ABSTRACT
In this article we answered the research question to what extent variation in extrinsic and intrinsic job preference orientations can be attributed to and explained by differences between individuals and between countries. We argued that socialization in school, economic deprivation, and job quality influence job preferences, and formulated testable hypotheses on the individual and country level. After first testing for cross-national equivalence of the latent constructs and assuring that factor solutions were satisfactory, we employed multiple response multilevel models on a subset of 22 countries in the European Values Study 1999/2000. The findings indicate that higher educational attainment, a high income, working in managerial and higher professional jobs, and having autonomy in one's job stimulate intrinsic job preference orientations, while particularly educational attainment and autonomy temper extrinsic work values. Workers in semi- and unskilled manual jobs have the highest extrinsic job preferences. On top of individual characteristics, living in a nation that invests much in human capital or has a high quality labour market is associated with lower levels of extrinsic job preferences. Moreover, countries with socio-economic features that reduce the risk of economic deprivation have a more intrinsically motivated workforce.
1. Introduction and research question
An intrinsically motivated work force is a desirable goal for present-day societies. An intrinsic job motivation implies that the aim of work lies in the work itself, and stresses the wish for making use of one's abilities and for personal development. An extrinsic job motivation regards work as a means to achieve goals outside work, and is therefore largely driven by the value of material rewards (De Witte et al. 2004; Ester et al. 2006). The shift to a service economy and the upgrading of the occupational structure (Skill-Biased Technological Change) have enlarged the difficulties in monitoring employees’ work due to the greater degree of autonomy and the more specific knowledge that characterize jobs nowadays (e.g., Berman et al. 1998; Spitz-Oener 2006). As a result, employers are more dependent on their employees’ desire to do their job well (Gallie 2007a). It has been demonstrated that, among the new skills that are supposed to be critical due to the changing characteristics of jobs, employers consider work attitudes, such as motivation, to be more important than cognitive or generic skills (Heckmann and Rubinstein 2001; Farkas 2003). Ester et al. (2006) particularly link the degree of intrinsic motivation of the work force to the degree in which a country is capable to operate effectively in a globalizing economy that requires flexibility and employability. Using multiple waves of the European Values Study, they show that in 11 out of 13 countries intrinsic job preferences grew in importance between 1981 and 1999, implying that as a consequence of growing job quality due to Skill-Biased Technological Change, workers became increasingly intrinsically motivated. In this study we investigate intrinsic and extrinsic job motivations in a cross-national design examining individual and contextual level explanations for variation in job preferences.
Recently, Gallie (2007a) has structured individual level explanations of job motivation into three domains: (a) socialization, (b) economic deprivation, and (c) quality of the job. We will follow this distinction, but our main contribution will be by extending it with hypotheses on the country level. Despite the potential beneficial impact job preferences have on the societal level, little is known about differences in job preferences between countries. In recent years, some studies compared job preferences between a limited set of (carefully selected) countries (e.g., Hult and Svallfors 2002; Gallie 2007a). These studies proposed interesting theoretical expectations about why countries would show different patterns of job preferences (Hult and Svallfors 2002; Gallie 2003, 2007a, b). In our view, to truly understand the value of these theoretical expectations, empirical tests need to be improved upon. We aim to elaborate on the cross-national approach in job preferences studies (a) by formulating testable hypotheses to explain differences between countries and (b) by introducing several empirical improvements.
First, we complement to earlier empirical research on this topic by analyzing a large set of countries (22), which enables us to test the ideas about the influence of countries’ characteristics by inserting specific theory-based country characteristics (multivariately) into multilevel models. Secondly, we explicitly take account of the positive relationship between intrinsic and extrinsic job preferences. The assumption that someone is either intrinsically or extrinsically motivated implies that if governments succeed in reducing extrinsic preferences, they automatically direct people into intrinsic preferences. Theoretically, this assumption seems plausible. However, empirically intrinsic and extrinsic job motivations are usually highly positively correlated (0.56 on average in the 22 countries under study). Therefore, instead of using single response models, we will apply multiple response models that take account of this positive correlation, resulting in more reliable estimates of the effects of individual and country characteristics on intrinsic and extrinsic job preferences.
Finally, we propose an improved measurement of job preferences by using scalar invariance models that test for cross-national equivalence (Muthén and Muthén 2006; Saris and Gallhofer 2007). This approach reduces the risk of comparing countries on a concept that has different meanings in the different contexts (see our data section for an elaborate discussion). To answer our research question to what extent can variation in job preferences be attributed to and explained by differences between individuals and between countries, we analyse data from the European Values Study 1999 that will be enriched with relevant country-level indicators derived from international statistical databanks.
2. Theory and hypotheses
2.1 Individual-level hypotheses
Following Gallie's (2007a) distinction into three mechanisms that explain differences between individual job preferences, we will discuss the expected influence of three kinds of factors: (a) socialization, (b) economic deprivation, and (c) quality of the job.
As opposed to a focus on material gains, socialization directed at enhancing personal autonomy and self-development is expected to have a long-lasting effect on job preferences (Argyris 1964; Gallie 2007a). The educational system is considered as a major socializing institution (Gallie 2007a). It is found that school curricula pay much attention to teaching pupils general, pro-social moral values (Cogan and Morris 2001; Solomon et al. 2001) and, particularly in the higher educational levels, there is a strong focus on self-development. Similarly, we can expect that in these higher levels, pupils are taught to focus on content rather than on material output, as intrinsic job preferences are generally perceived as more desirable values. Empirical research indeed shows that people who have reached higher levels of educational attainment tend to endorse an intrinsic work motivation and to reject an extrinsic work motivation (Lindsay and Knox 1984; Rose 2005; Gallie 2007a;Gesthuizen 2009).
The impact of economic deprivation corresponds closely to the idea of Maslow's (1943, 1970) hierarchy of needs. People will first be concerned with lower-order needs (food, shelter, security), and only if one has sufficient resources to fulfil those lower-order needs can people strive for higher-order needs (Pattie et al. 2004), such as self-actualisation (see also Inglehart 1990). In parallel, we connect the lower-order needs in Maslow's hierarchy to the material rewards that are central in extrinsic job preferences, and the higher-order needs such as self-actualisation to the characteristics that are valued in an intrinsic job orientation. We assume that people can only relax their focus on extrinsic – material – job features and strengthen their focus on intrinsic ones if their resources allow them to, i.e., if they do not suffer from economic deprivation. We therefore hypothesize that income is positively related to intrinsic job preferences and negatively related to extrinsic job preferences.
Finally, we assume that the work environment that employees are exposed to through their own job, directly influence their job preferences (Gallie 2007a). If the quality of a job is low, that is, little autonomy, non-interesting tasks, poor prospects, or physically hard work (Acemoglu 2001; Kalleberg et al. 2001), it is plausible that the employee will not be triggered to focus on the job itself, but will do his or her job in order to be able to find satisfaction outside work. This view corresponds with the extrinsic work orientation in which work is regarded as a means to achieve goals outside work. Using the same reasoning, it can be argued that high-quality jobs induce an intrinsic work orientation: those jobs meet the requirements for getting satisfaction out of the work itself, which is the essence of an intrinsic work orientation (Hackman and Oldham 1976). Indeed, intrinsic job and organizational characteristics have a stronger impact on overall job satisfaction, job involvement and organizational commitment than extrinsic ones (Kalleberg 1977; Gallie et al. 1998; Moynihan and Pandey 2007). Moreover, research using longitudinal panel data shows that job characteristics have a causal impact on job preferences. Most notably, Lindsay and Knox (1984) provide evidence that pupils with intrinsic work values attain higher diplomas, that higher educated individuals subsequently enrol in jobs that score high on intrinsic characteristics, but that in addition to this selection process, there is also a direct and strong positive impact of intrinsic job characteristics on subsequent intrinsic job preferences. We therefore expect that the quality of one's job (measured by occupational class and degree of autonomy) is positively related to intrinsic job preferences and negatively related to extrinsic job preferences.
The longitudinal study mentioned above exemplifies that there is no one-way causal relationship between quality of the job and job preferences. Other studies suggest similar conclusions. Mismatches between actual job characteristics and job preferences result in reduced job satisfaction (Kalleberg 1977; Rose 2005) and voluntary job mobility, which in turn reduces the mismatch between job characteristics and job preferences (Kalleberg and Mastekaasa 2001; Gesthuizen and Dagevos 2008). This implies that people with intrinsic job motivations will look for jobs that meet their intrinsic preferences. In a meta-analysis, Kristof-Brown et al. (2005) show that a negatively evaluated person-job, person-organization, person-group, and person-supervisor fit results in, for instance, weaker organizational commitment, lower job satisfaction, and a higher inclination to quit the job and start searching for another. Evidently, in this cross-national study we cannot make strong claims about the causal relationship between job quality and job preferences. Nevertheless, as we are able to control for the strongest factor of allocating people into lower or higher quality jobs – that is, educational attainment (see Lindsay and Knox 1984) – we believe that the effect of job quality that we observe is causal to a large extent.
2.2 Country-level hypotheses
We also use theoretical perspectives on socialization, economic deprivation, and the quality of the labour market to specify and explain which country characteristics affect job preferences. The general assumption that underlies the specific country-level hypotheses is that national contexts affect the value patterns of populations independent of individual characteristics. People are confronted with policy decisions, a cultural climate, and structural opportunities and limitations (Hult and Svallfors 2002; Gallie 2007a) to which they will react and accordingly adjust their values and behaviour to a certain extent.
Our first specific hypothesis is based on the assumption that governments also play an important role in socializing citizens (next to families, schools and networks). The choices governments make on spending public money signal what is valued highly in society (Curtis et al. 2001; Kääriäinen and Lehtonen 2006). Probably, governments underline the norm of intrinsic motivations more than extrinsic motivations, as an intrinsically motivated work force is in their best interest: it stimulates productivity and economic prosperity. The more strongly one is exposed to the societal norm of intrinsic motivation, the more likely one is to endorse it. Investing in human capital is one way in which a government can spread this norm, and that will inspire people to value intrinsic job features. Consequently, we expect that countries with high levels of human capital investments show higher levels of intrinsic job preferences and lower levels of extrinsic job preferences.
Welfare regimes differ markedly in the degree to which they prevent economic deprivation, for instance in case one becomes unemployed or disabled, or if one retires (Esping-Andersen 1990; Gallie and Paugam 2000; Nolan et al. 2000; Whelan and Maître 2005). A high level of social security takes away potential fear for the consequences of job loss since a reasonable living is secured even in case of unemployment, disability, and non-employment (Van Oorschot and Arts 2005). Furthermore, a higher income inequality increases the risk of falling deep, and high unemployment signals a country's incapability to provide job security. Following the hierarchy of needs principle (Maslow 1943, 1970), safeguarding basic needs allows for relaxation of a focus on extrinsic job features and for strengthening the focus on intrinsic job features. Indeed, in a comparison of four countries, Russell (1998) found that in stronger welfare regimes commitment to work was stronger, and less importance was attached to extrinsic job characteristics. We expect that the better countries are able to safeguard citizens against economic deprivation, the more intrinsically motivated their work force will be, and the less extrinsically motivated.
Finally, we assume that the quality of the labour market in general influences job preferences, over and above the quality of the job that a worker holds individually.1 The type of jobs – and hence the quality of jobs countries offer to their population – vary. The varieties of capitalism literature argues that, as compared to liberal market economies, coordinated market economies more strongly emphasize skill development and quality of production (Hall and Soskice 2001) and therefore offer higher quality jobs in terms of skill levels, autonomy, participation at work, and job security. In comparative perspective it has been shown that Scandinavian countries, which have co-ordinated market economies, score relatively high on the quality of job tasks and employee participation (Dobbin and Boychuk 1999; Gallie 2003), creating ‘an ethos in which employees attach particular importance to intrinsic characteristics of work’ (Gallie 2007a: 282; see also Hult and Svallfors 2002; Gallie 2007b). We connect to this and also hypothesize that the higher quality of the labour market in general (measured by the proportion in professional occupational classes combined with the average level of autonomy), the less extrinsically and the more intrinsically motivated individual workers are.2
3. Data
We used data from the European Values Study (EVS) 1999/2000 (release 2, May 2006 – Integrated Dataset). This large-scale data collection has been based on nationally representative samples and contains questions on values regarding several domains. Response rates are almost 60 percent on average. Detailed information on the European Values Study can be found in a publication by Halman (2001) and at www.europeanvaluesstudy.eu.
In the construction phase of the dependent variables, intrinsic and extrinsic job motivations, we started out with 31 countries: Austria, Belarus, Belgium, Bulgaria, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Germany, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, The Netherlands, Poland, Portugal, Romania, Russia, Slovakia, Slovenia, Spain, Sweden, Turkey, Ukraine, and the United Kingdom.3 The analyses are based on 22 countries (Germany, Austria, Denmark, Finland, Slovakia, Russia, Slovenia, Turkey, and Hungary are excluded) in order to be sure that measurements are comparable across the countries to be examined. Details will be outlined in the next section. We restricted our sample to the working labour force. After accounting for missing values on dependent and independent variables, our analyses are based on 13,211 respondents.
4. Measurements
4.1 Dependent variables: extrinsic and intrinsic job preferences
Generally, extrinsic job preferences are defined as valuing jobs for their material rewards, while a concern to make use of abilities defines intrinsic preferences (Gallie 2007a: 279; see also Kalleberg 1977; Ester et al. 2006; Gesthuizen and Dagevos 2008). Recently, Ester et al. (2006) summarized research on work values, in order to assess to what extent there is inter-scholarly agreement regarding indicators of extrinsic and intrinsic job preferences. Using the EVS 1999/2000 question ‘Here are some aspects of a job that people say are important. Please look at them and tell me which ones you personally think are important in a job’, they concluded that there is a high level of consensus that (1) good pay, (2) not too much pressure, (3) good job security, (4) good hours, and (5) generous holidays indicate extrinsic preferences, while (1) an opportunity to use initiative, (2) a job in which you feel you can achieve something, (3) a responsible job, (4) a job that is interesting, and (5) a job that meets one's abilities refer to intrinsic job preferences. We connect to these findings by also using these indicators, yet at the same time, we opt for several advancements.
Firstly, instead of treating these indicators as continuous indicators of a latent theoretical construct that is assumed to be measured on a continuous scale (as Ester et al. (2006) do), we treat them as binary dependent variables, the outcome of which depends on the respondent's true but unmeasured extrinsic and intrinsic job preferences. We use confirmatory factor analysis with binary/categorical factor indicators in Mplus (Muthén and Muthén 2006: 33) to estimate the respondent's factor scores on extrinsic and intrinsic job preferences, giving right to the level of measurement of the factor indicators, which are dichotomous (mentioned/not mentioned) and cannot be easily used in summated scales.
Secondly, we allow extrinsic and intrinsic factor scores to correlate. Recently, Gallie also used factor analysis to indicate similar latent constructs, but instead assumed that these orientations would be independent by employing varimax rotation in the data reduction analysis, after which the extrinsic score was subtracted from the intrinsic score (2007a: 283, 284). Our results show that this procedure is not very realistic: the average correlation is 0.56. We therefore decided to allow for two dimensions of work values, and for a correlation between them. Possibly, the positive correlation between intrinsic and extrinsic work values results from a common dependence on a third factor, namely work ethic. But even if we control for work ethic, the correlation between the two forms of motivation remains high.
Thirdly, instead of assuming that the factor indicators measure identical latent constructs across all countries (e.g., Ester et al. 2006; Gallie 2007a; Kalmijn and Kraaykamp 2007), we will actually test for cross-national equivalence by using scalar invariance multi group models4 in Mplus (Muthén and Muthén 2006). Saris and Gallhofer (2007: 341) claim that ‘… without scalar invariance the means of the computed composite scores cannot be used as indicators to compare means across countries’. Scalar invariance models constrain unstandardized factor loadings and the intercepts of the latent constructs to be equal across countries.5 In Mplus one can specify multi group models to test for measurement equivalence across groups while using factor indicators that are binary. As far as we know, this is the first attempt to reach that. We will only use those countries in our multilevel models for which we are sure that average scores on the dependent variables are allowed to be compared.
The results of our final model (results are not shown, but available upon request) show that the most restrictive scalar invariance model already shows a fit which approaches adequacy. The RMSEA (Root Mean Square Error of Approximation) is 0.089 (lower than 0.080 denotes an adequate fit; Hu and Bentler 1999). After allowing for correlations between the unexplained parts of the factor indicators, the RMSEA dropped to 0.079. However, for some countries the factor solution was unsatisfactory. Following common practice – albeit on the lenient side – we decided this to be the case if factor loadings and/or communalities were lower that 0.10, so that, on the basis of unreliable and therefore incomparable results, Germany, Austria, Denmark, Finland, Slovakia, Russia, Slovenia, and Turkey had to be excluded. We also decided to exclude Hungary given its outlier position on the average score on intrinsic job preferences,6 so that 22 countries remained. Finally, extrinsic and intrinsic job preferences are strongly correlated. The average correlation is 0.56, and for all countries except Slovakia, the correlation is positive.
4.2 Independent variables at the individual level
Education was measured by three categories. We assigned to the low-educated category those who completed lower secondary education at the highest. The intermediate category consists of the respondents who completed higher secondary education. The respondents who attained a tertiary-level diploma are categorized as high-educated. We also included a dummy for missing cases. Although parental education and occupation could have served as indicators for the socialization explanation, unfortunately they are unavailable in our data. Net household income is measured in deciles. We used mean substitution to replace missing values. As an indicator of job quality, occupational class is measured in eight categories: managers/employers, professional workers, middle level office workers, junior level office workers, skilled manual workers, semi- and unskilled manual workers, farmers, and members of the armed forces. A second indicator of job quality is the level of autonomy, that is: ‘how free are you to make decisions in your job?’, using a 10-point scale (from low to high freedom). Furthermore, we constructed six control variables. We measured age as a continuous variable that ranges from 1.8 to 6.4. To measure gender, we include a dummy for females. Household composition is captured by including having a partner and having children. A scale has been constructed to measure work ethic (Cronbach's alpha = 0.70). It consists of ‘a job is needed to develop one's talents’, ‘it is humiliating to receive money without working’, ‘people turn lazy if they are not working’, ‘working is a duty towards society’, and ‘work comes always first’. As the original 5-point scale varies between (1) strongly agree and (5) strongly disagree, the higher the average score of the respondent is on this scale, the lower the work ethic is. Finally, having a primary or secondary job is indicated by the question whether or not the respondent is the chief wage earner of the household.
4.3 Independent variables at the national level
We constructed national characteristics for each of the three theoretical domains. First, we used a measure to indicate to what extent a nation stimulates educational expansion. The measure human capital investments was based on the Eurostat 1999 indicator of the percentage of GDP that a government spends on the acquisition of human resources.7 We combined four characteristics that together indicate the extent to which countries are able to prevent from or protect against economic deprivation. Social security was measured as the percentage of GDP spent on social protection, which is provided by Eurostat.8 To indicate a nation's wealth, we include GDP derived from Eurostat.9 The degree of income inequality was measured by the GINI-coefficient calculated by the United Nations based on the year 1999. Unemployment rates are based on Eurostat figures. We used percentages from the year 1999, or the closest year if the figure for 1999 was missing.10 For reasons of parsimony, we computed factor scores for countries based on a principal factor analysis. The four indicators proved to load on one single dimension (factor loadings are higher than 0.45 and the eigenvalue is 2.29, so that the single factor explains almost 50 percent of the total variance). A higher score on this dimension indicates a lower level of economic deprivation. Finally, two measures were combined to indicate the average quality of the labour market in a country. First, we distilled the quality of jobs from the International Labor Organization database (LABORSTA): this is the percentage of individuals that worked in the first-digit-ISCO88 groups 1 (legislators, senior officials and managers), 2 (professionals) and 3 (technicians and associate professionals). Second, the quality of work conditions is the average level of autonomy in jobs within a country (aggregated from EVS). Again, we tested whether both indicators load on one single dimension. This is indeed the case, with an eigenvalue of 1.39 (39 percent the total variance).
| . | Minimum . | Maximum . | Mean . |
|---|---|---|---|
| Dependent variables | |||
| Extrinsic job preference | −1.80 | 1.28 | −0.19 |
| Intrinsic job preference | −5.42 | 5.63 | 0.51 |
| Independent individual characteristics | |||
| Low education | 0 | 1 | 0.25 |
| Medium education | 0 | 1 | 0.48 |
| High education | 0 | 1 | 0.26 |
| Education missing | 0 | 1 | 0.00 |
| Income | 1 | 10 | 5.46 |
| Managers/employers | 0 | 1 | 0.09 |
| Professional workers | 0 | 1 | 0.17 |
| Middle level office workers | 0 | 1 | 0.17 |
| Junior level office workers | 0 | 1 | 0.13 |
| Skilled manual workers | 0 | 1 | 0.24 |
| Semi- and unskilled manual workers | 0 | 1 | 0.18 |
| Farmers | 0 | 1 | 0.02 |
| Members of armed forces | 0 | 1 | 0.01 |
| Chief wage earner | 0 | 1 | 0.64 |
| No chief wage earner | 0 | 1 | 0.36 |
| Age | 1.80 | 6.40 | 3.91 |
| Female | 0 | 1 | 0.47 |
| No partner | 0 | 1 | 0.25 |
| Partner | 0 | 1 | 0.75 |
| No children | 0 | 1 | 0.28 |
| Children | 0 | 1 | 0.71 |
| Children missing | 0 | 1 | 0.01 |
| Autonomy | 0 | 10 | 6.46 |
| Work ethic | 1 | 5 | 2.55 |
| Independent country characteristics | |||
| Human capital investments | −1.82 | 2.21 | −0.05 |
| Risk reducing socio-economic factors | −1.14 | 2.54 | 0.15 |
| Quality of labour force and jobs | −1.12 | 1.48 | 0.11 |
| . | Minimum . | Maximum . | Mean . |
|---|---|---|---|
| Dependent variables | |||
| Extrinsic job preference | −1.80 | 1.28 | −0.19 |
| Intrinsic job preference | −5.42 | 5.63 | 0.51 |
| Independent individual characteristics | |||
| Low education | 0 | 1 | 0.25 |
| Medium education | 0 | 1 | 0.48 |
| High education | 0 | 1 | 0.26 |
| Education missing | 0 | 1 | 0.00 |
| Income | 1 | 10 | 5.46 |
| Managers/employers | 0 | 1 | 0.09 |
| Professional workers | 0 | 1 | 0.17 |
| Middle level office workers | 0 | 1 | 0.17 |
| Junior level office workers | 0 | 1 | 0.13 |
| Skilled manual workers | 0 | 1 | 0.24 |
| Semi- and unskilled manual workers | 0 | 1 | 0.18 |
| Farmers | 0 | 1 | 0.02 |
| Members of armed forces | 0 | 1 | 0.01 |
| Chief wage earner | 0 | 1 | 0.64 |
| No chief wage earner | 0 | 1 | 0.36 |
| Age | 1.80 | 6.40 | 3.91 |
| Female | 0 | 1 | 0.47 |
| No partner | 0 | 1 | 0.25 |
| Partner | 0 | 1 | 0.75 |
| No children | 0 | 1 | 0.28 |
| Children | 0 | 1 | 0.71 |
| Children missing | 0 | 1 | 0.01 |
| Autonomy | 0 | 10 | 6.46 |
| Work ethic | 1 | 5 | 2.55 |
| Independent country characteristics | |||
| Human capital investments | −1.82 | 2.21 | −0.05 |
| Risk reducing socio-economic factors | −1.14 | 2.54 | 0.15 |
| Quality of labour force and jobs | −1.12 | 1.48 | 0.11 |
Data: European Values Study 1999/2000, N=13,211.
5. Method: multilevel multiple response models
To test our hypotheses regarding the influence of individual and national characteristics on job preferences, we estimate a combination of multilevel and multiple response models. We used the saved factor scores from the final multi group scalar invariance model (the results are available upon request) as dependent variables. As individuals are nested within countries, two level models suffice to correct for clustering of units and to distinguish within from between country variance. We first present results from empty models (M1) to indicate the relative importance the individual and national context. In a second model (M2) we add individual characteristics, after which we add the national characteristics one at the time (M3–M5). We conclude with a final model in which all the effects of national characteristics are estimated simultaneously, while at the same time controlling for individual characteristics (M6).
We argued that it is important to take account of the positive correlation between the two job preference orientations. For instance if, as hypothesized, there is a negative relationship between educational attainment and extrinsic job preferences, its strength will be underestimated if one does not take account of the possibility that education at the same time has an indirect positive effect on extrinsic job preferences via an intrinsic job preference orientation. The reason is that educational attainment also positively affects intrinsic job preferences, which in turn is positively associated with the extrinsic dimension. To overcome such biases in the estimations we estimate multiple response multilevel models. These models take account of the indirect effects of individual and national characteristics on preferences for extrinsic job characteristics via preferences for intrinsic ones, and vice versa.
6. Results: individual and national level determinants of job preference orientations
We first assess to what extent multilevel modelling is appropriate. Table 2 shows significant country variances for both extrinsic and intrinsic job preferences. The intraclass correlations of 0.222 and 0.102, respectively, exemplify that there are large differences across countries in average levels of job preferences. On average, The Netherlands scores lowest on extrinsic job preferences, Croatia highest. Sweden is the most intrinsic country, while Latvia is the least intrinsic (the country scores on both dependent and independent contextual variables can be found in Table A1 in the appendix). The country variation is still significant if we take into account compositional differences between countries. Quite large drops in country variances are observed when the full set of national characteristics are added.
| . | Extrinsic job preferences . | Intrinsic job preferences . | ||||||
|---|---|---|---|---|---|---|---|---|
| . | (individual) . | (country) . | (individual) . | (country) . | ||||
| . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . |
| Empty model | 0.190 | 0.038 | 0.055 | 0.014 | 1.812 | 0.544 | 0.210 | 0.046 |
| +Individual | 0.188 | 0.037 | 0.052 | 0.013 | 1.786 | 0.535 | 0.194 | 0.042 |
| +Country | 0.188 | 0.037 | 0.035 | 0.008 | 1.786 | 0.535 | 0.154 | 0.043 |
| . | Extrinsic job preferences . | Intrinsic job preferences . | ||||||
|---|---|---|---|---|---|---|---|---|
| . | (individual) . | (country) . | (individual) . | (country) . | ||||
| . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . |
| Empty model | 0.190 | 0.038 | 0.055 | 0.014 | 1.812 | 0.544 | 0.210 | 0.046 |
| +Individual | 0.188 | 0.037 | 0.052 | 0.013 | 1.786 | 0.535 | 0.194 | 0.042 |
| +Country | 0.188 | 0.037 | 0.035 | 0.008 | 1.786 | 0.535 | 0.154 | 0.043 |
Data: European Values Study 1999/2000, N=13,211
Model 2 in Table 3 shows the extent to which individual characteristics are related to extrinsic and intrinsic job preferences. The higher educated find extrinsic work values less important than lower educated workers, while education is indeed positively related to an intrinsic work orientation. This corroborates our individual level socialization hypothesis. For intrinsic job preference orientations, income also acts in line with our predictions based on economic deprivation, as the relationship proves to be positive. Finally, our indicators of the quality of work also affect job preferences the way we hypothesized. As compared to managers/employers, middle level office, skilled, semi- and unskilled manual workers find extrinsic work values more important. Additionally, semi- and unskilled manual workers have a significantly lower score on intrinsic job preferences as compared to managers/employers, and other workers in service sector jobs. Furthermore, workers who have much autonomy in their work score lower on extrinsic work orientations then workers with lower levels of autonomy, while they find intrinsic job preference orientations more important.
| . | Extrinsic job preferences . | Intrinsic job preferences . | ||||||
|---|---|---|---|---|---|---|---|---|
| . | M1 . | M2 . | M1 . | M2 . | ||||
| . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . |
| Socialization | ||||||||
| Low education (ref.) | ||||||||
| Medium education | −0.042 | 0.017 | 0.067 | 0.056 | ||||
| High education | −0.057 | 0.013 | 0.161 | 0.057 | ||||
| Economic deprivation | ||||||||
| Income (1–10) | −0.002 | 0.003 | 0.011 | 0.005 | ||||
| Job quality | ||||||||
| Managers/employers (ref.) | ||||||||
| Professional workers | 0.033 | 0.019 | 0.039 | 0.048 | ||||
| Middle level office workers | 0.061 | 0.021 | 0.074 | 0.049 | ||||
| Junior level office workers | 0.036 | 0.024 | −0.041 | 0.058 | ||||
| Skilled manual workers | 0.050 | 0.026 | −0.032 | 0.068 | ||||
| Semi- and unskilled manual | ||||||||
| workers | 0.068 | 0.034 | −0.145 | 0.055 | ||||
| Farmers | 0.026 | 0.029 | −0.099 | 0.077 | ||||
| Members of armed forces | 0.045 | 0.039 | −0.003 | 0.084 | ||||
| Autonomy (1–10) | −0.005 | 0.002 | 0.026 | 0.009 | ||||
| Control variables | ||||||||
| Age (1.8–6.4) | −0.024 | 0.004 | −0.063 | 0.013 | ||||
| Female (0/1) | 0.003 | 0.009 | −0.029 | 0.030 | ||||
| Partner (0/1) | −0.001 | 0.009 | 0.013 | 0.029 | ||||
| Children (0/1) | 0.009 | 0.013 | 0.010 | 0.031 | ||||
| Work ethic (1–5) | −0.021 | 0.009 | −0.073 | 0.017 | ||||
| Chief wage earner (0/1) | 0.014 | 0.009 | 0.021 | 0.024 | ||||
| Constant | −0.165 | 0.049 | −0.001 | 0.086 | 0.519 | 0.096 | 0.653 | 0.123 |
| BIC (Sample size adjusted) | 54,784 | 54,340 | 54,784 | 54,340 | ||||
| . | Extrinsic job preferences . | Intrinsic job preferences . | ||||||
|---|---|---|---|---|---|---|---|---|
| . | M1 . | M2 . | M1 . | M2 . | ||||
| . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . |
| Socialization | ||||||||
| Low education (ref.) | ||||||||
| Medium education | −0.042 | 0.017 | 0.067 | 0.056 | ||||
| High education | −0.057 | 0.013 | 0.161 | 0.057 | ||||
| Economic deprivation | ||||||||
| Income (1–10) | −0.002 | 0.003 | 0.011 | 0.005 | ||||
| Job quality | ||||||||
| Managers/employers (ref.) | ||||||||
| Professional workers | 0.033 | 0.019 | 0.039 | 0.048 | ||||
| Middle level office workers | 0.061 | 0.021 | 0.074 | 0.049 | ||||
| Junior level office workers | 0.036 | 0.024 | −0.041 | 0.058 | ||||
| Skilled manual workers | 0.050 | 0.026 | −0.032 | 0.068 | ||||
| Semi- and unskilled manual | ||||||||
| workers | 0.068 | 0.034 | −0.145 | 0.055 | ||||
| Farmers | 0.026 | 0.029 | −0.099 | 0.077 | ||||
| Members of armed forces | 0.045 | 0.039 | −0.003 | 0.084 | ||||
| Autonomy (1–10) | −0.005 | 0.002 | 0.026 | 0.009 | ||||
| Control variables | ||||||||
| Age (1.8–6.4) | −0.024 | 0.004 | −0.063 | 0.013 | ||||
| Female (0/1) | 0.003 | 0.009 | −0.029 | 0.030 | ||||
| Partner (0/1) | −0.001 | 0.009 | 0.013 | 0.029 | ||||
| Children (0/1) | 0.009 | 0.013 | 0.010 | 0.031 | ||||
| Work ethic (1–5) | −0.021 | 0.009 | −0.073 | 0.017 | ||||
| Chief wage earner (0/1) | 0.014 | 0.009 | 0.021 | 0.024 | ||||
| Constant | −0.165 | 0.049 | −0.001 | 0.086 | 0.519 | 0.096 | 0.653 | 0.123 |
| BIC (Sample size adjusted) | 54,784 | 54,340 | 54,784 | 54,340 | ||||
Data: European Values Study 1999/2000, N=13,211.
Bold: Coefficient is at least twice its standard error.
Bold and Italic: Coefficient is between 1.5 and twice its standard error.
Table 4 depicts the contextual findings. According to our socialization hypothesis at the national level, investments in human capital should be negatively related to extrinsic job preferences, but positively to intrinsic ones. By way of such investments, governments signal that intrinsic motivations are valued highly. This hypothesis can only be partly confirmed. It is indeed inversely related to extrinsic preferences (see Models 3 and 6), but it does not substantially contribute to our understanding of country differences in average intrinsic job preferences.
| . | M3 . | M4 . | M5 . | M6 . | ||||
|---|---|---|---|---|---|---|---|---|
| . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . |
| Extrinsic job preferences | ||||||||
| Socialization | ||||||||
| Human capital investments | −0.109 | 0.035 | −0.085 | 0.042 | ||||
| Economic deprivation | ||||||||
| Risk reducing socio-economic factors | −0.044 | 0.040 | 0.065 | 0.060 | ||||
| Quality of the labour market | ||||||||
| Quality of labour force and jobs | −0.146 | 0.071 | −0.193 | 0.128 | ||||
| Constant | −0.012 | 0.074 | 0.001 | 0.086 | 0.004 | 0.083 | −0.005 | 0.072 |
| Intrinsic job preferences | ||||||||
| Socialization | ||||||||
| Human capital investments | −0.082 | 0.153 | −0.111 | 0.126 | ||||
| Economic deprivation | ||||||||
| Risk reducing socio-economic factors | 0.172 | 0.084 | 0.192 | 0.094 | ||||
| Quality of the labour market | ||||||||
| Quality of labour force and jobs | 0.181 | 0.169 | −0.009 | 0.195 | ||||
| Constant | 0.645 | 0.105 | 0.649 | 0.125 | 0.650 | 0.129 | 0.639 | 0.105 |
| BIC | 54,347 | 54,345 | 54,342 | 54,362 | ||||
| . | M3 . | M4 . | M5 . | M6 . | ||||
|---|---|---|---|---|---|---|---|---|
| . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . | Estimate . | SE . |
| Extrinsic job preferences | ||||||||
| Socialization | ||||||||
| Human capital investments | −0.109 | 0.035 | −0.085 | 0.042 | ||||
| Economic deprivation | ||||||||
| Risk reducing socio-economic factors | −0.044 | 0.040 | 0.065 | 0.060 | ||||
| Quality of the labour market | ||||||||
| Quality of labour force and jobs | −0.146 | 0.071 | −0.193 | 0.128 | ||||
| Constant | −0.012 | 0.074 | 0.001 | 0.086 | 0.004 | 0.083 | −0.005 | 0.072 |
| Intrinsic job preferences | ||||||||
| Socialization | ||||||||
| Human capital investments | −0.082 | 0.153 | −0.111 | 0.126 | ||||
| Economic deprivation | ||||||||
| Risk reducing socio-economic factors | 0.172 | 0.084 | 0.192 | 0.094 | ||||
| Quality of the labour market | ||||||||
| Quality of labour force and jobs | 0.181 | 0.169 | −0.009 | 0.195 | ||||
| Constant | 0.645 | 0.105 | 0.649 | 0.125 | 0.650 | 0.129 | 0.639 | 0.105 |
| BIC | 54,347 | 54,345 | 54,342 | 54,362 | ||||
Data: European Values Study 1999/2000, N=13,211.
Bold and Italic: Coefficient is between 1.5 and twice its standard error.
Bold: Coefficient is at least twice its standard error.
Note: Individual level variables are included, but not presented.
Our economic deprivation hypothesis argued that people are less likely to value extrinsic and more likely to value intrinsic job characteristics to the extent that countries are able to safeguard basic materialistic needs. We find clear support for this hypothesis as far as intrinsic job preferences are concerned, but find no evidence for an influence on extrinsic job preferences (see Models 4 and 6). Apparently, countries’ socioeconomic setup can stimulate intrinsic job preferences, but does not take away a focus on material rewards as expressed in extrinsic job preferences.
We also hypothesized that the general quality of a country's labour market affects job preferences. We do indeed find that the higher the quality of the labour market is in a country (in terms of the proportion of workers employed in professional occupational classes in combination with average national autonomy scores), the less important one finds extrinsic work values (see Models 5 and 6 for extrinsic job preferences). We do not observe a relationship between the quality of countries’ labour market on the one hand and intrinsic job preferences on the other hand. Note that on the individual level, having an upper class service sector job with much autonomy does determine job preferences. Apparently, being surrounded by many fellow countrymen who are exposed to a working environment that offers autonomy, opportunities for personal development and match one's abilities, does not additionally enhance intrinsic job preferences.
7. Conclusions and discussion
To what extent can variation in extrinsic and intrinsic job preferences be attributed to and explained by differences between individuals and differences between countries? To facilitate a valid answer to this question and a contribution to the literature, we performed cross-national scalar invariance tests to assess the extent to which job preferences are measured similarly across nations. Our confirmatory multi group factor analyses showed that for 22 of the 31 countries of the EVS 1999/2000, factor solutions were satisfactory, while showing a satisfactory fit for the scalar invariance model. To further facilitate a valid answer, we applied multiple response instead of single response models, which take into account the positive correlation between the dimensions of job preferences.
Our theoretical approach distinguished individual from national determinants. We applied the ideas of Gallie (2007a) that at the individual level socialization, economic deprivation and the quality of the job affect the preference for certain job characteristics. Above that, we argued that specific national contexts affect the value patterns of the population independent of individual characteristics, assuming that nations can influence their citizens’ preferences for intrinsic job characteristics through socialization, through safeguarding basic material needs, and through stimulating the evolution of a labour market in which high quality jobs prevail. Our general conclusion is that at the individual level our indicators of socialization (educational attainment), economic deprivation (income) and the quality of the job (occupational class and autonomy) indeed positively influence intrinsic job preferences, while educational attainment and autonomy negatively influence extrinsic work values. Moreover, manual workers proved to value extrinsic job characteristics the highest of all occupational classes. At the national level, the socialization mechanism and the quality of the labour market appear to diminish extrinsic job preferences, while safeguarding basic material needs increases intrinsic job preferences.
It is important to discuss the implications of the fact that extrinsic and intrinsic job preferences are generally positively correlated: people who find extrinsic job characteristics important also appreciate intrinsic work characteristics (and vice versa). This is even true when the general valuation of work is accounted for. Consequently, not accounting for the correlation between job preferences might lead to biased effects of individual and national characteristics, and thus invalid conclusions. Our findings show that individual and national level characteristics do not always affect intrinsic and extrinsic preferences in the opposite direction, even though theoretically this would seem plausible. Human capital investments and the quality of the labour market do negatively affect the average country level of extrinsic job preferences, but the positive relationship with intrinsic preferences is absent. It seems that countries can socialize their citizens to find material rewards less important (e.g., Curtis et al. 2001; Kääriäinen and Lehtonen 2006), but this does not automatically mean that their intrinsic values increase. The same holds for the impact of a country's quality of the labour market. If the logic of the variety of capital literature applies (e.g., Hall and Soskice 2001; Hult and Svallfors 2002; Gallie 2007a, b), its expectations are only partly confirmed. The higher the quality of a labour market, the less one adheres to an extrinsic orientation, but intrinsic orientations seem to be unaffected. The opposite holds for national level economic deprivation, a measure that we utilized to indicate the extent to which welfare regimes are able to protect their citizens against economic hardship. Theories on welfare regimes (Esping-Andersen 1990) would predict that the higher the level of protection, the lower a country's score on extrinsic job preferences and the higher on intrinsic job preferences. Yet only the latter prediction is empirically confirmed.12
We encourage future research to reveal more mechanisms how countries shape the job preferences of their populations. Gallie (2003, 2007a) suggested that policies to improve the quality of jobs might affect job preferences. In this study, this idea was translated in ‘outcomes’ of such policies in terms of the proportion of high service sector jobs in the labour force and the countries’ average level of autonomy in jobs. However, goal-oriented policies might lead to more than those objective outcomes. They might affect people's values through the mechanism of socialization as well. We believe this thought deserves more attention in future empirical country-comparative research on intrinsic and extrinsic job preferences.
Footnotes
To be able to validly estimate such a contextual effect, it is important to control for job quality at the individual level. Not doing so introduces the danger that the effects found at the contextual level are actually masked individual effects.
Theories on segmented labour markets argue that the primary sector contains high-quality jobs with good promotion opportunities, high incomes, job security and internal labour markets, whereas the secondary sector contains low-skilled and unstable jobs (Blossfeld and Mayer 1988; Althauser and Kalleberg 1990), and therefore are closely linked to our argumentation on the effects of employment systems. Unfortunately, the EVS does not allow for including measures closely linked to segmentation theories, such as type of contract, firm size, and industry. Previously, we argued that an intrinsically motivated workforce is especially important for countries that depend for a great part on the service sector, so that an intrinsically motivated workforce will stimulate the service sector and a highly developed service sector will stimulate intrinsic job motivations. Both lines of reasoning are plausible, but our cross-sectional data will not be able to separate the two.
We leave out Greece since its data are not based on a random sample, and Northern Ireland because no country-level information is available for sub-regions of the United Kingdom.
As this multi group model regresses the binary factor indicators on continuous latent variables, it uses the MLR-estimator (maximum likelihood estimation with robust standard errors), which produces parameter estimates with standard errors that are robust to non-normality. Moreover, we used the theta instead of the default delta parameterization because the former allows for residual variances of the observed factor indicators.
This is achieved by constraining the thresholds of the binary dependent factor indicators to be equal across countries. These thresholds depict the point at which the transition is made from finding a job characteristic unimportant to finding it important. The less a job characteristic is found to be important in reality – that is, the more difficult the item – the higher the threshold. This in fact means that the model takes account of differences in the item difficulties: it assigns different thresholds to the factor indicators dependent on their difficulties, and it constraints these thresholds to be equal across countries.
As Hungary's average is extremely high as compared to the other countries, we have performed sensitivity analyses, which have shown that Hungary is an influential case.
This Eurostat indicator comprises direct public expenditure for educational institutions, support for students and their families with scholarships and public loans, and transferring public subsidies for educational activities to private firms or non-profit organizations.
Eurostat describes it as social benefits, which consist of transfers, in cash or kind, to households and individuals to relieve them of the burden of a defined set of risks and needs.
For the scarce cases where values were missing (for instance Ukraine and Belarus), we extracted them from the Penn World Tables. The correlation between both GDP measures is 0.93.
According to Eurostat, unemployed persons comprise persons aged 15 to 74 who were without work during the reference week, currently available for work and actively seeking work.
The centring is based on the total sample of countries in our study. Therefore the averages of the national characteristics can deviate a bit from zero.
Alternatively, this finding could be due to a higher level of modernization of countries in which citizens have more post-materialistic values and therefore value work more for intrinsic reasons. In earlier versions we actually included measures of post-materialism for the same reasons, and they proved to have strong predictive power. However, we decided to exclude these measures for the reason that conceptually they come so close to our dependent variable, we were afraid of introducing tautology to our models.
Appendix
| . | Extrinsic job preferences . | Intrinsic job preferences . | Socialization . | Economic deprivation . | Quality of the labour market . |
|---|---|---|---|---|---|
| France | −0.42 | 0.37 | 0.75 | 0.69 | 0.34 |
| United Kingdom | −0.01 | 0.24 | −0.61 | 0.72 | 0.42 |
| Italy | 0.07 | 0.89 | −0.48 | 0.44 | −0.21 |
| Spain | 0.02 | 0.06 | −0.80 | −0.01 | −0.15 |
| Portugal | −0.18 | 0.04 | 0.24 | −0.41 | −0.53 |
| The Netherlands | −0.73 | 0.45 | −0.42 | 0.77 | 1.48 |
| Belgium | −0.46 | 0.26 | 0.82 | 0.71 | 0.81 |
| Sweden | −0.37 | 1.42 | 2.21 | 1.08 | 1.13 |
| Iceland | −0.27 | 0.50 | 0.76 | 2.01 | 0.66 |
| Ireland | 0.04 | 0.84 | −0.67 | 0.71 | 0.08 |
| Estonia | −0.17 | 0.12 | 0.93 | −0.94 | 0.12 |
| Latvia | −0.49 | −0.27 | 0.63 | −0.97 | −0.52 |
| Lithuania | −0.10 | 0.09 | 0.94 | −0.96 | −0.91 |
| Poland | 0.00 | 0.67 | −0.40 | −0.84 | −0.51 |
| Czech Rep. | −0.39 | 0.03 | −1.14 | −0.75 | 0.04 |
| Romania | −0.08 | 1.35 | −1.82 | −1.13 | −1.12 |
| Bulgaria | 0.01 | 0.72 | −0.72 | −0.99 | −0.15 |
| Croatia | 0.26 | 0.49 | −0.91 | −0.81 | −0.58 |
| Malta | 0.12 | 1.03 | −0.79 | −0.49 | 0.18 |
| Luxembourg | −0.08 | 1.07 | −1.44 | 2.54 | 1.02 |
| Ukraine | 0.00 | 0.33 | −0.04 | −0.81 | −0.31 |
| Belarus | −0.40 | −0.06 | −0.04 | −0.56 | −0.72 |
| . | Extrinsic job preferences . | Intrinsic job preferences . | Socialization . | Economic deprivation . | Quality of the labour market . |
|---|---|---|---|---|---|
| France | −0.42 | 0.37 | 0.75 | 0.69 | 0.34 |
| United Kingdom | −0.01 | 0.24 | −0.61 | 0.72 | 0.42 |
| Italy | 0.07 | 0.89 | −0.48 | 0.44 | −0.21 |
| Spain | 0.02 | 0.06 | −0.80 | −0.01 | −0.15 |
| Portugal | −0.18 | 0.04 | 0.24 | −0.41 | −0.53 |
| The Netherlands | −0.73 | 0.45 | −0.42 | 0.77 | 1.48 |
| Belgium | −0.46 | 0.26 | 0.82 | 0.71 | 0.81 |
| Sweden | −0.37 | 1.42 | 2.21 | 1.08 | 1.13 |
| Iceland | −0.27 | 0.50 | 0.76 | 2.01 | 0.66 |
| Ireland | 0.04 | 0.84 | −0.67 | 0.71 | 0.08 |
| Estonia | −0.17 | 0.12 | 0.93 | −0.94 | 0.12 |
| Latvia | −0.49 | −0.27 | 0.63 | −0.97 | −0.52 |
| Lithuania | −0.10 | 0.09 | 0.94 | −0.96 | −0.91 |
| Poland | 0.00 | 0.67 | −0.40 | −0.84 | −0.51 |
| Czech Rep. | −0.39 | 0.03 | −1.14 | −0.75 | 0.04 |
| Romania | −0.08 | 1.35 | −1.82 | −1.13 | −1.12 |
| Bulgaria | 0.01 | 0.72 | −0.72 | −0.99 | −0.15 |
| Croatia | 0.26 | 0.49 | −0.91 | −0.81 | −0.58 |
| Malta | 0.12 | 1.03 | −0.79 | −0.49 | 0.18 |
| Luxembourg | −0.08 | 1.07 | −1.44 | 2.54 | 1.02 |
| Ukraine | 0.00 | 0.33 | −0.04 | −0.81 | −0.31 |
| Belarus | −0.40 | −0.06 | −0.04 | −0.56 | −0.72 |
Data: European Values Study 1999/2000.
References
Maurice Gesthuizen is assistant professor in the Department of Sociology at Radboud University Nijmegen, The Netherlands. His main research interests are in the fields of educational inequality, economic vulnerability, social capital and their interrelationships, as well as in the field of job rewards. He studies these topics in longitudinal and comparative perspective and has recently published in journals such as Acta Politica, Acta Sociologica, the European Sociological Review, Quality and Quantity, Research in Social Stratification and Mobility, Scandinavian Political Studies, the International Journal of Manpower and Work, Employment and Society.
Ellen Verbakel is assistant professor in the Department of Sociology at Tilburg University, The Netherlands. Her research interests include work and family, in particular labour market careers of couples and comparative research on values.