Job Preferences in Comparative Perspective 1989–2015: A Multidimensional Evaluation of Individual and Contextual Influences

This article aims to provide a comparative assessment of work values across countries as well as over time. Differences and similarities in job preferences for eight central value dimensions are examined across nineteen countries between 1989 and 2015, made possible by four survey rounds from the International Social Survey, Work Orientation modules. Analyses of how extrinsic and intrinsic work values are related to individual and contextual factors are guided by contrasting theoretical approaches—modernization theory and a welfare-state institutional perspective. Four main results are reported. First, secure and interesting jobs are the most preferred job qualities, universally important to nearly all employees throughout all survey years. Second, values are markedly stable over time, but vary more across countries. Third, large majorities simultaneously value work autonomy, high income, advancement opportunities, jobs perceived as useful to society or helpful to others, indicating how individuals generally, are both intrinsically and extrinsically oriented toward work, with some gendered differences. Fourth partly in support of welfare-state institutional expectations, work values differ across countries mostly in relation to economic equality rather than economic development, so that both extrinsic and intrinsic work values are more important in more unequal societies.

For most individuals, work is central to their lives, commonly involving half of their waking hours. Forming the basis of both an individual's livelihood and a country's economic development, promoting work quantity has always been a priority in modern welfare states. More recently, the promotion of job quality has also come to the fore. It was included in the European Union strategy in the Lisbon Agenda 2000, in both its capacity to endorse comparative advantages in increasingly global markets, and its contribution to individual well-being (Eurofound 2012;European Commission 2008). Work qualities span extrinsic and intrinsic aspects, the former pertaining to income, security, and status, and the latter relating to the subjective perception of whether work is e.g., inherently interesting, useful to society, or helpful to others. Also, the arrangement of work contains central qualities relating to aspects of job autonomy, such as opportunities to work independently and control the organization of work or working times (Borg 1990;Davoine, Erhel, and Guergoat 2008;Handel 2005;Kalleberg 1977;Mottaz 1985;Ros, Schwartz, and Surkiss 1999).
Although working (wo)man has frequently been assumed a homo economicus, principally motivated by economic incentives, much research in the social sciences finds substantial variation in work-related values related to the capacity of work to offer social integration, structure to daily life, a creative outlet, a sense of meaningfulness, and a contribution to society (e.g., Jahoda 1982). Comparative studies of work commitment find great variation not only among individuals within countries but also across countries with regard to individuals' motivation to work, measured on an extrinsic-intrinsic continuum (Esser 2005b;Gallie and Alm 2000;Hult and Svallfors 2002;Svallfors, Halvorsen, and Andersen, 2001). In parallel, research also finds strong correlations between work qualities and higher levels of job satisfaction (Kalleberg and Griffin 1978;Kalleberg 1977;Pichler and Wallace 2009;Westover 2012), life satisfaction (Drobnič, Beham, and Präg, 2010); physical health (Aronsson and Blom 2010), and well-being (e.g., Esser 2017;Eurofound 2012). In a Swedish study, being in one's most preferred job was found to be the most beneficial labor market factor for long-term health (Aronsson and Blom 2010). In sum, individual job preferences appear to be of central importance both for theories of work motivation and theories of personal well-being. Comparative studies, however, are scarce. Findings from a comparison of job preferences across five countries in 1996 and 2001 reported notable cross-country variation in values, although these could not be identified as types of either production or welfare regimes, but rather to differences in workforce composition, where differences in job quality play an important role for intrinsic work values (Gallie 2007b). To date, however, no broader comparative assessment has been made of how job preferences vary across several dimensions and over a longer period.
The overarching purpose of this article is to gain knowledge of how job preferences have developed over time in a relatively large number of "western" countries. We do this by analyzing how job preferences vary for eight central job-quality dimensions, across 19 countries, and at four points in time between 1989 and 2015. The broad scope of this study, for both the number of countries and the period studied, is made possible by combining four waves of the International Social Survey (ISSP), Work Orientation modules of 1989, 1997, 2005, and 2015. These modules include questions on job preferences spanning extrinsic and intrinsic job qualities, and several theoretically interesting socioeconomic and demographic characteristics (gender, age, education, and labor market status) expected to be relevant to explaining variation in work values.

JOB PREFERENCES IN COMPARATIVE PERSPECTIVE
Moreover, analyses are guided by potentially substantial contextual variation in job preferences. This notion draws on two main perspectives that, in contrasting ways, explain variation over time and across countries. Modernization theory seeks to explain valuation change primarily in relation to societies' economic development, increasing levels of education, and female labor force participation (Inglehart 1990;Inglehart and Abramson 1994). A welfarestate institutionally centered perspective attaches more importance to the normative importance of the economic distribution within countries, and to the different ways in which welfare states provide social protection, also relating to the regulation of labor markets (Bowles and Gintis 2000;Gallie 2007b; Korpi and Palme 1998). So, a second aim of this study is to examine how variations in job preferences across countries and time may be associated with these central contextual factors. For this purpose, the survey data are combined with country-level data that allow for simultaneous assessment of individual-and country-level factors in multilevel analyses, which is appropriate, given the nested structure of the data, that is, individuals within countries (over time).

JOB PREFERENCES WITHIN AND ACROSS CONTEXTS: THEORY AND PREVIOUS RESEARCH
Following a comparative cases approach, 19 high-income, western, long-standing democracies are included in this study. Work valuations in these countries during the period studied can be expected to share universal and common components, not least in relation to how remuneration from work ultimately ensures subsistence. Nevertheless, the extent to which work valuations are extrinsically or intrinsically oriented may vary across country-specific contexts. Here, it is fruitful to appreciate the complexity and variety in individuals' work experiences, which motivate a multidimensional approach to understanding work values that may simultaneously span both extrinsic and intrinsic domains. The extrinsic domain includes, for example, income, security, prestige, status, respect, acceptance, and power (i.e., external benefits of the job), and the intrinsic domain entails, for example, taking pride in one's work, feelings of accomplishment, self-realization, happiness, self-respect, social identity, and a sense of contribution to society (i.e., job-inherent qualities). An important dimension of work also includes the practical arrangement of work relating to job autonomy, which includes, for example, control over task performance and organization of work and working times (see, e.g., Davoine et al. 2008 for an overview of related job-quality dimensions). The control over working times may be specifically relevant to individuals whose time is shared more equally between work and caretaking responsibilities (Lyness et al. 2012). The relative importance of extrinsic or intrinsic valuations has long been debated. In contrast to economists, who have often emphasized the economic incentive of work, early classical sociologists shared a view of the potential of work to contain fundamental intrinsic capacities as a "self-determined productive activity," and thus a necessary and positive good in human lives (Marx), conducive to human dignity (Weber), or functioning as a connection between people in mutually dependent "organic" relationships (Durkheim). In this understanding, work holds the potential to bring out self-realizing social activities in human nature. However, pessimism about the value of work has also been expressed in view of the development of the content and purpose of work with increasing industrialization and specialization, argued to result in alienation (Marx), (over-)rationalization (Weber), and (dis-)integration (Durkheim).
At the individual level, explanations of work-value formation have been linked to relatively general social processes relating to early socialization, the extent of economic deprivation, and the significance of the work environment itself (e.g., Gallie 2007a). First, long-term socialization processes can be understood to mold preferences toward an intrinsic and autonomous orientation in order to promote self-realization and initiative at work, representing a continuation of early socialization focused on enhancing personal autonomy and self-development (Argyris 1964). Individual differences may, from this perspective reflect the varying extent of individuals' experiences within the educational system, leading to expectations of stronger intrinsic values with longer/higher education (Inglehart 1977). In addition, early gendered socialization can be expected to influence the development of subsequent gendered work values (Alwin, Braun, and Scott 1992). In contexts more extensively marked by traditional division of gender roles, women could be expected to place greater emphasis on intrinsic jobs values, whereas men's work values would be more extrinsically oriented (Betz and O'Connell 1989). This expectation may also be related to how labor markets are markedly gendered, with more women employed in the (public) care services promoting what is known as care-rational motivations (Waerness 1984(Waerness , 2003, which would be reflected in stronger intrinsic valuations among women. Second, from another perspective, it has been proposed that job preferences will vary in relation to a hierarchy of human needs (Inglehart 1977;Maslow 1954). In this view, extrinsic values relating to income and security are regarded as more fundamental. Once such needs are met, for example, with decreasing economic pressure, values may shift toward (higher order) intrinsic self-realisation values. Arguably, such changes in values also relate to countries' opportunity structures, as provided by the level of economic development, but also to the extent that labor market organization and welfare-state institutions promote full employment on (more) gender equal terms and provide basic economic security. More accentuated gendered work values could also be expected in the age group prone to family formation involving the presence of small children in the household, thus introducing a life-course component in the analytical framework of this study (Lyness et al. 2012).
Third, the formation of job values may be influenced by the quality of work itself. Higher-quality jobs that, for example, offer variety, task identity, and autonomy are more often perceived as meaningful and evoking stronger internal motivation (Hackman and Oldham 1976), as supported by longitudinal research (Lindsay and Knox 1984;Mortimer and Lorence 1979). Comparative research indicates substantial cross-country differences in job quality, where more coordinated employment relations, especially the presence of stronger unions, appear to promote higher job quality, for example, more autonomy (Edlund and Grönlund 2008;Esser and Olsen 2012) or flexibility in working hours (Esser and Olsen 2018;Lyness et al. 2012). In sum, there is a theoretical and empirical rationale for considering the direct effects of the compositional differences of gender, education, and age.
At the individual level the importance of job quality may be assessed in relation to individual labor market positions. Predominantly lower extrinsic, as well as more limited intrinsic job qualities, are generally found in more precarious jobs that provide a lower degree of attachment to the labor market, such as part-time jobs, where women are typically overrepresented (Clark 2005a(Clark , 2005bOlsen 2012, 2018;Greenan, Kalugina, and Walkowiak 2014;Kalleberg 2009Kalleberg , 2011Loughlin and Murray 2013;Stier and Yaish 2014). Assuming decreasing extrinsic qualities with more precarious or distant labor market positions, stronger extrinsic and intrinsic values would be expected among full-time workers, but to a lesser extent among part-time workers and weaker among unemployed persons or persons engaged mainly in (unpaid) domestic work. Such expectations may by confounded in relation to the degree of (voluntary) self-selection into these positions, in which case intrinsic motivations in particular may be more predominant in, for example, more care-rational occupations and possibly involve more job autonomy. Data restrictions, however, limit further elaboration of these considerations in this study.

Country Contexts and Preferences
In addition to individual variation in job preferences, a large comparative literature has addressed contextual variation in postmaterialist values across countries. According to modernization theory, economic development is considered the main driver behind shifting valuations toward more "postmaterialist" concerns in increasingly postindustrial or "welfare" societies (see, e.g., Inglehart 1977;Inglehart and Abramson 1994;Inglehart and Baker 2000). In addition, rising levels of income and education, as well as increased occupational specialization, are expected to elicit consequences such as, for example, changing gender roles, attitudes toward authorities, and more. In line with such changes, values will shift away from economic and physical security toward a greater emphasis on freedom, self-expression, wellbeing, trust, and tolerance (Inglehart and Baker 2000:20, 49). It is less studied whether, and if so how, these changes also include a shift in work values, because work values are not specifically included in the modernization survey indexes commonly used. However, it has been suggested that the proposed value change from materialist to postmaterialist concerns implies a move away from more traditional work values as inherited from the Weberian Protestant ethic and instead toward a more individualized appreciation of work and its capacity for self-expression (Gundelach 2002:145-46), that is, increasing levels of concern about intrinsic valuations. Consequently, this perspective leads us to expect higher levels of intrinsic job preferences in countries with higher levels of economic development and higher rates of female labor force participation. They can also be more prevalent among individuals with higher levels of education or specialization in their work. Moreover, and importantly, we should expect changes over time in tandem with increases in economic output, educational levels, and female labor force participation in most countries, although with some restrictions related to the present comparison of high-income welfare states.
The empirical research concerned with evaluating change in postmodern values has drawn mainly upon materialist/postmaterialist indexes constructed in the early 1970s, later incorporated in the World Values Survey, which today covers more than 80 countries in repetitive rounds, approximately every 10 years. This research finds substantive value changes in line with expectations, toward more individualistic values related to self-expression in various forms in more affluent countries, although these comparisons are usually broader, comparing countries from all parts of the world and at all levels of economic development (e.g., Inglehart 1977;Inglehart and Abramson 1994;Inglehart and Baker 2000). However, findings from a comparison of 20 relatively rich economies in 2000, which draws on a separate survey item asking to what extent work is valued as a duty toward society, that is, a more collective commitment and more traditional Protestant work ethic, or at the other end of the continuum, a more individualized orientation, revealed contrasting findings. A stronger sense of duty was found in welfare states providing more generous income-related unemployment protection, also demanding longer periods of contribution (i.e., participation in work) to qualify for unemployment benefits. At the same time, a weaker sense of duty was seen in countries with an exceedingly long period for provision of unemployment benefits (Esser 2005a). In this way, more generous social protection may still, in important ways, align with strong, traditional work ethics, especially when combined with pronounced requirements of participation to qualify for benefits.
Although modernization theory leads us to expect shifting work values in "welfare" societies, with a primary focus on the level of economic development, this perspective largely disregards how welfare states are associated with a substantially different distribution of economic incomes or with distinct configurations of social protection against common social risks such as unemployment and sickness, or an economic buffer through times of family formation. To this end, we propose a welfare-state institutional approach based on an understanding of significant differences between welfare-state institutions and how these may effectively explain the relative importance of extrinsic or intrinsic work values in the postindustrial countries compared here (following, e.g., Edlund and Grönlund 2008;Esser 2005b;Gallie 2007b). To the extent that more encompassing (generous) welfare states provide more extensive social protection, a shift in emphasis from extrinsic to intrinsic work values can be expected. Supportive of this argument is also the understanding of social protection as collective resources that benefit not only those in direct need of social protection, but all individuals who are qualified to access (extensive) social protection if need be. In this way, all individuals benefit from a sense of security that lowers the potential stress around temporary loss of income, either in relation to unemployment (Sjöberg 2010) or sickness (Esser 2017).
Previous broader comparative research on variation in job preference is rather limited. When five European countries were compared, higher levels of intrinsic preferences were found in Scandinavian countries, whereas extrinsic preferences were stronger in Britain and Germany. This was consistent with arguments relating to both welfare and quality-of-work policies, but contrary to expectations relying on "production systems" perspectives (Gallie 2007b). Nevertheless, a study of British job preferences (1992British job preferences ( -2006 found intrinsic job preferences to be increasing, a development associated with rising levels of education, improvement in jobs with respect to skill, learning opportunities, and employee involvement, and higher incomes and security (Gallie, Felstead, and Green 2012). Studies with a specific focus on job security, as a particularly important job quality for most people, indicate that such values are quite widespread and relatively stable over time (Clark 2005b;Esser and Olsen 2018;Gallie 2007b;Kalleberg 2009). Related comparative research on employment commitment along an intrinsic-extrinsic continuum found substantial differences between countries (Berglund 2001;Esser 2005b;Hult and Svallfors 2002), also when unemployed people were compared across countries (Gallie and Alm 2000), generally lending evidence for the existence of stronger intrinsic values in more encompassing welfare states and more regulated labor markets (Esser 2005b). Yet, existing comparative research has been based on data for specific points in time or for a relatively short time-period.
In sum, these theoretical viewpoints warrant evaluation of the role of contextual factors for variations in work valuations across countries and over time. While modernization suggests higher explanatory value for economic development, educational level, and female labor force participation, welfare-state and labor-market institutional perspectives center instead on the economic distribution within countries, labor-market structures, and the organization of social protection, which are also related to female labor force participation. Motivated by theory and previous research, three working hypotheses can be formulated about the expected associations between contextual factors and extrinsic as well as intrinsic work values. Here, arguments relating to intrinsic values are also extended to valuations of job autonomy, insofar as these are expected to correlate negatively with extrinsic values (see the discussion above). Thus: . in line with modernization theory, stronger intrinsic values and weaker extrinsic values are expected in contexts of higher levels of economic development; . in line with both modernization theory and an institutional perspective, stronger intrinsic job preferences and weaker extrinsic preferences are expected in contexts of higher levels of female labor force participation; . in line with an institutional perspective, stronger intrinsic job preferences and weaker extrinsic job preferences are expected in contexts of lower income inequality, more generous welfare provision, and gender-equal family policies, as well as more regulated labor markets. Notably, the study design, including 19 western countries at four points in time between 1989 and 2015, facilitates the exploration of these working hypotheses from both crosssectional (differences between countries) and longitudinal (change over time) perspectives.

Data and Variables
Comparative survey data on job preferences draw on four rounds of the International Social Survey Program (ISSP), Work Orientations modules of 1989, 1997, 2005, and 2015. The subsample includes working-age (18-59 years) individuals who work (full-or part-time), are unemployed, or are engaged mainly in (unpaid) home work. In total, 43,179 individuals across 19 countries are compared. Countries are unevenly represented in the four rounds. 1

Measures of Job Preferences
The eight indicators of job preferences were phrased as statements: "For you personally, how important do you think each of the following would be if you were choosing a job?" Answers were available on a scale of 1-5, reflecting the degree of importance-"not important at all," "not important," "neither important nor unimportant," "important," and "very important"-in relation to having a(n) 'secure job; 'high income'; 'good opportunities for advancement'; 'interesting job'; 'job that allows someone to help other people'; 'job that is useful to society'; 'job that allows someone to work independently' and 'job that allows someone to decide their times or days of work'.

Individual-Level Variables
Individual characteristics include gender (binary distinction between men and women) and age, contrasting four age groups (18-24, 25-34, 35-44, and 45-59) to capture mainly life-course related job preferences. Educational level distinguishes between those with and without (any level of) tertiary education. 2 Labor market status separates employees working full-time (�30 hours/week), part-time (<30 hours/week), unemployed people, and (unpaid) "home workers."

Contextual (Macro-Level) Characteristics
The individual-level survey data are combined with country-level data for the multivariate analyses. Guiding the empirical assessment of welfare-state characteristics and associated societal outcomes, this study takes an institutional approach rather than a regime approach. This permits evaluation of how more specific contextual aspects may be associated with job preferences, rather than evaluating broad "regime-packages," which may conceal important country differences. By including continuous measures of specific and contrasting contextual dimensions, we hope to reach a more informed understanding of which factors are most relevant in shaping work valuations (cf. discussion in Bergqvist, Yngwe, and Lundberg 2013).
Measures of the contextual variables are shown in Table 1. For enhanced overview, countries have been grouped in four geographical clusters that, to some extent, reflect distinct types of welfare states (cf Arts and Gelissen 2002;Esping-Andersen 1990;Korpi and Palme 1998), which also, to some degree, correspond to corporatist arrangements in the labor market (Jahn 2016), or so-called employment regimes (Gallie 2007a). Although contextual measures certainly exhibit several clustered commonalities, there are notable within-cluster heterogeneity and cross-cluster commonalities, which add empirical rationales for the chosen analytical strategy of drawing on specific and continuous contextual measures (see above). Over time, overall contextual measures reveal relative stability or an increasing trend, in a few instances even dramatic increases, especially in female labor force participation and family policies designed to encourage more equal sharing of paid and unpaid work through dual-earner/dual-carer policies. 3 Corporatist arrangement is the exception, where all Nordic countries, except Norway, saw decreasing levels of corporatism. Substantial decreases also occurred in Belgium and Spain. Slighter decreases occurred in Austria and Switzerland as well, although not fundamentally altering their respective very high and relatively low degree of organization. Although nearly all Anglophone countries saw major increases, these countries still display negative measures, with Ireland as the only exception, where the corporatism index measure jumped to a relatively high level (for a more thorough discussion of developments in corporatism, see Jahn 2016).
Turning to the magnitudes of contextual measures, first, countries' levels of economic development do not in any obvious way reflect the four geographical country groups, although levels are distinctly lower in southern European countries. Norway and Switzerland stand out as the richest countries. In contrast, the measure of income equality demonstrates a relatively clustered pattern. Inequalities are lowest in the Nordic countries (below 27 in all countries by 2015). Western continental European countries display average or slightly below average levels of inequality, whereas inequalities are greatest in southern Europe and the Anglo-Saxon countries, especially Portugal, the United Kingdom, and the United States. Inequality in most countries tends to increase somewhat over time, except in Finland, France, and Portugal. To some extent in these cases, this may relate to the proximity in time between the included survey years.
Also, somewhat clustered patterns are seen in relation to welfare-state (social insurance) generosity, family policies, and female labor force participation. The Nordic countries typically combine high levels of female labor force participation with higher measures in the two  Note: "First" refers to the year each country was first surveyed, "D to Last" is the change between the first and last survey years of each country, and "Mean" is the average across survey years of each country. All data refer to each survey year or nearest preceding year for which data are available. For 2015, data on social protection generosity, family policies, and labor market organization refer to the latest data available in 2010. GDP per capita is in '000 ppp$ constant 2010 prices; Female labor force participation (FLFP) refers to population ages 18-64; social-protection generosity is the average percentage of both replacement and coverage rate of unemployment insurance benefits and sickness cash benefits (of a production worker wage net of taxes, average for single and couple family households, in 1985, 1995, 2005, and 2010); Labor market organization is an additive index of the importance of collective actors, work council representation, and collective bargaining, where higher scores represent greater degrees of labor market organization/corporatist arrangement (Jahn 2016). dimensions of social policy, but levels in Denmark are somewhat lower. Although female labor force participation is also relatively high in the Anglo-Saxon countries, with Ireland as a laggard, levels of public-policy effort are generally low, with social protection more open to market solutions. Two exceptions include Canada, where levels of public-policy effort are distinctly higher in both policy dimensions, and Ireland with a relatively high level of welfare-state generosity. 4 Western continental and southern European countries tend to display intermediate levels in both policy dimensions, although with some notable exceptions. Welfare-state generosity in Italy is distinctly lower than in Spain and Portugal. In Austria and Germany, the orientation of family policies has taken leaps toward more egalitarian sharing of paid and unpaid work. With the exception of these two countries, southern and western European countries' family policies still typically encourage a more traditional division of labor (Ferrarini 2006). This overview suggests considerable cross-sectional as well as longitudinal variation in contextual factors, which prompts the central question of whether, and if so how, these macro-level characteristics may account for differences in job preferences, both across countries and over time.

RESULTS
First, descriptive results are presented, followed by results from multivariate analyses. To start, Table 2 shows descriptive statistics for work values in the eight dimensions examined. The proportions in the first row sum the proportions of individuals who attribute significant importance in each value dimension, either as very important or important. These summed proportions convey a central understanding. Large majorities simultaneously express preferences in several job-quality dimensions, above all having a secure and interesting job-that is, both extrinsic and intrinsic qualities. Only a few regard any of these job qualities as unimportant or very unimportant (on average 5 percent and 1 percent, respectively). Also, the mid-response -- alternative "neither/nor" is chosen by relatively few of the respondents, on average 16 percent, with the most (26 percent) related to control over work time. The two most universally valued job qualities-having a secure and interesting job-are preferred by 94-95 percent of all individuals. Moreover, most regarded these job qualities (57 percent and 53 percent, respectively) as "very important," which contrasts with valuations in the other six dimensions, where the most common response is "important." These value dimensions also vary the least across individuals (as indicated by the standard deviation). The third most preferred job quality relates to work autonomy, that is, being able to work independently. On average, this is valued by four out of five respondents (80 percent), where every third respondent considers it to be "very important." Next, four job qualities are valued by equally large majorities (71-75 percent), which include two extrinsic and two intrinsic job qualities: high income and advancement opportunities as well as jobs perceived as useful to society and helpful to others. Roughly half of all respondents regard these aspects as "important," whereas approximately every fourth or fifth respondent (19-25 percent) regards them as "very important." Finally, the value dimension preferred by the smallest majority (on average 59 percent) relates to control over one's working time. Nevertheless, every fifth respondent assigned high importance to control over working times, although individual differences are the largest in this value dimension, as indicated by the standard deviation. Overall, these findings clearly convey how people's work orientations are composite, so that most people, regardless of their country of residence, strongly value both extrinsic and intrinsic aspects of work. 5 These general descriptive statistics, however, may conceal interesting country-specific differences and over-time variation, which are described next.
For subsequent analyses, explorative factor analysis (principal component analysis) was applied to the data to identify potential underlying valuation factors, which could efficiently reduce the number of value dimensions for analyses. Initial analysis revealed that the indicator for "interesting job" did not load exclusively on one factor when countries were analyzed separately. For these reasons, this item is kept separate, facilitating a factor solution that is comparable across countries. Remaining items were grouped into three theoretically substantiated factors: (1) extrinsic values (job security, high income, and advancement opportunities), (2) intrinsic values (help others and useful to society), and preferences for (3) job autonomy (independent work and control over one's work time). The results from factor analyses based on the whole sample are presented in the Appendix (Table A1, which is also the solution with acceptable robustness across countries as well as gender).
Three additive indexes were constructed, one for each latent factor/dimension. Figures 1A-B shows the averages across countries and time for the three value dimensions/indexes and the single item relating to preferences for an interesting job. Figure 1A shows results relating to extrinsic (triangular markers) and intrinsic (square markers) values, while Figure 1B presents valuations of job autonomy (round markers) and the importance of having an interesting job (diamond markers). Lines between markers indicate changes over time (between adjacent survey years, or else separate markers). Countries are again grouped geographically along the x-axis (similar to Table 1, alphabetically within groups).
An overview of Figure 1A-B shows how the strongest valuations (on average 4.47) concern having an interesting job. This is a consequence of how the equally strong valuations of job security are included within the extrinsic index. Nevertheless, in relative terms, valuations of an interesting job are higher, on average in all but two countries, than the second most preferred value dimension, extrinsic values (on average 4.08). Only in Spain do extrinsic preferences  ISSP 1989ISSP , 1997ISSP , 2005ISSP , 2015. Note: Extrinsic index includes items "secure job," "high income," and "advancement opportunities." Intrinsic index includes "useful to society" and "helpful to others." Job autonomy includes ability to "work independently" and "control over work time and hours." Having an interesting job is a single item. supersede having an interesting job. In Portugal, these values are preferred equally. All in all, the overall universality and stability in the preference for having an interesting job is striking. Only the Dutch are somewhat less concerned with having an interesting job (in the three surveyed years). 6 Turning to a comparison of the three composite indexes, the general pattern indicates how extrinsic factors are more important in most countries than both job autonomy and intrinsic values (highly contingent on job security being included in this index). Cross-country variation, however, is substantial, especially in relation to extrinsic preferences. These are less important in Nordic and western European countries, whereas they are of greater importance in the Anglo-Saxon countries, and especially in the three southern European countries. The particularly low average of Danish respondents' extrinsic values relates to their lower concern with job security. 7 Somewhat unexpected, from both a modernization theoretical and institutional perspective, is the cross-national pattern of intrinsic preferences, which to some extent parallels that of extrinsic preferences, although with less variation across countries. Also, British values throughout the period are more like those in northern and western Europe. The importance of extrinsic values overall is that they are notably stable over time, although with pronounced increases in Spain and the United States. However, intrinsic values have increased over time in several countries, especially in Belgium, Switzerland, Germany, New Zealand, and the United States. 8 Exceptions include decreasing intrinsic valuations in Denmark and, to some extent, in Australia.
Valuations of job autonomy ( Figure 1B) show less variation across countries that have also become somewhat more similar over time given the increased importance of job autonomy, especially in Spain, Portugal, and New Zealand. In Norway, too, valuations of autonomy increased in the 1990s, while they are overall least important in France and, in later survey years, Australia, the United States (which saw decreases over time), and the UK (continuous low valuations). By 2015, however, it becomes difficult to discern any obvious pattern related to geographical groupings.
In sum, Figure 1 conveys an overall picture of relative stability in job preferences over time in most countries. Instead, it is across countries that work valuations differ more markedly. To explore how these differences may relate to both sociodemographic and contextual factors, results from multivariate analyses are presented next. Given the universality and stability over time in preferences for having an interesting job, restricting the amount of variation to be explained by contextual characteristics at the macro level, this item is excluded from the multivariate analyses below.

Multivariate Analyses
The multivariate analyses are based on multilevel modeling, which is the appropriate statistical technique when data are nested in a hierarchical structure. Three data levels are distinguished: individuals ¼ Level 1; year-country ¼ Level 2; and country ¼ Level 3. This three-level design makes it possible to distinguish differences between countries (cross-sectional effects) from variation over time within countries (longitudinal effects).
The results presented in Table 3 show associations between sociodemographic factors and job preferences. Estimates are unstandardized regression coefficients with associated standard    ISSP 1989ISSP , 1997ISSP , 2005ISSP , 2015 errors, and represent the average (fixed) effect across countries and country-years. 9 All interpretations, however, focus on broader patterns rather than exact numerical estimates. Estimated models also provide intraclass coefficients (ICC), shown in the lower section of the table. These indicate the proportion of variation located at each level as compared with the total variation (at all levels). 10 An overview of contextual variation largely confirms the above conclusions, how values generally vary decidedly less over time (ICC-country*year) than across countries (ICC-country). Across-country variation is the largest in the extrinsic dimension (Models A0-A1), somewhat lower in intrinsic values (Models B0-B1), and least with regard to job autonomy (Models C0-C1).
To start, extrinsic values are weaker among people with higher education, who instead have stronger preferences for intrinsic values and especially job autonomy than those with lower education, results that are well in line with theoretical expectations. 11 Still, it should be noted that the two items included in the intrinsic factor indicate more collectively oriented intrinsic values (usefulness to society and helpful to others), rather than more individualistic orientations (toward self-realization), which is what modernization theory specifically stipulates should be of increasing importance with higher levels of education.
Next, job preferences are distinctly gendered in relation to intrinsic values. On average, women express stronger intrinsic preferences. Preferences for job autonomy are also gendered, but to a more limited extent. Women tend to be more concerned with job autonomy than men, as may be expected, because women are more often in part-time work and generally take more responsibility for reproductive work in the home (also in the models controlling for labor market status). Notably men's and women's preferences for extrinsic values do not differ. Men and women share extrinsic values in a highly similar fashion. 12 Regarding age-related job preferences, younger people are more extrinsically oriented than older people, in a relatively linear relationship. Inversely, older people (45-59) tend to have stronger intrinsic preferences than younger people, which are weakest among those ages 25-35, which is the age span coinciding with early family formation and more pronounced breadwinner obligations. Preferences for job autonomy (C1) are the strongest among 35-44-year-olds, the age group likely to have school-age children. Analyses do not allow distinguishing between life-cycle dynamics and a generation/cohort component, but our informed interpretation is that these orientations reflect mainly life-cycle-related career trajectories in the labor market. On average, older people are more likely than younger people to have achieved job security, high income, and career advancement in combination with diminished breadwinner obligations, something that might make them less inclined to prioritize (additional) extrinsic job aspects, but rather intrinsic values and job autonomy.
Results relating to labor market status show distinct differences. Part-time employees have weaker extrinsic preferences, but stronger preferences for intrinsic qualities and job autonomy than full-time employees. As discussed above, this may relate to selection, that is, how parttime employees, of which women form the majority, may choose to work less, partly in relation to stronger valuations of other spheres in life, and partly in line with care-rational valuations more prevalent in care-service-oriented (part-time) jobs. Noteworthy is the similarity between unemployed persons' and full-time employees' preferences for intrinsic values and job autonomy. Unemployed people, however, are less concerned with extrinsic values.
Again, this may be related to the extent that selection into unemployment is voluntary or involuntary. Unemployed people's extrinsic preferences may differ before (more voluntary) unemployment or may decrease upon (more involuntarily) becoming unemployed, as part of coping strategies to decrease stress around lower economic security (e.g., Lazarus and Folkman 1984). Home workers' job preferences are weaker in all dimensions than full-time employees' work values, resonating with distinctly lower levels of commitment to paid work among home workers (Esser 2005b).
Finally, concerning the development of job preferences over time, what stands out is the somewhat increasing importance of intrinsic preferences, a finding that fits well with modernization theory. There is, however, no unambiguous inverse effect for extrinsic preferences. Although these preferences are somewhat stronger in 1989, the decrease over time is not continuous for all survey years. In the models relating to job autonomy, valuations are generally stable over time. In only one model, variation over time supersedes cross-country variation (C0, relating to job autonomy), although both ICC-proportions are quite small (2.08 percent and 1.36 percent, respectively). Lastly, a note on the importance of compositional effects on job preference. The highly similar proportions of contextual variation across models without controls (A-C0), as compared with models including controls (A-C1), convey how the inclusion of individual characteristics does not explain much of disclosed cross-country or over-time variation, but rather leaves significant variation to be explained by the inclusion of contextual factors in the next step of analyses.

Accounting for Variation in Preferences at the Contextual Level
A "hybrid model" is used to estimate relationships between preferences and contextual factors using multilevel modeling. This method distinguishes cross-sectional and longitudinal associations by entering each contextual variable into models in two ways. First, the country-level variable (CS in Table 3) indicates the average score of a specific contextual attribute spanning all survey years. Second, the longitudinal country-year variable (LT in Table 3) indicates the differential effect of a country's score for a specific year from that country's overall mean value spanning all survey years (for further discussion, see Fairbrother 2014;Schmidt-Catran and Fairbrother 2015).
A methodological challenge in most comparative studies is the restricted number of contextual units. Given limitations in statistical power, only one or a few contextual variables are thus entered in each model for reasonable accuracy in estimates. Nevertheless, the relative importance of different contextual variables may be evaluated by comparing models based on fixed-effect estimates, random effect variances, and overall model fit. With regard to the latter, the Bayesian information criterion (BIC) weights model fit and parsimony, where lower BIC values suggest more effective representations of the data. Table 4 reports the results from multilevel models when job preferences are regressed on the contextual variables. Each model (shown horizontally in the table) includes also the full set of independent individual-level variables (see Table 3), but are not shown for reasons of parsimony. First, the associations between extrinsic preferences (A) and the level of economic development in Models A2 and A8 are negative, but only significant in Model A2 when the effect of income inequality is not accounted for. The longitudinal estimates for     Table 2), but are not shown in the model. A categorical variable for year is also included to control for common longitudinal trending (not shown).
Source : ISSP 1989: ISSP , 1997: ISSP , 2005: ISSP , 2015 gross domestic product (GDP) per capita are insignificant throughout, suggesting that extrinsic preferences have not changed over time as a function of a country's economic development. All models that include income inequality (A3, A8-10, and A13) indicate a relatively strong cross-sectional association, which is also robust to the inclusion of the other contextual variables. This factor, however, does not explain changes over time. The importance of female labor force participation is assessed in Models A4, A9, A11, and A13. These models consistently show statistically significant negative cross-sectional associations but no longitudinal relationship. Model A5 shows how extrinsic preferences are weaker in countries with more gender-equal family policies, with decreasing importance over time, and also robust to the inclusion of other contextual factors (in Models A10-13). Finally, extrinsic work valuations are unrelated to both the generosity of social protection and the level of labor market organization, although a weak longitudinal association is found with social protection (Model A6), which is no longer statistically significant when also controlling for policies (Model A12).
A comparison of the measures of model fit (BIC) and random variances across models indicates which models and the associated variable combinations fit the data most efficiently. Best model fit, despite its relative simplicity, is Model A3, demonstrating the effect of income inequality on extrinsic valuations. In the case of cross-country variation, inclusion of income inequality in Model A3 as compared to the model that only includes individual-level factors (Model A1), reduces residual variance from 3.81 to 1.53, suggesting that income inequality accounts for as much as 60 percent of country differences in extrinsic preferences. In contrast, the inclusion of several contextual factors in subsequent models does not improve model fit in relation to additional model complexity. While BIC scores of these models are only marginally larger, an informed interpretation of estimate significance and explained variance could be warranted. Although estimate sizes in some models are not trivial, several are quite negligible despite their significance, especially in relation to longitudinal effects. For example, when the family policy indicator is included in Model A5 as compared to model A1, the residual variance at the country-year level is reduced from 0.15 to 0.12, implying how changes in family policy explain about 20 percent of initially quite limited variation over time in extrinsic preferences.
Next, results relating to intrinsic job preferences are seen in Models B2-B11. Models B2 and B8 cover economic affluence (GDP per capita). The cross-sectional estimate is insignificant in both models, suggesting that differences in countries' economic affluence do not explain cross-country variation in intrinsic preferences. However, in line with modernization theory, there is a statistically significant positive longitudinal association between GDP per capita and intrinsic preferences (Model B2), which is also robust to the inclusion of economic inequality (Model B8).
In the effects of income inequality (Models B3 and B8-11), there is a positive crosssectional, but a negative longitudinal, association with intrinsic preferences. Moreover, all models relating to female labor force participation (B4), dual-family policy (B5), socialinsurance generosity (B6), and labor market organization (B7) produce results opposite to those theoretically predicted (i.e., negative as opposed to expected positive associations). However, only significance of female labor force participation is robust to the inclusion of income inequality in models (see Models B9-11). With regard to BIC estimates, similarly to models relating to extrinsic values, best model fit is seen for Model B3, which includes income inequality only at both the cross-sectional and longitudinal levels. Again, estimates at both the cross-sectional and longitudinal levels are significant. However, the opposite signs of these estimates indicate the importance of distinguishing cross-sectional from longitudinal associations in the analysis. In general, one can argue that the longitudinal estimate is more reliable since it controls for any unobserved heterogeneity at the country level (cf Schmidt-Catran 2014). This finding is also in line with the stipulated expectation from a welfare state institutional perspective. However, comparing the random variance components across Models B1 and B3 suggests that Model B3 accounts for about 50 percent of the cross-sectional variation in intrinsic preferences, but only about 15 percent of the longitudinal variation within countries (which is limited to begin with).
Finally, valuations of job autonomy are not associated with economic development (Model C2), but show a negative relationship with economic inequality at both the cross-sectional and longitudinal levels (Model C3), where the longitudinal association remains statistically significant when additional contextual variables are also included (Models C8-10). With regard to family policies (Model C5), social-insurance generosity (Model C6), and labor market organization (Model C7), results show positive cross-sectional associations indicating stronger preferences for job autonomy in more encompassing welfare states, with more gender-equal family policies and more regulated labor markets, although only the effect of labor market organization retains its significance when economic inequality is taken into account (see Models C8-10). However, the marginally larger BIC estimates of these models as compared to the model including only individual-level characteristics (C1 , Table 3), indicate the limited importance of contextual characteristics for explaining the limited crossnational variation in these valuations.
Summarizing these findings, results suggest that contextual factors matter for job preferences, primarily in relation to income inequality. First, as expected from an institutional perspective, in the variation between countries, extrinsic preferences are indeed stronger in more unequal countries, while preferences for job autonomy are somewhat weaker in more unequal contexts. More unexpected is the finding that also intrinsic preferences are stronger in more unequal countries. Second, relating to valuation changes over time, in line with theoretical expectations, intrinsic preferences tend to decrease with growing inequality over time within countries. Notably, the cross-sectional and longitudinal effects of income inequality are not consistent as to their "signs," "sizes," or "statistical significance," making caution necessary in drawing strong conclusions about the generality of these estimated relationships.
In relation to the other contextual factors considered in this study, the overall absence of strong and consistent effects, especially in relation to social protection, is somewhat unexpected. Although some associations are statistically significant, their inclusion accounts for limited cross-country variation as compared to models with only individual-level characteristics included. For example, comparing Models A-C5 to Models A-C1 shows that the inclusion of the family policy measure as the single contextual measure accounts for 11 percent, 14 percent, and 21 percent of the between-country variation in extrinsic, intrinsic, and autonomy valuations respectively. These are arguably relatively small proportions, especially when considering that cross-country variation in valuations of job autonomy is very limited to start with. Moreover, indicators of model fit do not improve for these models, indicating limited explanatory value to compensate for additional model complexity. In relation to previous research that has shown a strong (negative) correspondence between welfare state generosity and income inequality outcomes (Korpi and Palme 1998;Marx, Salanauskaite, and Verbist, 2016), results appear to call into question the validity of social protection indicators. Possible explanations are further discussed in the concluding section.

DISCUSSION AND CONCLUSIONS
Drawing on uniquely broad multidimensional survey data spanning nearly three decades, this article's overarching aim is to explore how job preferences in eight central dimensions vary across countries and over time, between 1989 and 2015. In addition, hierarchical multivariate analyses evaluated the importance of individual and contextual factors for such variation. Guided by two contrasting theoretical perspectives-modernization theory and a welfarestate institutional perspective-different emphasis is accorded to individual and contextual factors for explaining the development of extrinsic and intrinsic job preferences. Related comparative research has demonstrated considerable attitude differences in work commitment across modern welfare states, but comparative studies of job preferences are few. This article thus contributes an improved understanding of individual job preferences relevant to both theories of work motivation, theories of personal well-being, and comparative welfare-state and labor market research concerned with empirical assessments of value formation in macro-level contexts.
Four main findings are reported. First, secure and interesting jobs, that is, both extrinsic and intrinsic valuations, are unambiguously the most universally preferred job qualities, valued as important by 94-95 percent of all employees, with striking similarities across countries and over time, attesting to the central significance of these job qualities. Second, values are overall relatively stable in the period studied, although the importance of intrinsic values has increased somewhat. Rather, it is mainly across countries that work values differ. Although the international survey data provide a unique opportunity to evaluate valuation change during a quarter of a century, this is still a relatively short and limited period in the history of working (wo)man. All of the countries compared have already taken major steps toward being service-and knowledge-intensive economies, where relatively small proportions of employees are involved in manual industrial or agricultural labor. From this perspective, it is notable how large proportions of all employees, independent of skill, express strong job preferences across several extrinsic and intrinsic value dimensions. Restricted availability of data on occupations in the earliest survey unfortunately precluded closer evaluation of the occupational gradient in work valuations.
Third, large majorities (every third or fourth respondent) also express appreciation for several other intrinsic and extrinsic dimensions-for work autonomy, jobs with high income, advancement opportunities, jobs perceived as useful to society or helpful to others. These findings accentuate the importance of acknowledging the multidimensionality of work values, their complexity, and the way most individuals are simultaneously intrinsically and extrinsically oriented toward work. In other words, most people value several aspects of work, differing rather in their relative importance, and possibly timing in their combinations. The latter is supported by the greater importance of extrinsic values and job autonomy JOB PREFERENCES IN COMPARATIVE PERSPECTIVE (control over work organization and work times) among individuals in ages most prone to family formation and child-rearing. Despite such commonalities, values were found to be clearly gendered and related to education; women are more intrinsically oriented and more concerned with job autonomy, whereas people with higher education express less concern about extrinsic values, but more concern with intrinsic values and job autonomy. Overall, these micro-level findings are in line with expectations of modernization theory, as well as theories on early socialization and prolonged socialization through extended education systems. Gendered values also support arguments raised from a care-rational reasoning, to the extent that women are overrepresented in the care-service sector, which bears further assessment in occupational specific analyses. To the extent that education is also strongly correlated with income, the idea of hierarchical needs, proposed by Maslow, receives some support in so that people with higher education are less extrinsically oriented toward work.
Fourth, multivariate analyses attributed most explanatory power of cross-country variation in work values to income inequality. Extrinsic as well as intrinsic work values are more important in more unequal societies, which in part supports expectations of the welfare-state institutional perspective. In support of modernization theoretical arguments, increasingly intrinsic valuations over time are, to a limited extent, related to increasing economic affluence. In relation to the somewhat unexpected finding of stronger intrinsic valuations in countries with more unequal income distributions, it must be noted how the intrinsic value factor is based on two distinct, collectively oriented intrinsic survey items: jobs valued as useful to society and those valued as helpful to others. In other words, they do not capture the individualistic self-realization value orientation emphasized by modernization theory. Notably, no positive longitudinal associations were found between income inequality and job preferences, but in fact a weak negative relationship in the case of intrinsic preferences. This points to the need for more research before more detailed conclusions can be drawn concerning the relationship between economic inequality and job preferences.
The analysis consistently supported neither of the two theoretical perspectives. Modernization theory predicts more substantial valuation changes over time in tandem with growing economic development and female labor force participation. Instead, job preferences are remarkably stable over the time period examined, where the limited temporal change observed is only explained to a minor extent by the contextual characteristics considered in this study. From a welfare-state institutionalist perspective, the limited contextual effects of social policy indicators were unexpected. To the extent that encompassing Nordic welfare states have been found more redistributive, producing lower security inequalities as compared to basic security-oriented welfare states in Anglophone countries (see, e.g., Korpi and Palme 1998), a stronger association between more generous social protection and job preferences would be expected, although indirectly and to the extent that such measures do in fact reduce income inequalities.
This study is not without limitations. Although it addresses eight central value dimensions, still more important dimensions would have improved robustness in the construction of valuation factors, not least to better evaluate the relevance of central theorems of modernization theory. Specifically, this concerns the central importance of individualistic intrinsic dimensions reflecting aspects of self-realization. Another limitation is that the design of the survey questionnaire does not prompt individuals to rank or relate work-value dimensions directly to each other, requiring the indirect inference of the importance of these dimensions. In addition, the potential of hierarchical regression models would obviously be enhanced by including more countries in the analysis, and expanding variation in work values if, for example, postcommunist countries, low-and middle-income countries, or Asian economies were also included. Yet, such an expanded scope also relies on the availability of quality contextual measures.
While the comparative perspective in this study has indicated some importance of contextual factors, especially income equality, for variations in work valuations across countries, it is rather the similarities in valuations across these (rich) countries over the time period studied that stand out-large majorities across all countries are strongly both extrinsically and intrinsically oriented toward work. People generally prefer a variety of qualities in a job, above all secure and interesting jobs. In relation to the current policy pursuit of increased participation and full-time work for extended shares of the labor force, but also attainment of long sustainable working lives, it also seems reasonable to increase concern with a number of central job qualities that are highly valued and may also be expected to avert potentially costly workrelated forms of stress and ill-health. Assessing how social policies and labor market regulations may successfully promote various job qualities that can accommodate widely shared job preferences and be conducive to better work-related health and stronger work motivation, would certainly be fruitful steps for future research.

There is No Disclosure
"No potential conflict of interest was reported by the authors."

ACKNOWLEDGMENTS
For valuable comments the authors wish to thank Markus Hadler and participants in the social policy seminar at the Swedish Institute for Social Research (SOFI), Stockholm University October 13, 2017. This work was in part supported by the Swedish Research Council, grants 2012-5503 and 2013-1724. For updated family policy data we are indebted to Katharina Wesolowski (SOFI). NOTES 5. In part, this may be a consequence of the survey design, which permits respondents to express equally high valuations for all dimensions, and thus offers a possibly more indirect indication of which dimensions are more important, as compared with survey questions that had prompted respondents to rank or choose one or a few value dimensions over others.
6. In contrast to majorities in all other countries, in the Netherlands, a minority values an interesting job as "very important" (39 percent). 7. In Denmark, a minority (40 percent) values a secure job as "very important," in contrast to majorities in the other countries.
8. Closer inspection of the data shows how increases over time are especially prevalent in valuations of control over working times, possibly relating to increasing female labor force participation, which in many countries involves part-time work.
9. For example, Model A1 shows that, on average, across countries and over time, those with higher education have slightly weaker extrinsic preferences than those with lower education (B ¼ −0.10, SE ¼ 0.01).
10. This is calculated as the variation at one level divided by the sum of variation across all levels. For example, in Model A0, the cross-sectional ICC (ICC-Country) is calculated as 4.24/(30.06 þ 4.24 þ0.24) ¼ 12.28 percent.
11. These results pertain to all individuals across all countries, but do not assess the extent to which these differences are prevalent in each country. Closer inspection of the data reveals how education is not a distinguishing factor for extrinsic valuations in the Nordic or Anglophone countries, excluding the United States, or in relation to job autonomy in the Anglophone countries.
12. Country-specific analyses (not in the table) show that only in the Netherlands are women significantly less extrinsically oriented than men.