Unemployment and Subjective Well-being: Comparing Younger and Older Job Seekers

Introduction Unemployment has been shown generally to cause lower levels of self-esteem and self-efficacy as well as to elevate depression, anxiety and stress (Bartley, 1994; Breslin and Mustard, 2003; Briar et al., 1980; Creed, Muller, and Machin, 2001; Eisenberg and Lazarsfeld, 1938). Unemployment research has been conducted for over eight decades and has encompassed a variety of methods, of various geographic places, and a wealth of theories (Fryer, 2002; Mckee-Ryan et al., 2005). One influential model is Jahoha’s (1982) deprivation model in which wellbeing is construed as sustained by a number of a manifest functions such as pay, together with latent functions associated with having a job, in particular time structure, activity, social contact, sharing goals with others, and social status (Cole, 2007; Paul and Batinic, 2010). Wanberg (2012) conducted a review of the research literature from 2000 to 2010, examining the relationship between unemployment and psychological health. She concluded that there is a causation effect of unemployment leading to worse mental health and also partial support for the ‘selection’ hypothesis according to which some people become unemployed due to a lower psychological wellbeing. Wanberg (2012) also noted the role of age in mediating the relationship between unemployment and psychological health although evidence on this is sparse. During the last decade, the understanding of risk factors predicting psychological health has improved markedly. Wanberg (2012) concluded that five risk factors have been identified: work role centrality, coping resources, cognitive appraisal, coping strategies and demographics. Apart from demographics, these variables are all related to individual and psychological competences. However, a substantial body of literature limits investigations of the link between mental health and unemployment to only exploring intra-psychological variables without adequate attention to the historical and societal context (Creed et al., 2001; Eisenberg and Lazarsfeld, 1938; Goldsmith, Veum, and Darity, 1996; Nurmi, Salmela-Aro, and Ruotsalainen, 1994; Schaufeli, 1997). According to critical unemployment studies, investigating the link between unemployment and a deterioration of subjective well-being by only focusing on psychological variables poses a problem in the sense that research thus contributes to an individualization and privatization of the unemployment issue (Fryer and Stambe 2014). In order to take this critique into consideration, we turn to governmentality studies developed by Michel Foucault (Foucault et al. 2007) and his successors Mitchel Dean (2010) and Nicolas Rose (1996). Here, we find perspectives and insights highlighting the need to investigate a phenomenon in its historically and culturally contingent context. Subjectivity is here viewed as inherently intertwined with how citizens are governed in a broad sense and in relation to specific, institutionalized practices (Dean, 2010). In various ORIGINAL ARTICLE


Introduction
Unemployment has been shown generally to cause lower levels of self-esteem and self-efficacy as well as to elevate depression, anxiety and stress (Bartley, 1994;Breslin and Mustard, 2003;Briar et al., 1980;Creed, Muller, and Machin, 2001;Eisenberg and Lazarsfeld, 1938). Unemployment research has been conducted for over eight decades and has encompassed a variety of methods, of various geographic places, and a wealth of theories (Fryer, 2002;Mckee-Ryan et al., 2005). One influential model is Jahoha's (1982) deprivation model in which wellbeing is construed as sustained by a number of a manifest functions such as pay, together with latent functions associated with having a job, in particular time structure, activity, social contact, sharing goals with others, and social status (Cole, 2007;Paul and Batinic, 2010). Wanberg (2012) conducted a review of the research literature from 2000 to 2010, examining the relationship between unemployment and psychological health. She concluded that there is a causation effect of unemployment leading to worse mental health and also partial support for the 'selection' hypothesis according to which some people become unemployed due to a lower psychological wellbeing. Wanberg (2012) also noted the role of age in mediating the relationship between unemployment and psychological health although evidence on this is sparse.
During the last decade, the understanding of risk factors predicting psychological health has improved markedly. Wanberg (2012) concluded that five risk factors have been identified: work role centrality, coping resources, cognitive appraisal, coping strategies and demographics. Apart from demographics, these variables are all related to individual and psychological competences. However, a substantial body of literature limits investigations of the link between mental health and unemployment to only exploring intra-psychological variables without adequate attention to the historical and societal context (Creed et al., 2001;Eisenberg and Lazarsfeld, 1938;Goldsmith, Veum, and Darity, 1996;Nurmi, Salmela-Aro, and Ruotsalainen, 1994;Schaufeli, 1997).
According to critical unemployment studies, investigating the link between unemployment and a deterioration of subjective well-being by only focusing on psychological variables poses a problem in the sense that research thus contributes to an individualization and privatization of the unemployment issue (Fryer and Stambe 2014). In order to take this critique into consideration, we turn to governmentality studies developed by Michel Foucault (Foucault et al. 2007) and his successors Mitchel Dean (2010) and Nicolas Rose (1996). Here, we find perspectives and insights highlighting the need to investigate a phenomenon in its historically and culturally contingent context. Subjectivity is here viewed as inherently intertwined with how citizens are governed in a broad sense and in relation to specific, institutionalized practices (Dean, 2010). In various fields, in medicine, in education, in management and in unemployment, scholars have shown that what it means to be a patient, a student, a manager or an employee, and an unemployed person is greatly affected by the institutional and societal context. From this strand of research, a growing literature investigates the relationship between neoliberal policies and how the unemployed person is increasingly constructed as an active job-seeker rather than a passive recipient of welfare benefits (Boland, 2016;Dean 1995;Fryer and Stambe 2014;Pultz, 2016). Here, self-responsibility plays a particularly important role (Dean 1995;Walters 1994). Also from critical social psychological approaches a general tendency to view social issues as being private and individualised challenges has been identified (Beck, 1992;Willig, 2013). Exploring how unemployed people are governed in a neoliberal era, Fryer and Stampe (2014) note that the privatization and stigmatization of the condition of being unemployed governs not only unemployed people, but in fact the entire population, echoing the key characteristics of biopolitics (Foucault 2008). Representing unemployment as socially undesirable is an active part of valorising work, productivity and self-sufficiency in contemporary societies in which work is intimately connected with identity (Foucault, 2008;Hartmann and Honneth, 2006). Interestingly, Mascini, Achterberg, and Houtman (2013) found that unemployed people tend to view risks such as being poor, unemployed or homeless, in individualized terms, blaming individuals for being in that situation. In comparison, people in employment tended to perceive the same issues in more collective terms, blaming society or other external factors. The authors explain the results by concluding that "the people who suffer most from neoliberalism accept its ideology" (p. 1220).
Overall, there is empirical support suggesting that the neoliberal ideology and the neoliberal policies that have dispersed through most western welfare states since the 1980s -or 1990s -as is the case with Denmark -in various ways affect how unemployment is experienced and that it somehow modulates the relationship between unemployment and subjective well-being (Engelbreth Larsen, 2013;Mik-Meyer and Villadsen, 2013). Neoliberal policies cover the retrenchment in passive labour market measures (paying unemployment benefits) and the introduction of an activation regime making receiving benefits conditional on the job seeker's own effort to get a job (Bengtsson, Frederiksen, and Larsen, 2015;Dean 1995;Immervoll, 2012). We aim here to contribute to developing and employing variables that shed light on this overall hypothesis by exploring whether these neoliberal pressures affects different age groups similarly or not.
Contemporary labour markets are becoming increasingly precarious with organizational downsizing and an increase in nonstandard work arrangements leading to a decline in job security (Glavin and Schieman, 2014;Kalleberg, 2008Kalleberg, , 2009. To the unemployed person this insecurity demands a flexibility and mobility in terms of geography, salary and profession and these abilities are idealised in today's labour markets (Hartmann and Honneth, 2006;Mendenhall et al., 2008). In order to explore flexibility and its link to subjective well-being we investigate whether younger and older people become more flexible in terms of their job search and we look into how flexibility is associated with well-being.
The work of Sharone (2007Sharone ( , 2013 also offers insights into the timely challenges associated with unemployment during and after the Great Recession. Sharone (2013) conceptualizes subjective responses as answers to concrete institutional settings and he identifies self-blame as the dominating vulnerability for unemployed people in the US (see also Newman, 1988). Here, hiring systems emphasize personality and interpersonal skills rather than professional skills, and this finding has to some extent been replicated in a Danish context (Pultz and Hviid, 2016). Self-responsibility is increasingly enacted as a way to govern and control citizens in contemporary welfare states, and in an unemployment context this manifests as self-blame. The logic is that when a person is encouraged to understand him-or herself as a self-responsible and capable agent, it backfires when something in life develops in an undesirable way. In neoliberal times people are encouraged to ascribe responsibility to themselves for whether they succeed or fail. Following prior work of Pultz, Teasdale & Bang (n.d.), we measure the level of self-blame and compare across the two age groups and investigate the link to subjective well-being.
The negative representation of unemployed people has been deemed important in relation to well-being in qualitative research (Bakke, 1933;Sharone, 2013) but to our knowledge this has not been studied in quantitative studies. We begin here to fill in this gap by also developing a scale for unemployment shame and by measuring the level of shame and its relation to well-being.
Related to this issue, on the basis of discursive analyses Gibson (2011) identified the importance of effortfulness when evaluating whether a person deserves receiving unemployment benefit from the state. The meaning of work has also been associated with the neoliberal ideology and studies. To investigate effortfulness, we explore work ethics among the three groups in our study and examine its relationship to subjective well-being. People with higher work ethics tend to suffer more when they become unemployed (Hoorn and Maseland, 2013;Lalive and Stutzer, 2003;McKee-Ryan et al., 2005).
Despite the tendency to reiterate the link between unemployment and deterioration of subjective well-being without addressing the variation and heterogeneity in this group, some studies have sought to portray the complexity of the phenomena and we aim here to contribute to this strand of research. Pultz and Mørch (2015) have recently reported on a group of young people who choose to become unemployed in order to pursue their ambitions in various creative fields and who perceive unemployment benefits as ' entrepreneurial support'. However, it remains yet to be investigated how widespread this practice is and how it relates to subjective well-being. We begin filling in this gap by examining how identification with the condition of unemployment relates to overall well-being but also to matters of self-blame and shame. In addition, we investigate the link to Jahoda's latent functions which, it is claimed, sustain well-being among people with jobs and of which unemployed people are usually deprived. According to a study by Paul, Greithner & Moser (2010), people outside of the workforce, for instance people on maternity or paternity leave, experience some level of deprivation of the latent functions compared to people in jobs, although not as much as unemployed people.
To reiterate, we explore the relationship between the subjective well-being and unemployment by comparing young unemployed people (YU) to an age-matched group of young persons in employment (YE), and a group of older unemployed persons (OU). Overall, inspired by governmentality studies, the aim is to unravel the importance of historically sensitive variables related to neoliberal policies and neoliberal understandings and secondly, to see how these affect two age-groups perhaps in different ways. Inspired by McDonald and Elder (2006) we agree that unemployment changes across the life course and therefore examining unemployment without taking age into consideration blurs some of the variation which we deem key when investigating such complex phenomena. Andersen (2009) studied the influence of social class on the relationship between unemployment and subjective well-being and she concluded that middle-class people experience a bigger negative effect of unemployment as compared to lower and higher classes. Based on this insight we focus our analyses on well-educated young people and thus contribute with empirical evidence about what Standing (2011) has termed 'the academic precariat'.

Danish context
Denmark is a universalist welfare state meaning, inter alia, that all citizens are entitled to unemployment benefit or social assistance when they are unable to support themselves financially (Bengtsson, Frederiksen, and Larsen, 2015). In Esping-Andersen's influential typology of welfare states, he distinguishes between social-democratic, liberal and conservative welfare states. Denmark is grouped among the social-democratic welfare states and is characterized by a high degree of de-commodification meaning the degree to which "a person can maintain a livelihood without reliance on the market" (Esping-Andersen 1990, p. 21). However, recent empirical studies, such as Scruggs & Allan (2007), document that there is less variation between the different types of welfare states today compared to before the onset of neoliberal policies. Accordingly, Dingeldey (2007) notes that it makes more sense to talk about re-commodified states rather than de-commodified as receiving unemployment benefits is made increasingly conditional on living up to a number of demands of activity and availability emphasising duty more than right.
The unemployment system consists of a state provision (job centres) and unemployment funds that manage paying benefit to unemployed members. In recent years, a strong activation policy has been introduced and the unemployment fund can sanction financially any person who fails to follow the rules and demands in the area (Jørgensen and Thomsen, 2016). During the last two decades, the judicial and economic terms have been made more restrictive for unemployed people (Engelbreth Larsen, 2013). Today, unemployed people with insurance are entitled to receive a benefit payment for up to two years of unemployment and it requires one year of fulltime working to retain that right. Whilst unemployed they must make seven job-applications per month and participate in mandatory activation programs in order to receive unemployment benefits.

Subjects
The subjects for the present survey were derived from member registers of two Danish unemployment funds for persons with an academic education. Subjects in three groups, a primary group with two control groups, were selected at random within the age ranges listed in Table 1. They were invited by email to participate in a study concerning unemployment and were provided with a customized link to fill out the survey. The primary group, Young Unemployed (YU), comprised persons registered as unemployed and in the age range 21-35 years. The target group selected comprised 4,468 persons, of whom 357 completed the questionnaire (8%). An age-matched control group was selected comprising persons within the same age range as the YU group but who were registered as being in employment (YE). The target group selected here comprised 2,389 persons, of whom 183 completed the questionnaire (8%). An age-comparison group was selected comprising older unemployed persons (OU) in the age range 41 to 63. Within this group, responses were obtained from 52 persons. The sizes of the three groups were determined within economic and logistic constraints. Demographic data for the three groups are presented in Table 1. Approximately 90% of subjects in all three groups have a Master's level degree, this being usual for Danish university educations.

Measures
The questionnaire was specifically designed for this study. It was informed by consulting the unemployment literature as indicated below where the various measures are introduced. It was also informed by qualitative interviews with 39 job seekers that were part of a mixedmethods research design in a larger doctoral project. The study was first piloted with qualitative feedback from 20 respondents. The survey was then altered based on their responses and evaluations. Except where otherwise indicated, the questions were to be responded to using a five-point Likert scale ranging from 1 = low to 5 = high. The questionnaire comprised two sections: a) questions were asked concerning general conditions to be completed by all three groups and b) specific questions concerning being unemployed which were directed to the YU and OU groups only.
Questions were asked concerning general conditions as follows with examples: Subjective well-being, 'Taken all in all, how happy would you say that you are?' (4-point scale) used in the European Values Survey (EVS, 2015) and the single item from the Life Satisfaction scale 'How satisfied are you with your life at the moment? (Diener et al., 1985). These two items correlate overall (r = .78) and were added to form a general scale of well-being with scores ranging from 2 to 9. Respondents who gave a response of 1or 2 to Life Satisfaction Scale elicited another set of measures from DASS (Crawford & Henry, 2003). The Depression Anxiety Stress Scale is 42-item self-report instrument designed to measure the three, related, negative emotional states of depression, anxiety and tension/stress. Work ethic, 'Even if I won the lottery, I would continue to work.' (Morse and Weiss, 1955) and 'I get bored without work.' (Rowley & Feather, 1987). These two items were averaged to form a scale of 'Work ethic' (r = .42).
We incorporated items inspired by Jahoda's fivecomponent model of mental health in relation to employment (Paul and Batinic, 2010). These were Activity, 'I am generally busy during the course of the day'; Time Structure, 'I lack structure in my everyday life' (Reversed) and 'My days are generally well-organised.' The mean of two items relating to managing time well (r = .63); Social Contact, 'I feel alone.' (Reversed) and 'I spend a lot of time together with others'; The mean of two items relating to meaningful social relationships (r = .47); Collective Purpose, 'I often engage in social activities'; Status, 'I feel that as unemployed I am looked down upon'. This last Jahoda item applied to the YU and OU groups only.
Additionally, specific questions were asked concerning being unemployed which were directed to the YU and OU groups only. These comprised the following: Selfperception, 'Even though I am unemployed I do not consider myself to be unemployed' developed in relation to this empirical study. Two scales were derived from a Mokken analysis (Schuur, 2011) of data from the YU group of which self-blame has been introduced in Pultz, Teasdale & Bang (n.d.). A Self-blame scale was formed from nine items: 'It is primarily my own fault that I am still unemployed'; 'I am unemployed because my job search effort has been insufficient'; 'If I improved my job interview skills I would have a better shot at finding a job'; 'I have a tendency to blame myself that I am still unemployed'; 'Sometimes I am afraid that something is wrong with me that prevents me from finding a job'; 'I am unemployed because I am not sufficiently professionally competent'; 'I am unemployed because my network is not big enough'; 'I am unemployed because I am not sufficiently outgoing'; 'I have to be more outgoing to get a job'. The Self-blame scales had a Cronbach's alphas of .82 and .80 for the YU and OU groups respectively. The second Mokken scale concerned Unemployment shame and comprised the following five items: 'I feel fine telling people that I am unemployed' (reversed); 'I experience being unemployed as a personal defeat'; 'I feel guilty receiving unemployment benefit'; 'Other people with a job see me as lazy'; 'I feel looked down upon as unemployed'. The scale was developed for the purpose of this study and the various items were developed on the basis of issues raised in interviews and by consulting the research on shame and unemployment (Alm, 2001;Rantakeisu, Starrin & Hagquist, 1999). The Unemployment shame scale had Cronbach's Alphas of .77 and .71 for the YU and OU groups respectively.
Although derived from the non-parametric Mokken analysis we elected to further test the dichotomous factorial structure of these 14 items by using Confirmatory Factor Analysis (CFA), since there exists, to our knowledge, no satisfactory non-parametric equivalent of the CFA. For purposes of the CFA we combined across the YU and OU groups and allowing the two factors to correlate. The resulting model could be improved since it has a significant Chi-Square value of 602 (df = 76, p < .01) and a rather high Root Mean Square Error of Approximation (RMSEA) of .13 (LO 90 = .12 and HI 90 = .14). For present purposes however, it can be considered satisfactory that the nine standardized factor loadings for Self-Blame were positive and ranged from moderate to good (range .49-.75) as were the loadings for Shaming (range .48-.78) and the two factors correlated .45; thus, being substantially in agreement with the Mokken analysis.
An additional item included in the present study was Job Flexibility which was defined as the number of types of employment (out of nine) currently being sought minus the number sought at the beginning of the current period of unemployment. A positive value thus indicates an increased flexibility over the time period. Being a single variable, it is not possible to examine the internal reliability of the Flexibility measure. However, since the information in the two source variables is comparatively objective there is reason to believe that the calculated Flexibility measure is itself valid.
Finally, we have included a dichotomous coding of the duration of the current period of receiving unemployment benefit as 0-12 months and 13-24 months.

Statistical procedures
Visual inspection revealed that the Likert scale measurement variables, albeit derived from the nonparametric Mokken procedure, showed no marked deviations from normality. We have therefore analysed all measurement variables using parametric procedures, namely t-tests, Anovas, post-hoc Bonferroni tests and multiple linear regressions. For all significance tests, alpha was set to .05, two-tailed where appropriate. Effect sizes for two-group comparisons were expressed, for significant differences only, as Cohen's ' d', calculated from an averaged standard deviation for the two groups involved or as Multiple R 2 for the regression models. Categorical variables were analysed using Fisher's Exact Test and Odds Ratios.

Results
The YU group of survey responders did not differ in mean age from the cohort originally contacted. However, females were significantly over-represented in the YU group compared to their cohort (Odds Ratio 2.32, 95% CI = 1.82-2.95). There were minor differences between the Young Unemployed (YU) and Young Employed (YE) groups with respect to age and gender distribution although it is noticeable that there is a relatively high proportion of females in all three groups (See Table 1). There are, however, substantially fewer among the YU group who are in a partner relationship than there are among the YE group (Odds Ratio 2.18, 95% CI = 1.46-3.26).
In view of the gender bias in the YU group we examined all three groups for gender differences in Well-Being but none of the three comparisons was significant (p > .1). However, analysis of variance showed a highly significant effect of group on Well-Being (F(2,575) = 68.6, p < .001) and post hoc Bonferroni tests revealed that the YU group reported substantially lower Well-Being than the YE group (p < .001) and with a large effect size. The YU and Older Unemployed (OU) groups did not differ (p = .6) (See Table 2).
Work Ethic yielded a significant overall effect (F(2,586) = 24.6, p < .001) and both pairwise comparisons are statistically significant (p < .01) with the YU group having a higher mean than the YE or OU groups, the latter effect being much larger than the former.
Results for the five Jahoda-inspired measures are also shown in Table 2. For three of the variables there were highly significant effects of group (F(2,586) > 5, p < .001). For Activity, the YU group mean was significantly below the means for both the YE and OU groups, although the effect size was much greater for the YU-YE comparisons. For Time Structure, the YE group mean was significantly (p = .017) above that for the YU group, albeit with a modest effect size. For Social Contact, the YE group mean was significantly above that for the YU group (p < .001) with a substantial effect size. The YU group did not differ from the OU group. There were no overall differences between the three groups with respect to Collective Purpose or Status, the latter being necessarily restricted to the YU and OU comparison.
The DASS questionnaire was triggered by low scores on one of the Well-Being item "Taken all in all, how happy would you say that you are?" Only nine of 179 in the YE group met this criterion, and the group was therefore excluded from analyses of the DASS scales. There was no significant difference in the proportions meeting the DASS criterion among the YU and OU groups (Fisher's Exact Test, p = .5). Notwithstanding that, there were significant differences (p < .001) between those subjects in the YU and OU groups who did complete the DASS. The YU groups had markedly lower means than the OU group on Depression, Anxiety and Stress, with effect sizes ranging from 1.5 to 1.9 (See Table 2). The comparisons between the YU and OU groups on the unemployment variables are shown in Table 3. The YU group were significantly more inclined to blame themselves for being unemployed (p < .001) and expressed more shame about being unemployed (p < .01), the effect sizes being large and moderate respectively.
A much lower proportion of the YU group had been unemployed for over 12 months (17%) than among the OU group (38%), (OR = 3.1, 95% CI = 1.61-5.78). As seen in Table 3, within the YU group, those who had been unemployed for 13-24 months had lower mean levels of Well-Being that did those who had been unemployed for 13-24 months. Table 4 shows the results, for the YU and OU groups separately, from a multiple linear regression predicting Well-Being from the self-perception as not being unemployed and the neoliberal variables, together with the duration of the present period of unemployment. The models proved to have satisfactory validity with no excess of large standardized residuals (outside of ±2.58) and Cook's Distances well below 1.0 in all cases, Leverage values were well within acceptable ranges (<.08 and <.34 respectively and Mahalanobis distances <35 in almost all cases (Tabachnick and Fidell 2007). It should be noted that, owing to the large discrepancy in sample sizes for the two groups, statistical power was much greater for the YU group than for the OU group.  It can be seen that among the YU group there was a statistically significant effect of Self-Perception with positive Well-Being positively associated with not considering oneself as unemployed. The dimensions of blaming oneself and being ashamed are both significantly and negatively associated with well-being. Among the OU group, none of the variables attain statistical significance, but it is noteworthy that the standardized betas show a pattern very similar to that for the YU group.
In a second set of linear regression models (see Table 5) we have examined the relationship between self-perception as not unemployed and the five Jahoda dimensions. Again, the validity of the models assessed from standardized residuals, Cook's Distances, Leverage and Mahalanobis Distances and was found to be satisfactory. For the YU group, among the five Jahoda variables, activity, social contact and status were all statistically significantly and positively associated with the self-perception dimension. For the OU group only activity was significantly (and positively) associated with self-perception. But here again the patterns of coefficients for the YU and OU groups are quite similar. It is worthwhile adding that, within the YU group, self-perception correlated significantly (p < .01) and negatively with self-blame (r = -.17) shame (r = -.32) and work ethic (-.24). Among the OU group only the corresponding correlation with self-blame was significant (r = -.43, p <, .01) although the correlation with shame also approached significance (r = -.27, p = .054).
It is worth noting that although there is a significant effect of duration of unemployment on well-being for the YU, but not the OU group (see Table 3), the effect appears to be absorbed by other predictor variables in our regression model (See Table 4).

Discussion
In this article, we have sought to illuminate how historically sensitive variables related to the neoliberal development and dispersion of activation policies affect the relationship between subjective well-being and unemployment in younger and older unemployed people respectively. Informed by governmentality studies we have demonstrated that how a person acts and reacts while being unemployed does not solely rely on that individual's personal characteristics or good or bad coping strategies, but is shaped by the surrounding society and how the actual governing of citizens take place. The contextual and societal dimension is often neglected in survey studies that usually focus on purely psychological variables, although there are exceptions (Boland and Griffin, 2015;Celik, 2008;Harris, 2001). Drawing on insights from governmentality studies, we develop a historically sensitive approach to investigating subjective well-being by taking into account that governmental policies and available representations and discourses influence how a person experiences being unemployed in the Danish welfare state which is characterised as a re-commodified social-democratic welfare state.
Among variables related to the effects of neoliberal policies and understandings of the unemployment experience can be counted self-blame, shame, work ethic, and flexibility. The present study thus enables a dialogue between customized historically-sensitive variables and more traditional variables often employed in the unemployment literature, such as measures of wellbeing and Jahoda's latent functions (1982). The study thus contributes with theoretical and methodological suggestions for how to study the link between unemployment and subjective well-being in times of dominating neoliberal understandings, ideologies and policies.
While we find that both the younger and older unemployed group experience a negative effect of unemployment compared to the employed control group, our results suggest that this is not the whole story. In fact, similar responses in subjective well-being in the younger and older unemployed groups are associated with a very different pattern of relationships to key variables in this study. The key similarities indicate that independently of age, unemployment affects subjective well-being negatively and both younger and older unemployed people lack time structure and activity as suggested by Jahoda (1982) and replicated many times since (Paul and Batinic, 2010). However, younger people self-report a higher level of self-blame and shame, they report a higher level of work ethic and they become more flexible in terms of applying for a broader range of jobs as time passes albeit not at a statistically significant level. We interpret these findings as indications that neoliberal policies and understanding impact upon different age groups in different ways, with young unemployed people being more affected by them.
In contrast, the older group have significantly higher levels of depression, anxiety and stress as measured by the DASS (Lovibond & Lovibond, 1995). Comparing our results to norms and cut-off scores distinguishing between symptom levels for the DASS, the young unemployed group report on average being in the mild to moderate range whereas the older unemployed people report symptom levels in the severe to very severe range. Importantly, the DASS was conditioned on reporting a low level of subjective well-being and therefore the responses are not representative of all individuals in the two groups. Thus, we limit our interpretations to the comparison between those younger and the older unemployed people for whom a clinically important symptom level is compellingly supported by the empirical material. In that sense, despite the fact that the young and older unemployed group give similar answers to the subjective well-being measure, our results provide partial support for the trends documented in the literature suggesting that older people experience a larger negative effect of unemployment on psychological well-being compared to younger unemployed people (Breslin and Mustard, 2003;Briar et al. 1980;Goldsmith, Veum, and Darity, 1996;Theodossiou, 1998). We have demonstrated that a global measure such as subjective well-being which is widely used in the research literature, such as in the European Values Survey (EVS, 2015); in fact, can conceal important and interesting differences. Hence, the measure is useful when distinguishing roughly between very different groups, such as the young unemployed people and the control group of employed people, however when it comes to more fine-grained analysis this measure is insufficient.
The neoliberal activation policies introduced since the 1990s in Denmark have increasingly made receiving unemployment benefits conditional on job-search activities and studies investigating governmental practices highlight the importance of self-responsibility and active job search among unemployed people. We speculate that this increase in active labour market measures (Jørgensen and Thomsen, 2016) affects how unemployment is experienced. Interestingly, the young unemployed people self-report a higher level of work ethic than both the group of employed young people and the older unemployed group. This finding echoes previous research by Gibson (2009) suggesting that effort is central when evaluating a person's merit in relation to receiving unemployment benefits in western welfare states. Previous research by Johnson, Sage, and Mortimer (2012) has demonstrated how work values tend to adapt to work situations with people in well-paying and steady jobs valuing extrinsic values such as pay and stability more than people in insecure, or no, current employment. The explanation of this phenomenon termed the 'reinforcement and accentuation model' avoids the psychological distress associated with cognitive dissonance experienced if one's values are not in accordance with one's situation. Following this logic, unemployed people should report less work ethic compared to the employed group. This is, in fact, also the case with the older unemployed group who as predicted by the model show a lower level of work ethics than the young employed group. However, the young unemployed group in fact exhibit the highest level of work ethic leading to a discrepancy between the value reported and actual behaviour. We speculate that the heightened work ethic is an indication that the young unemployed people feel ashamed of their situationespecially when they are seen as primarily to blame for it -and therefore compensate by demonstrating values in accordance with neoliberal ideology valorising work, productivity and efficiency (Honneth & Hartman, 2001). As we believe that work ethic is somehow linked to shame, we would urge future research to investigate this association more closely. Further correlational analyses show that indeed shame and work ethic are significantly positively correlated for the young unemployed group, and in contrast they are negatively correlated for the older unemployed group, albeit not at a statistically significant level, possibly due to the small sample size of the older unemployed group.
From sociological literature and especially from governmentality studies, self-responsibility has been identified as pivotal in contemporary ways of governing citizens (Dean, 1995;Lazzarato, 2009;Walters, 1994). In an unemployment context, self-responsibility is manifested as self-blame (Sharone, 2013). Contemporary labour markets are increasingly characterized as 'precarious' (Standing, 2011) and it seems that when the responsibility for unemployment is largely placed on the individual, what is in the individual's own control is exactly the lowering of expectations or the increasing of flexibility. On the reverse side, this strategy is correlated with a lower level of well-being. Whereas the experience of agency and control is usually identified as a buffer against the deteriorating effect of unemployment on self-blame, new research identifies limits to the beneficial effects of perceived control in the face of uncertainty (Glavin and Schieman, 2014). The authors stress that unrealistic expectations lead to frustration, cognitive dissonance and psychological distress. A similar pattern might be expected with regard to flexibility but more research is needed to address this question.
It is our belief that a key contribution of our multivariate regression analyses is that we have demonstrated that not considering oneself as unemployed serves to elevate the level of well-being experienced by young unemployed people. Drawing on governmentality studies we tentatively interpret this finding in terms of the entrepreneur (Foucault 2008;Read 2009). As Foucault (2008) notes in 'The Birth of Bio-Politics', a certain type of subject is produced in the neoliberal era whom he is termed 'Homo Economicus' characterized as an innovative, entrepreneurial type invested in optimizing human capital in the competitive labour markets. Foucault writes: 'homo economicus is an entrepreneur, an entrepreneur of himself' (Foucault 2008). Our results here echo previous work done by Pultz and Mørch (2015) who in a qualitative study have demonstrated the 'freelancer' as a figure that increasingly appears as a way of dealing with contemporary demands to become a qualified member of society (Andersen and Mørch, 2005). By not viewing themselves as being unemployed, this group to some extent avoids the negative stigma associated with the unemployment label.
The multivariate regression analyses suggest that, in addition to experiencing improved subjective wellbeing, these people experience less deprivation in terms of Jahoda's latent functions such as social contact and activity. Jahoda's influential latent deprivation model proposes that manifest work entails a number of latent functions that serve to protect mental health. Jahoda (1991) speculated that it would most likely not be possible for the unemployed individual to organize or create the latent functions (time structure, social support, meaningful activity, social appraisal, collective purpose) without gaining employment. She writes that it would require: 'personal initiative that is rare among all strata of the population' (Jahoda, 1982, p. 94). Our results tentatively suggest that in today's insecure and precarious labour markets that sort of initiative is perhaps not at all rare. Reflecting upon the discrepancy between Jahoda's model and newer empirical studies investigating the agentic practices of young people, Barr & Orford (2002) ascribe some of the difference to changed time and space and they speculate whether the initiative required to construct the latent functions described by Jahoda might be a city phenomenon with young people in urban areas suffering less than unemployed people in rural areas (Ball & Orford 2002). Fryer and Payne (1984) also speculate that agency entails taking initiative and, in the case of unemployment, for some proactive people it seems possible to distinguish between employment and meaningful activity. If an unemployed person experiences being invested and engaged in meaningful activities, this will serve as a protection against the deterioration of mental health which is so widely documented in the unemployment literature. Subcultures and artistic milieus have long existed (Bain & McLean, 2013;Demetry, Thurk, & Fine, 2013;Thornton, 1995) but the present results tentatively imply that the practice is not confined to a few artists, as approximately one-third of the survey respondents agree to the statement that they do not perceive themselves as unemployed. This could indicate that the practice is becoming more widespread in a labour market that is increasingly precarious and consists of freelancers and project contracts (Fogh Jensen, 2009).
A few limitations of our study deserve mentioning. The low response rate constitutes a potential bias even though this challenge is recognized in many online survey studies. Comparing the YU respondents to the cohort from which they were recruited, respondents do not differ on age or amount of work experience, however there are more women among the respondents compared to the cohort. Given a lack of further information on nonresponders it is only possible to speculate on the reasons for the low response rate. We believe that a multiplicity of factors may have operated. First, the contacted persons were informed that the online questionnaire would take '30-45' minutes to complete, which itself could have discouraged many of them from responding. Second, some of those contacted had in the short interval perhaps gained employment, or knew that they were about to gain employment. A third reason is that people may have been guarded against a questionnaire which they could have been sure would request very personal information. Related to this point, the topic at hand is sensitive and sensitive subjects have been shown to have lower response rates (Tourangeau & Yin, 2007). A fourth reason is that they have been exposed to other forms of online contact and information-gathering and have thus felt a 'response-fatigue'. One could conjecture that more conscientious and compliant persons would be more likely to complete that questionnaire. This is consistent with the over-representation of women in the YU group, since females are typically found to score higher than men on measures of conscientiousness (Schmidt, et al., 2008). At the same time, it is noteworthy that there were no gender differences in well-being in any of our three groups.
Beyond that possibility, however, it is difficult to conjecture the direction and magnitude of any sampling biases caused by the low response rate and any future research should aim to achieve higher response rates as well as more information about non-respondents.
In all self-reported data, there is the possibility of social desirability bias, i.e., the tendency to provide an answer that is culturally accepted or socially approved. Furthermore, the specific wording of a particular item shapes the distribution of answers. Our survey has been cross-sectional and without a longitudinal dimension we are unable to investigate the timing and development of the key variables. Without results based on gathering information in multiple waves, we are primarily restricted to between-subjects analyses rather than within-subjects analyses. Due to the momentary nature of our data we are unable to shed light on the interesting debate between the aforementioned hypotheses of causation versus selection and we cannot discount the possibility that the difference between the unemployed groups and the control group arises because lower mental or physical health results in reduced employability rather than unemployment causing the negative effects. Longitudinal data would be preferable in relation to research questions addressing changes across the scope of time and, ideally, the data should be collected before and after the onset of neoliberal technologies, which in Denmark were introduced in 1994 ( Bengtson et al., 2015).
The present exploratory study thus points to interesting dynamics in today's societies, but it is a limitation that we are unable to rule out the possibility that shame and self-blame have also been dominant reactions to unemployment before the onset of neoliberalism. Consulting the literature, however, in particular Sharone (2013), shows how subjective responses to unemployment largely depend on the hiring systems in a given context. Comparing Israeli and American job seekers experience with unemployment, Sharone concludes that while Americans seem to blame themselves, Israeli job-seekers instead feel powerless. For current purposes, it is noteworthy that Sharone relates self-blame to neoliberal and individualist cultural and institutional resources. Unfortunately, unemployment shame was not directly targeted in this study. For future work, incorporating a longitudinal element to the research design would be preferable. In addition, more direct measures of the group of entrepreneurs would strengthen the design and perhaps contribute to informing us about the historical contingent practices of what it means to be unemployed in the Danish welfare state today.

Conclusions
In view of the low return rate, conclusions from the present study must be drawn with caution. However, as they stand, our results presented in this article suggest that historically sensitive variables related to the neoliberal development and policies greatly influence the relationship between subjective well-being and unemployment. They also suggest that it affects different age groups in different ways, with young unemployed people overall being more sensitive than older unemployed people to the neoliberal pressures, here defined as self-blame, unemployment shame, flexibility and work ethic. Measuring subjective well-being is useful when comparing across very different groups but it is not adequate when comparing groups that both experience unemployment. A key contribution is the identification of formally unemployed people who do not perceive themselves as unemployed. They do not suffer the same apparent deteriorating effects of unemployment on subjective well-being. Building on that, we emphasize the need to take into account the heterogeneity in how people act and react while being unemployed. This heterogeneity furthermore implicates that people in different (un)employment situations might benefit from different interventions and it is worth taking this into consideration in the development of labour market policies. Furthermore, our work suggests that future research should address Jahoda's influential model and update it in relation to today's precarious labour markets in which people are governed to act by taking initiative and demonstrating a large degree of flexibility.

Ethics and Consent
The study was approved by the Danish Data Protection Agency.

Funding Information
This study was financed by a PhD stipendium awarded to SP.