meaning and validity of the measures used. Therefore, we also examined measurement invariance over time.

Model fit parameters to evaluate the models are the Chi square statistic, the root mean square error of approximation (RMSEA) and its associated pclose statistic, the squared root mean residual (SRMR) and Hoelter’s N.

3.3 Measurement Model for Feelings of Relative Deprivation and Invariance Over Time

First, we explored the factor structure of relative deprivation in separate models for each wave. We specified latent variables that were reflected by the indicators (cf. confirmatory factor analyses; Brown 2006). The model fits were moderate (RMSEA = 0.10, 0.13 and 0.13). Modification indices pointed out that the measurement models could be improved significantly by adding three additional associations between the individual items, i.e. between items (a) and (b), between items (c) and (e), and between items (c) and (f) [4]. The models in which these were also specified fitted more adequately (RMSEA = 0.06, 0.09 and 0.00).

Next, we specified a panel measurement model that contained the latent variables for feelings of relative deprivation at the three waves in one model. We added covariation between the three latent variables and we specified covariances between the errors of a measured indicator at the different waves (correlated measurement error). Fit for this model was adequate (Chi square = 308, df = 105, RMSEA = 0.043, pclose = 0.986, SRMR = 0.030, Hoelter’s N = 362).

Finally, we tested for measurement invariance, by comparing the previous (unrestricted) model with a model in which the factor loadings were restricted to be equal at each point in time (factorial invariance). This test of measurement invariance turned out to be statistically significant (LL ratio test: Chi square = 48, df = 10, p < 0.000). However, inspection of the size of the different factor loadings did not reveal large differences, and model fit was still very good (Chi square = 356, df = 115, RMSEA = 0.044, pclose = 0.961, SRMR = 0.037, Hoelter’s N = 343). This is why our final panel measurement model assumes stability of loadings across the three waves. Finally, to further guarantee measurement invariance over age, the size of the three item associations within each wave was made equivalent over time, although these equivalence restrictions again implied a loss of model fit (LL ratio test: Chi square = 21, df = 6, p = 0.002). However, this model had still good overall fit indicators (Chi square = 377, df = 121, RMSEA = 0.045, pclose = 0.959, SRMR = 0.038, Hoelter’s N = 341). In this final panel measurement model, all indicators had strong loadings on the latent factor (loadings >0.5, standardized factor loadings range from 0.63 to 0.86), which consequently can be interpreted as a stable indicator for feelings of relative deprivation. With an RMSEA that was lower than 0.05 (and a pclose well above the 0.10 limit that is commonly used), this measurement model could be considered to be a well fitting model.

4 Results

We combined the latent variable measurement model for feelings of relative deprivation (discussed above) with the occupational status measures in a three-wave cross-lagged effects panel model. As explained in the methods section, analyses are controlled for all previous measurements of relative deprivation and occupational status, at ages 23 and 26 (=stability effects). The most relevant effect parameters of this structural model estimating the reciprocal effects of feelings of relative deprivation and occupational status are presented in Table 1. The stability effects for feelings of relative deprivation were rather low. This suggests quite some change over time in relative positions taken up on this attitude among this group of young adults. Stability effects for occupational status were much higher [5].

Table 1 Cross-lagged models for feelings of relative deprivation and occupational status at ages 23–26–29 (Working People; N = 1,067)
Full size table

Next we turn to the substantively more interesting cross-lagged effects. Because we had three points in time, two comparisons can be made (age 23–26 and age 26–29). Between the age of 23 and 26, it appeared that effects in both directions were statistically significant: a higher occupational status at the age of 23 decreased the feelings of relative deprivation at age 26 (ß = −0.095), and higher feelings of relative deprivation at age 23 tend to result in a lower occupational status at age 26 (ß = −0.052). Between the age of 26 and 29, only the effect of occupational status was statistically significant, and in fact quite large (ß = −0.197), not the reverse effect of feelings of relative deprivation.

We also analysed a second model with control variables, in which, in addition to the other effects, both the feelings of relative deprivation and the occupational status at each wave were regressed on the background variables respondent’s gender, educational level, year repetition in secondary education and father’s educational level. As can be seen in Table 1, this did not change the parameters fundamentally, although most cross-lagged effects become smaller. The effect of occupational status at the age of 23 also is no longer statistically significant.

Taken together, these findings with regard to the cross-lagged effects indicate that whereas in general reciprocal effects between occupational status and feelings of relative deprivation exist, this reciprocity does not manifest itself within one period of observation. In the first part of the observation period (age 23–26) only a cross-lagged effect for feelings of relative deprivation on occupational status was observed (after controlling for background characteristics). In the second part of our observation window (age 26–29) only the effect from occupational status on feelings of relative deprivation proved to be significant. The importance of that variation in timing of effects between position and attitudes becomes clear when also considering the effects of the background characteristics.

Table 2 shows the effect parameters for the background characteristics in the second model. Each of the background variables did have some effect on the position and the attitude between the age of 23 and 29, although the educational level is clearly the most important predictor, for both occupational status and feelings of relative deprivation, with, as expected, higher occupational status and lower feelings of deprivation among those with higher qualifications. In addition, young adults from more privileged families tended to have lower scores on feelings of relative deprivation and higher occupational status; women also had more positive perceptions of their position in society. Contrary to our expectations, young adults who had had to repeat a year while in secondary education did not show more negative perceptions and were not less successful in securing a job with high occupational status.

Table 2 The effects of the control variables
Full size table

Most interesting was how the effects of the background characteristics changed over time. From the effect parameters we can infer that the educational level was very important at the age of 23, after which it loses some of its importance. In this, we recapitulate that due to the inclusion of the stability effects, for the age of 26 and 29 the effect parameters can be interpreted as affecting changes, while this is not the case for the first observations at the age of 23.

Nonetheless, in particular for the feelings of relative deprivation, by the age of 26 and 29 the educational level no more predicted changes in this attitude. Combined with our previous findings concerning the cross-lagged effects between our main variables of interest, clearly, over age, occupational status takes the place of educational level as most important predictor of how young adults perceive their position in society. In addition, also gender and father’s educational level, while relevant at age 23 for both feelings of deprivation and occupational status, did not have any clear effects by the age of 29 anymore.

All in all, then, the general pattern that emerged concerning the impact of the social background characteristics is one in which individuals’ feelings of relative deprivation and occupational status, become increasingly independent from people’s social background with increasing age. That observation, however, does not downplay the overall significance of that background for labour market outcomes. In particular the educational level of the father affects the occupational status at labour market entry significantly, even when controlling for labour market entrants’ educational background. Furthermore, we find high stability effects for occupational status. The dynamic nature of our model thus suggests a rather strong social reproduction.

5 Conclusions and Discussion

This study focused on the mutual relationship between feelings of relative deprivation and occupational status among young adults. In doing so, we have aimed to add to the literature on the causal links between objective characteristics of people’s socio-professional position and their opinions, attitudes and world views. We also expanded that literature by focusing (1) on the status of people’s occupation—rather than on the causes and consequences of unemployment—and (2) on feelings of relative deprivation—rather than opinions strictly related to the private sphere such as job/life satisfaction or mental health.

The results did indeed reveal reciprocal effects between feelings of relative deprivation and occupation status, but these effects vary in strength as well as in timing. The effect of feelings of relative deprivation on occupational status (controlled for background characteristics) is significant, but rather weak, and only noticeable between the age of 23 and 26. The effect of occupational status on feelings of relative deprivation on the other hand is much stronger but manifests itself only between the age of 26 and 29. Another finding is that, with regard to the control variables, the educational effects on feelings of relative deprivation by large have disappeared at the age of 29.

The general picture that emerges is one in which feelings of relative deprivation, which at a very young age mainly coincide with the barriers erected by different educational tracks in secondary education, gradually become determined more by the labour market position that is attained. In that sense, these feelings of deprivation ‘outgrow’ their roots in the educational system. This finding corroborates authors who assume strong causal effects of the labour market position on the attitudes of people. From a societal point of view however, this observation is less hopeful than it seems, as we do indeed find effects of feelings of relative deprivation on the occupational status in the first part of the observational window (age 23–26). Moreover, finding that the direct effects of people’s social background characteristics and education fade out throughout the professional career, does not undermine their strong indirect effects and associated social reproduction. The dynamic character of our analyses clarifies how social (dis)advantage is transformed in individual (dis)advantage.

People who feel deprived of a bright future, also effectively tend to realize a lower occupational status. That effect may be rather weak, but supplements all other effects that lead to a low occupational status, and thus is relevant. In addition, these effects are measured at a point in time—labour market entry—of which research demonstrates that it is crucial for the further development of the occupational career. A labour market career in that context resembles the trajectory of a football after being kicked: the sooner the deviation, the larger the consequences. It is that property which renders feelings of relative deprivation important and therefore deserves to be studied in a better way than has been the case up till now. Our research findings also suggest new possibilities for future research on this topic.

In this article we analysed the relationship between feelings of relative deprivation and occupational status. Needless to say that, in the context of the relationship between opinions and labour market position, this is but one of the possible aspects of employment that can be assessed. Our focus in this paper was mainly inspired by recent concerns with respect to the growth of the number of low-quality jobs (e.g., Uchitelle 2006). As a consequence, the analysis was restricted to respondents who worked at the three points of observation. Some scholars consider the difference between employment and unemployment to be fundamental (e.g. Karsten and Batanic 2010). Further research should assess whether and how the transition from employment to unemployment (or vice versa) is related to attitude change. Moreover, more complex measures that combine different indicators of what can be called ‘a problematic labour market entry’ (such as the number of different jobs, periods of unemployment, short-term contracts, …) may yield a more complex picture of the general relationship between objective characteristics of someone’s social position and subjective interpretations.

At the theoretical level we devoted particular attention to clarifying the conceptual status of feelings of relative deprivation. We argued that the vague group identification in relative deprivation enables these feelings to be easily shared by other people and as such to provide a common ground for the development of alternative status hierarchies. That is the reason why feelings of relative deprivation and not satisfaction with life strongly predicts support for populism and political extremism (Elchardus and Spruyt 2012). That element combined with the empirical finding that between the ages of 26–29 these feelings are rather strongly influenced by changes in the status of people’s occupations, raises two questions, which provide a starting point for new research. First, how exactly does the relationship between position and attitude we found here relate to macro-tendencies such as the dualisation of the labour market? One hypothesis that can directly be derived from our findings is that feelings of relative deprivation and their societal consequences are higher in countries where this process of dualisation is more pronounced. The region we studied, Flanders, is known for its well-developed system of labour market regulation (with legal minimum wages and strong employment protection) and social security system. It would be interesting to compare the results we obtained with those from countries like Germany, the UK or the US. Secondly, even though it is substantial, the relationship between occupational status and feelings of relative deprivation is not perfect. Further research should clarify under which circumstances such spill-over effects occur. Relative deprivation theory stresses the importance of comparison processes. This suggests that besides objective job qualities, the context in which these work conditions are experienced could be of crucial importance. People performing low-quality jobs in a heterogeneous network, for example, may be more susceptible to feelings of relative deprivation because when they compare themselves with their significant others the outcome will be far more negative than when comparing with people engaged in a homogenous network.

6 Notes

[1] Secondary education in Flanders is characterized by strong educational tracking. Pupils are divided over four tracks: (1) vocational education that prepares for the labour market, from which few pupils go on to higher education and almost none to universities; (2) technical education preparing for the labour market and for some types of higher technical training, from which a small proportion also go on to universities; (3) general education, from which almost all pupils (more than 90 %) move on to higher education and about half go to universities and (4) an arts education that prepares for higher education in the arts, which comprises only about 2 % of the pupils (De Ro 2008). In this way the educational track position is in Flanders highly consequential for the final educational attainment level and indirectly for the occupational status.

[2] An important issue related to the use of panel data concerns selective attrition. Retention in the later waves indeed showed some variation according to the key variables in this article. With regard to the working situation, no difference was found between the ages of 23 and 26 (B = 0.008, p = 0.648); however, those who were working at the age of 26 had a somewhat higher probability of participating in the survey at age 29 (B = 0.624, p = 0.000). With regard to relative deprivation, again, those who dropped out between ages 23 and 26 do not differ from those who participated in the survey at age 26 (B = 0.021, p = 0.572); however, between ages 26 and 29 drop-out was selective with lower retention probability among those demonstrating higher scores for relative deprivation (B = −0.105, p = 0.006).

We scrutinized the attrition further by means of logistic regression analyses in which we included all variables used in our analyses. The results (available from the authors on request) show selective attrition according to educational level, and some other effects, but overall, explanatory power is rather low (6 and 2 % of the participation probability at wave 2 and 3 respectively).

[3] For the sake of parsimony we report in this article only on the effect of father’s educational level. Incorporating additional indicators of the family background, in particular mother’s educational level, did not alter our results (results available from the authors upon request).

[4] These were added to the panel measurement model in an iterative way, based on additional tests of the measurement model for each age separately, and on the modification indices of the integrated panel measurement model.

[5] We undertook some additional analyses to check whether methodological issues could account for this unexpected finding. A comparison with a model in which the feelings of relative deprivation were measured as a sum score instead of a latent variable showed that it was not due to the correction for measurement error for relative deprivation. Another explanation may be the rather strong formulation of the items of relative deprivation. Only about 20–40 % of respondents agree with any of the statements. Estimates for the same model but using polychorical and polyserial correlations instead of the variance–covariance matrix, which is more appropriate for ordinal variables, indeed reveal somewhat higher stability-effects for relative deprivation (but still rather low). In addition, also the cross-lagged effects of relative deprivation are somewhat higher, as well as increased explained variances particularly for relative deprivation.