Germany s working class left behind: the growing class gap in life satisfaction between 1984 and 2013

Transcription

1 Germany s working class left behind: the growing class gap in life satisfaction between 1984 and 2013 Oliver Lipps (Swiss Centre of Expertise in the Social Sciences FORS, Lausanne, Switzerland, oliver.lipps@fors.unil.ch) & Daniel Oesch (Life Course and Inequality Research Centre LINES, University of Lausanne, Switzerland, daniel.oesch@unil.ch) Abstract The 1990s and 2000s were a gloomy period for Germany s working class, hit by mass unemployment, welfare retrenchment and wage stagnation. We examine whether rising economic inequality went along with a widening class gap in subjective well-being: Did the destiny of people at the top and the bottom of the social hierarchy diverge with respect to how satisfied they were with their lives? We answer this question by analysing the evolution of life satisfaction across classes and income quartiles for Germany, based on the Socio-Economic Panel, We check the validity of these results by examining another indicator of subjective well-being self-reported health and by comparing the findings for Germany with those for Switzerland, where economic inequality remained stable over the last two decades. In Germany and Switzerland, higher income quartiles and classes consistently report higher life satisfaction and better health than lower quartiles and classes. Clearly, people who earn a higher income and are in a more advantageous class position are more satisfied with their lives. In Germany, but not in Switzerland, the gap in life satisfaction and subjective health has widened substantially over the last two decades between the upper middle class and the low-skilled working class. Keywords: Germany, Switzerland, life satisfaction, subjective well-being, health, social class, inequality Paper to be presented at Universidad Carlos III, Madrid, 8 April

2 Introduction The two decades between the fall of the Wall and the Great Recession were not a good period for Germany s working class. Mass unemployment, welfare retrenchment and the decline of trade unions weakened low educated workers position in the labour market. As a result, the 1990s and 2000s brought a rise in atypical employment, low-paid work and wage inequality (Bosch and Kalina 2008, Kolev and Saget 2010: 5). While real wages increased at the top of the earnings distribution, they declined at the bottom (Antonczyk et al. 2010). The widening gap in economic resources between the top and the bottom of Germany s wage structure between the upper-middle and the working class is well documented (Bosch 2009, Streeck 2009). The question raised in this paper is whether this rise in economic inequality has gone along with an increasing gap in workers subjective well-being. Did the destiny of people at the top and the bottom also diverge with respect to how satisfied they are with their lives? Our paper answers this question by comparing the evolution in life satisfaction of high earners and low earners, the upper-middle and the working class in Germany over the last three decades. There is a growing body of research which analyses inequality in life satisfaction over time and across countries, commonly measured with the variance of life satisfaction within the entire population (e. g. Alesina et al. 2004, Delhey and Kohler 2011, Stevenson and Wolfers 2008). Our primary goal is not to examine the inequality in life satisfaction per se. Rather, we wish to trace the evolution in life satisfaction of people in the two categories set at the top and bottom of the social hierarchy, be it in terms of earnings (the first and fourth income quartile) or occupation (the upper-middle class and the low-skilled working class). We know that Germany s working class has lost out in terms of earnings (Bosch 2009) but was it also left behind in terms of life satisfaction? Of course, an increasing gap in well-being between social classes in Germany may be driven by other causes than widening economic inequality. Notably, it may be the result of international trends such as globalization (and offshoring) or skill-biased technological change (and automation). These trends arguably make life more difficult for low-educated people everywhere in Western Europe. We address this issue by comparing the evolution in well-being in Germany with that in Switzerland, a neighbouring country that was exposed to the same shifts in trade and technology, but where wage inequality and the proportion of lowpaid work remained stable over the last two decades. Of course, a two-country study does not allow us to identify causal links between income inequality and subjective well-being. Rather, the empirical objective of our study is 2

3 descriptive and consists in documenting the evolution in well-being of different socioeconomics groups over time. Our analysis is set at the individual level and based on data stemming from the German Socio-Economic Panel (SOEP), , and the Swiss Household Panel (SHP), In order to increase the validity of our analysis, we use two indicators to measure subjective well-being life satisfaction and self-reported health and two indicators to measure people s position in the labour market social class and income quartile. Our paper is structured as follows. Section II reviews the literature on inequality in subjective well-being and argues why we expect an increasing class gap in life satisfaction in Germany. Section III presents our data, measures and estimation method. Sections IV and V show evidence for the increase in income inequality and the rising class gap in subjective well-being in Germany, but not in Switzerland. The concluding section VI sums up our findings. Income inequality and inequality in subjective well-being Research on subjective well-being has turned into a growth industry in recent years, spurred by the increasing availability of surveys that include questions about life satisfaction and happiness. If one solely focuses on studies covering at least a decade worth of data, there are several contributions that examine the link between income inequality and inequality in subjective well-being. Based on individual-level data from 72 countries measured at three time points, Ovaska and Takashima (2010: 214) find that larger income inequality goes along with larger inequality of happiness within a given country. Likewise, in an analysis of 44 US states ( ) and 12 European countries ( ), Alesina et al. (2004: 2035) report that individuals are, on average, less happy in periods when income inequality is high than when it is low. The same conclusion is drawn by Oishi et al. (2011: 1099) for the United States over a 37-year period and by Hajdu and Hajdu (2014: 5) for Europe over the 2000s. A literature review by the OECD therefore concludes that higher levels of economic inequality are associated with lower average levels of subjective well-being (Senik 2009: 3). However, this seemingly unambiguous association is called into question by another wellestablished finding: inequality in life satisfaction has been falling in most Western countries since the early 1980s (Clark et al. 2014). This evolution is puzzling as we observe, over the same period, a U-turn in income inequality with widening gaps between top earners and low earners (Alderson and Nielsen 2002, OECD 2011a). Yet despite rising income inequality, there is ample evidence of a downward trend in general inequality in life satisfaction. This 3

4 finding emerges from multi-country studies based on the World Values Survey between the 1970s and 2000s (Veenhoven 2005: 474, Clark et al. 2014) as well as from single country surveys for Australia in the 2000s, Britain in the 1990s and 2000s (Clark et al. 2014: 29) or Ireland between 1994 and 2001 (Madden 2011: 676). The evolution is more ambiguous for the United States and Germany. In the US, happiness inequality decreased from its highest level in the 1970s all the way through the 1990s, but possibly started to rise again in the 2000s (Dutta and Foster 2013). When decomposing changes in the distribution of subjective well-being between 1972 and 2006, Stevenson and Wolfers (2008: 74) find that the black-white gap in happiness has declined and the gender gap disappeared. Yet the inequality in happiness by education has increased: happiness was rising among college graduates, decreasing among respondents with some college, and falling sharply among those with a high school degree or less. In Germany, inequality in happiness steadily decreased during the 1980s and early 1990s. Starting from the mid-1990s, the evolution has either been described as trendless fluctuation (Clark et al. 2012) or, less persuasively, as showing a slight increase in happiness inequality (Becchetti et al. 2013). What seems undisputed is that the 1990s saw a break in the downward trend towards less inequality in subjective well-being. Becchetti et al. (2013: 20) argue that happiness inequality in Germany ceased to fall because of the depressive impact that rising levels of unemployment had on well-being. One explanation for why inequality in life satisfaction may have decreased in many Western countries (but perhaps not in Germany or the US) is that economic growth allows people to get access to more public goods such as education, health, public infrastructure and social protection (Clark et al. 2014: 13). Accordingly, GDP growth may not raise life satisfaction on average, but reduce the dispersion in life satisfaction because it leads to a better provision of public goods. Yet if income growth goes along with an increase in income inequality, the effect of economic growth may be undone and inequality in life satisfaction will rise as a result possibly the scenario which prevailed in Germany and the US after the 1990s (Clark et al. 2014: 13). Why would we expect a rising gap in life satisfaction across different social classes in Germany since the 1990s? Over the last two decades, the economic destiny of the working and the upper-middle class has diverged strongly. During the post-reunification recession, unemployment rose rapidly and remained at over 8 per cent for 15 years in a row, from 1993 to But unemployment afflicted low educated workers in Germany much more than high educated workers. Over the 1990s, the unemployment rate of workers with no more than nine 4

5 to ten years of formal education was, on average, 2.6 times higher than that of workers with tertiary education. Over the 2000s, this ratio increased to 3.8 (OECD, Education at a Glance, various years). In addition, during this long crisis, there was a weakening of three collective institutions that impose public obligations on employers and thus primarily benefit workers with little bargaining power: trade unions, works councils and the welfare state (Streeck 2009: 170). First, German trade unions came under pressure after the long recession of the early 1990s when unemployment rates rose rapidly and membership figures began to dwindle. Between 1991 and 2011, union density was exactly halved, dropping from an all-time high of 36 to 18 per cent (Visser 2011). As a consequence, collective bargaining was increasingly called into question by employers, and coverage with collective agreements in the market sector fell from 78 per cent in 1990 to 58 per cent in 2010 (Visser 2013: 93). In parallel, the outreach of collective bargaining was curtailed by employers increasing use of exit options such as (i) the expansion in outsourcing and temporary agency work, (ii) the rise of mini-jobs, and (iii) the growing importance of temporarily posted workers from other European countries (Bosch 2009: 346). Works councils were a second collective institution that came under pressure. Employers increased use of exit options coincided with a long-term decline in the number of firms with works councils. Since work councils supervise the enforcement of collective agreements, their decline further reduced the reach of collective bargaining as agreements became less binding on individual firms (Streeck 2009: 40). Taken together, these changes significantly curtailed the inclusiveness of Germany s pay-setting institutions (Bosch et al. 2010: 91). In the absence of a legal minimum wage (not introduced before 2015), the erosion of collective pay-setting weakened the labour market position of those workers who had little individual bargaining power. Third, the German welfare state, notably the unemployment benefit system, was redesigned in the early 2000s to make welfare benefits less attractive. Under pressure from persistently high unemployment, Gerhard Schröder s government implemented several reforms between 2003 and 2005, amongst which the Hartz laws, which aimed at encouraging the unemployed to take up low-paid jobs. These reforms first introduced new types of marginal employment and stimulated self-employment (Hartz II in 2003) and then reduced both the level and duration of unemployment benefits (Hartz IV in 2005). The explicit aim was to increase the willingness of low-skilled workers to accept low-paid jobs. 5

6 The triple influence of persistently high unemployment, the weakening of collective paysetting institutions and cutbacks in unemployment benefits put Germany s working class under increased market pressure and reduced the wage floor. The result was, after the 1990s, an unprecedented rise in low-wage (and often atypical) work (Bosch and Kalina 2008) and income inequality (Biewen and Juhsasz 2012: 624). Since 2000, income inequality in Germany increased more than twice as much as in the average of the OECD (OECD 2011b). Based on these arguments, we find two contradicting expectations as to the evolution of life satisfaction of different social classes. One expectation is that economic growth leads to a better provision with public goods and hence to happiness moderation (Clark et al. 2014). Despite three recessions in the early 1990s, the early 2000s and 2009, German GDP expanded between 1992 and 2012 by almost a third (29 per cent in real terms). Based on this substantial increase in society s aggregate income, we would expect a stable or even decreasing gap in life satisfaction between the upper-middle and the working class. A second argument highlights the fact that over the last two decades, Germany s institutional safeguards have become less effective at sheltering low-skilled workers from market forces and that unemployment and income inequality have strongly risen. As a result, we would expect that life satisfaction has fallen in the low-skilled working class, while it remained stable, or may even have increased, in the upper-middle class. Of course, a widening gap in life satisfaction between the working class and the upper middle class in Germany may well be due to other factors than increasing wage inequality. Low-educated workers may have come under pressure everywhere in the OECD as a consequence of international trade and offshoring (Krugman 2008) or skill-biased technological change and automation (Berman et al. 1998). Life may thus have become more difficult for low-educated workers in Europe and North America regardless whether collective pay-setting institutions were dismantled or not. An ideal case to contrast the German experience and the inequality hypothesis is neighbouring Switzerland. Switzerland shares several institutional features with Germany, notably the strong focus on vocational education and industry-based collective bargaining. Moreover, there was also a hefty increase in Switzerland s GDP between 1992 and 2012 of 37 per cent. But contrary to Germany, Switzerland s unemployment rate remained comparably low and wage-setting institutions were not dismantled: the coverage rate with collective agreements remained stable at around 50 per cent between 1991 and 2011 (Oesch 2011) and unemployment benefits remain comparatively generous with a replacement rate of between 70 and 80 per cent. Accordingly, the trend in wage inequality in Switzerland was basically flat 6

7 and thus stands in stark contrast to the steep increase in Germany (OECD 2011, Kuhn and Grabka 2012, Kuhn and Suter 2015). By comparing the evolution of life satisfaction in Germany with that in Switzerland, we obtain a simple difference-in-difference design. Switzerland thus serves us as a contrast case where wage inequality (and unemployment) remained more or less constant over the last two decades. Data, measures and estimation method Our empirical analysis is based on longitudinal data from the German Socio-Economic Panel, waves (SOEPv30, Wagner et al. 2007). For the comparison with Switzerland, we draw on data stemming from the Swiss Household Panel, waves (SHP, Voorpostel et al. 2014). These nationally representative surveys collect information on a wide range of respondents socio-demographic characteristics, labour force participation and subjective well-being. In order to analyse the same population over time, we limit our analysis to West Germany. In addition, we restrict our sample to respondents aged between 20 and 65 years who are in paid employment, either as employees or self-employed, working at least 10 hours per week on a yearly average. This leaves us with large samples of 6,799 (1984) to 7,833 individuals (2013) in Germany and 4,162 (1999) to 3,832 individuals (2013) in Switzerland. Our main dependent variable is people s self-reported life satisfaction and based on the question: How satisfied are you with your life, all things considered?, with answers ranging from 0 (completely dissatisfied) to 10 (completely satisfied). The variable s distribution is shown in the appendix (see figure A.1). Although this 11-point scale is, strictly speaking, an ordinal variable, we follow the established practice in the well-being literature (Ferrer-i- Carbonell and Frijters 2004, Clark et al. 2008) and treat it as a cardinal variable. This allows us to compute the mean value of life satisfaction. Some sociologists are sceptical whether the question about self-reported life satisfaction captures the concept of subjective well-being. We address this issue by resorting to a second measure that, while also linked to subjective well-being, may seem more tangible to these sceptics: a question about self-reported health where respondents are asked to rate their health on a five-point scale, ranging from 1 (very bad) to 5 (very good). This question has been asked in the German Socio-Economic Panel first in 1992 and then continuously from 1994 to In the Swiss Household Panel, it has been included from the onset in Our key independent variable is social class where we use an aggregated version of the schema proposed by Erikson and Goldthorpe (1992) and distinguish four hierarchically ordered classes: (i) the upper and upper-middle class containing large employers, managers 7

8 and professionals; (ii) the lower-middle class including small employers, technicians and associate professionals; (iii) the skilled working class including clerks, craftsmen, skilled service and sales workers; (iv) the low-skilled working class containing assemblers, machine operators, farmhands, low-skilled service and sales workers, and elementary occupations more generally. Employed and self-employed respondents are attributed to the four classes based on their main occupation as measured with isco-88 4-digit. 1 As for those individuals working as employers and having more than 9 employees, we allocate them to the uppermiddle class regardless of their occupation. Likewise, small employers (1 to 9 employees) whose occupation is neither manager nor professional are attributed to the lower-middle class (for more details, see Oesch 2006). Unlike sociologists, who are used to measuring people s hierarchical position in the employment structure with a class variable, economists more commonly measure the same concept with income groups. In order to increase the validity of our measurement, we use both indicators: occupational class and income quartile. We divide our sample of working adults into four income quartiles, based on hourly work income. The 1 st quartile encompasses the 25% of respondents with the lowest work income and the 4 th quartile the 25% of respondents with the highest work income. We expect the four occupational classes and four income quartiles to measure the same underlying phenomenon of social hierarchy. Alongside class and income quartile, we control for sex, age, age squared, region (Bundesländer for Germany, greater regions for Switzerland), nationality (national vs. nonnational) and living together with a partner. Of particular interest are period effects which we measure in five-year spells which roughly correspond to different political and economic cycles (84-89, 90-94, 95-99, 00-04, 05-09, 10-13). Moreover, in a last set of models, we introduce contextual variables for income inequality, unemployment and how worried people are about their job security. While job security is measured on the individual level (1 very concerned and 3 not concerned at all ), income inequality (p75/p25) and the unemployment rate are measured at the level of the Bundesland or the greater region to give us more variance. Table A.1 in the appendix shows the descriptive statistics of our measures. Our analytical approach consists, first, of an extensive description of how life satisfaction evolved in the different classes and income quartiles. Rather than computing an aggregate measure of dispersion for the entire population (typically the variance or standard deviation), we compare the evolution of life satisfaction across social classes and income quartiles. We I Codes are available from the authors. 8

9 do so by calculating the yearly means in life satisfaction for each class and quartile. In order to reduce the influence of short-term fluctuations and to better grasp the time trend, we use weighted scatterplot smoothing (LOWESS). This means that we compute the mean life satisfaction in a given year by borrowing some additional information from adjacent years, where nearest neighbouring years get higher weights and more distant years lower weights. This provides us with a line across years which best fits the data, but without imposing a functional form. We check our descriptive results with a multivariate model. Classes (or income quartiles) are largely composed of the same people over time as entry into a class (or an income quartile) is contingent on time-constant factor such as social origin and, above all, education. 2 A medical doctor is likely to belong to the upper-middle class (and 4 th quartile) over the entire career, a farmhand without post-mandatory education unlikely to leave the low-skilled working class (and 1 st quartile). This pleads for a random-effects model which captures both the variation between and within individuals. In contrast, a fixed-effects model only uses the within-individual variation which means that there is no variation for the workers who do not change classes over time. The general equation of our random-effects linear regression model is given as: yy iiii = ββ 1 + ββ 2 cccccccccc iiii + ββ 3 pppppppppppp iiii + ββ 4 cccccccccc iiii ppppppppppdd iiii + ββ 5 WW iiii + uu ii + εε iiii The independent variable yy ii is the 11-point measure of an individual i at time t. WW iiii is a vector of controls such as sex, age, age squared, region, nationality and having a partner, uu ii a normally distributed variable to capture unobserved person-specific heterogeneity and ε it the idiosyncratic error term. The random effects model stipulates that uu ii is uncorrelated with the explanatory variables, as well as εε iiii with the explanatory variables and with uu ii. Our two main predictors are social class (or, alternatively, income quartile), cccccccccc iiii, and 5-year spells which capture the period effects, pppppppppppp iiii, with the main interest being on a third predictor, the interaction between class and period, cccccccccc iiii pppppppppppp iiii. If there is a different evolution in the life satisfaction between classes (or income quartiles) over periods, this effect will be picked up by the interaction term. We correct for auto-correlation in our panel data by using panelcorrected robust standard errors. 2 A transition matrix for Germany shows that only a small share of workers is mobile across classes: 88 per cent in the low-skilled working class and 92 per cent in the upper-middle class do not change class from one year to the next over the period of observation. There is somewhat more mobility across income quartiles, but a majority still remains immobile over time: 68 per cent in the 1 st and 77 per cent in the 4 th quartile. 9

10 We are aware of the fact that our analysis of life satisfaction in two countries has little causal traction. We are much better prepared to describe how life satisfaction evolved for different classes and quartiles than to explain why it changed as it did. The evolution of life satisfaction across income quartiles and classes Before we model life satisfaction over time in more detail, we need to show that income inequality effectively increased between classes and quartiles in Germany, but not in Switzerland. Figure 1 depicts the evolution of hourly earnings for the four quartiles and the four classes in Germany and Switzerland. In Germany, the earnings gap between the bottom and top quartile constantly widened between 1984 and In 1984, the median hourly work income in the 4 th quartile was 2.5 times that of the 1 st quartile; this ratio rose to 2.9 in 1994, 3.4 in 2004 and 3.6 in In Switzerland, the hourly earnings of the four quartiles evolved in parallel over a shorter time period, the ratio between the 4 th and 1 st quartile remaining basically stable at 2.7 (1999) and 2.6 (2013). By definition, earnings in the four quartiles are hierarchically rank-ordered, being higher in the 2 nd than the 1 st quartile, and higher in the 4 th than the 3 rd quartile. In contrast, our social classes are constructed without reference to earnings. Still, figure 1 shows for both countries a clear rank-order of classes in terms of hourly earnings: the upper-middle class has the highest and the low-skilled working class the lowest median earnings over the entire period. Moreover, earnings inequality in Germany did not only widen across income quartiles, but also across social classes. At the end of the post-reunification boom in 1993, the median earning in the upper-middle class was 1.88 times that of the low-skilled working class. Twenty years later, this ratio had mounted to In Switzerland, the gap between the top and bottom class remained stable from 1999 to 2013 at around 1.6. Another figure emphasizes the extent to which Germany s low-skilled working class was left behind economically. Between 1999 and 2013, the median hourly earnings of the lowskilled working class rose by a meagre 3 per cent aggregated over 14 years and in nominal terms (that is, before correcting for inflation). In comparison, the median hourly earnings increased by 11 per cent for the upper-middle class. About here Figure 1 How did subjective well-being evolve in Germany over the same period? The upper-half of figure 2 compares the evolution in life satisfaction between the four quartiles (left) and four 10

11 classes (right); the lower-half plots the difference in the evolution of the highest and the lowest quartile (left) or the highest and lowest class (right) respectively. Regardless whether we look at income quartiles or classes, we observe the same pattern, with higher quartiles and higher classes consistently reporting higher life satisfaction than lower quartiles and lower classes. Clearly, people who earn a higher income and are in a more advantageous class position are more satisfied with their lives. Over the last thirty years, life satisfaction in Germany fell during the 1980s for all quartiles and all classes. In the early 1990s, the fall came to a standstill for the top quartile and the upper-middle class, but not for the bottom quartiles and lower classes. Their mean life satisfaction decreased until 2007/8 when it finally began to rise again. Accordingly, the gap in life satisfaction between the top and bottom was the smallest at the end of the postreunification boom in 1992/3 and the largest in 2007 to 2009 when the recovery of the German labour market finally began in earnest. At its maximum, the class disparity in life satisfaction was 0.55 points on the 11-point life satisfaction scale and hence roughly twice the gap observed between the more satisfied people who have a partner and the less satisfied people who do not have a partner. 3 About here Figure 2 In comparison, workers seem more satisfied with their lives in Switzerland than Germany. Mean levels in Switzerland exceed those in Germany by over half a point (see figure 3). Not only levels of life satisfaction are higher, but also inequality in life satisfaction is smaller in Switzerland. The disparity between the highest and lowest class (or the top and bottom quartile) is only half as large as in Germany, fluctuating between a minimum of 0.12 and a maximum of Contrary to Germany, we do not observe a clear trend in the disparity of life satisfaction over time in Switzerland. After 2005, the gap in life satisfaction decreased across quartiles, but increased across classes. While we observe a downward trend in life satisfaction for all classes in Switzerland, there was a particularly strong drop for the low-skilled working class after About here Figure 3 3 The life satisfaction of respondents with a partner was, on average, 0.24 points higher than for respondents without a partner (calculated over the waves for West Germany). 11

12 Does a single question on life satisfaction tell a robust story about increasing inequality across income and class? We address this concern by replicating our analysis with another measure for subjective well-being, self-reported health (rescaled: 0(very bad) to 4 (very good)). Figure 4 shows how self-assessed health evolved across quartiles and classes for Germany between 1992 and Again, we find that high earners and the upper-middle class are not only better off economically, but also rate their health more positively than low earners and the working-class. The health gap between income quartiles is very small in the early 1990s, but constantly widens after 1996 and reaches its maximum in Inequality in self-reported health is much larger when measured across classes than across income quartiles. The health gap between classes is already sizeable in the 1990s and further increases in the 2000s. Interestingly, over the 2000s self-assessed health improves for the top quartile and the highest class, but steadily declines for the bottom quartile and the lowest class. We find again a smaller difference in self-reported health across income quartiles and classes in Switzerland than in Germany (see figure A.1 in the appendix). In Switzerland, selfassessed health largely evolved in parallel for the four quartiles and classes until 2008, declining everywhere. After 2008, the drop is more marked for the bottom quartile and the lowest class than the rest of the workforce. In both countries, we observe much larger differences in health across classes than income quartiles and the low-skilled working class stands out as reporting much poorer health than the other three classes. Class thus seems to capture a dimension of disadvantage that is not grasped as clearly by income quartiles. One explanation is that class provides a better proxy for permanent income than a single measure of current income which may lump together, in the same quartile, a young accountant and a mid-aged stone-mason (Erikson and Goldthorpe 2010: 219). Another explanation is that occupational class captures the strenuousness of a job typically the extent of physical labour and provides thus a closer link to health than income. Finally, class also correlates more closely with years of education than do income quartiles (r=0.63 vs. r=0.37), and the strong nexus between education and health is well known (e.g. Huijts et al. 2010). About here Figure 4 A multivariate analysis of the effect of class on life satisfaction We examine the validity of our descriptive results by analyzing the link between class and life satisfaction in the multivariate setting of random-effects model, controlling for 5-year periods, 12

13 sex, age, age squared, nationality, region and whether the person lives together with a partner. Table 1 shows the coefficients of interest only: social class or income quartile and the interactions between class and 5-year period (for the full model, see tables A.2 and A.3 in the appendix). For Germany, we find a strongly positive effect of being in a higher income quartile or in a higher class as compared to being in the lowest income quartile or the lowest class. Measured with our 11-point life satisfaction scale, the mean difference between people in the lowest and the highest income quartile is 0.39 points, between people in the lowest and the highest class 0.25 points, net of differences due to sex, age, nationality, region, or partnership status. Note that the happiness gap between people with and without a partner is also 0.25 points in Germany. For Switzerland, the differences in life satisfaction are smaller and only significant between income quartile with the top quartile being 0.14 points more satisfied than the bottom quartile. In both countries, we find consistently negative coefficients for age and periods, confirming the general downward trend in life satisfaction in both countries that we observed above. However, the interaction terms are only significant between periods and class in Germany. Compared to the upper-middle class and the pre-reunification period , life satisfaction of the low-skilled working class decreased more strongly in (-0.70 points), (-0.15) and (-0.11). For Switzerland, the interaction terms are small and not statistically significant, suggesting that there was no-class specific trend in life satisfaction since About here Table 1 In a last analysis, we tested the effect on life satisfaction of three variables that possibly moderate the widening class gap in life satisfaction in Germany: the evolution in income inequality, aggregate unemployment, and job insecurity. (see table A.4 in the appendix). The results of these additional models suggest that a region s level of income inequality and unemployment rate reduces workers subjective life satisfaction. However, these two variables do not explain why the class gap in life satisfaction has widened over time the main effect of class and the interaction terms between class and period remain unchanged. This is not surprising insofar as it is unlikely that inequality and unemployment per se, at a region s level, have a major impact on workers life satisfaction. Rather, it is the consequences on people s individual lives of inequality and unemployment that should reduce life satisfaction: wage stagnation, an unemployed family member or greater job insecurity. In 13

14 effect, subjective job security is associated with higher life satisfaction. And introducing subjective job security into our model, both the main effect of class and the interaction of class with period decrease by about a fifth. Clearly, lower classes are more worried about keeping their job than higher classes and this gap in worries has widened over the 2000s (see table A.4 in the appendix). At the same time, even when controlling for the class disparity in subjective job security, we still obtain a class gap in life satisfaction that is almost as large as the happiness gap between the partnered and unpartnered respondents in the German sample. Conclusion This paper started out with the argument that the 1990s and 2000s were a gloomy period for Germany s working class. The period between the end of the post-reunification boom in 1993 and the late 2000s when Germany s labour market finally recovered were marked by mass unemployment, a spread of atypical jobs, welfare retrenchment and increasing income inequality. When economic insecurity increases and the influence of collective pay-setting institutions decreases, workers with little individual bargaining power the low-skilled working class are likely to suffer most. The expected outcome is a rising class gap in life satisfaction. However, Germany s economy continued to expand over most of the 1990s and 2000s. Despite several cyclical slumps, German GDP increased by almost a third between 1992 and A society that gets richer has more resources to spend on public goods, notably health care, education and public infrastructure. Accordingly, two decades of non-negligible growth may have led, in Germany as elsewhere in Western Europe, to a reduction in the class gap in life satisfaction. Our data provide an unambiguous answer to these two competing hypothesis: inequality trumps growth. The class gap in life satisfaction and subjective health has widened in Germany between the early 1990s and the late 2000s. Regardless of whether workers hierarchical position in the labour market is measured with income quartiles or occupational classes, life satisfaction evolved more favourably among people in the top quartile and the highest class than among those in the bottom quartile and lowest class. Only after 2008, when the German labour market finally began to recover and unemployment rates declined, do we observe a slow decline in the class gap in life satisfaction. A similar conclusion is reached when we look at subjective health instead of life satisfaction: high earners and the upper- 14

15 middle class rate their health more positively than low earners and the working-class and the class gap in health constantly widens after the mid-1990s. Germany s widening class gap in subjective well-being also stands out when compared to Switzerland. In Switzerland, classes and income quartiles also vary in their life satisfaction and subjective health, but the difference is smaller and we do not observe a trend towards an increasing disparity over time. Germany s class gap in subjective well-being is not trivial; our multivariate analysis suggests that the difference in life satisfaction between the highest and the lowest income quartile, the class at the top and that at the bottom, is as large or larger than that between (the more satisfied) people who have a partner and (the less satisfied) people with no partner. However, our attempt to explain the differential evolution in life satisfaction by introducing context measures for income inequality or unemployment proves ineffective. The class gap in life satisfaction does not disappear once we account for income inequality or unemployment at the aggregate level. We know that the class gap in life satisfaction widened in Germany after the early 1990s. But the jury is still out over the exact reasons why Germany s working class was left behind in terms of life satisfaction and subjective health. 15

16 References Alderson, A. and Nielsen, F. (2002). Globalization and the Great U-Turn: Income Inequality Trends in 16 OECD Countries, American Journal of Sociology 107 (5): Antonczyk, D., Fitzenberger, B. and Sommerfeld, K. (2010). Rising wage inequality, the decline of collective bargaining, and the gender wage gap, Labour Economics 17: Becchetti, L. Massari, R. and Naticchioni, P. (2013). The drivers of happiness inequality: suggestions for promoting social cohesion, Oxford Economic Papers 66(2): Berman, E., Bound, J. and Machin, S. (1998). Implications of Skill-Biased Technological Change: International Evidence. Quarterly Journal of Economics, 113(4): Bosch, G. (2009). Low-wage work in five European countries and the United States. International Labour Review, 148(4): Bosch, G. and Kalina, T. (2008). Low-Wage Work in Germany: An Overview. In: Bosch, G. and Weinkopf, C. (eds). Low-Wage Work in Germany, New York: Russell Sage: Biewen, M. and Juhsasz, A. (2012). Understanding rising income inequality in Germany, 1999/ /2006, Review of Income and Wealth 58(4): Clark, A., Diener, E., Georgellis, Y. and Lucas, R. (2008). Lags and leads in life satisfaction: A test of the baseline hypothesis. Economic Journal, 118: Clark, A., Flèche, S. and Senik, C. (2014). Economic Growth Evens-Out Happiness: Evidence from Six Surveys, CEP Discussion Paper 1306, London School of Economics. Delhey, J. and Kohler, U. (2011). Is happiness inequality immune to income inequality? New evidence through instrument-effect-corrected standard deviations, Social Science Research 40 (2011): Dutta, I. and Foster, J. (2013). Inequality of happiness in the U.S.: , Review of Income and Wealth 59 (3): Erikson, R. and Goldthorpe, J. (1992). The Constant Flux. Oxford: Oxford University Press. Erikson, R. and Goldthorpe, J. (2010). Has social mobility in Britain decreased? Reconciling divergent findings on income and class mobility, British Journal of Sociology 61(2): Ferrer-i-Carbonell, A. and Frijters, P. (2004). How important is methodology for the estimates of the determinants of happiness? Economic Journal, 114: Grabka, M. and Kuhn, U. (2012). The evolution of income inequality in Germany and Switzerland since the turn of the millennium. SOEPpapers on Multidisciplinary Panel Data Research 464, DIW Berlin. Hajdu, T. and Hajdu, G. (2014) Reduction of Income Inequality and Subjective Well-Being in Europe, Economics Discussion Paper Halaby, C. (2004). Panel Models in Sociological Research: Theory into Practice, Annual Review of Sociology, 30: Huijts, T., C. Monden & G. Kraaykamp (2010). Education, educational heterogamy, and selfassessed health in Europe. A multilevel study of spousal effects in 29 European countries, European Sociological Review, 26, 3:

17 Kolev, A. and Saget, C. (2010). Are Middle-Paid Jobs in OECD Countries Disappearing? An Overview. International Labour Office Working paper 96, Geneva. Krugman, R. (2008). Trade and Wages, reconsidered. Brookings Papers on Economic Activity, Spring: Madden, D. (2011). The Impact of an Economic Boom on the Level and Distribution of Subjective Well-Being: Ireland, , Journal of Happiness Studies 12: OECD (2011b). Divided We Stand: Why Inequality Keeps Rising. Paris: OECD. OECD (2011a). Growing income inequality in OECD countries: What drives it and how can policy tackle it? Paris. Oesch, D. (2006). Redrawing the Class Map: Stratification and Institutions in Britain, Germany, Sweden and Switzerland. Basingstoke: Palgrave Macmillan. Oesch, D. (2011). Swiss Trade Unions and Industrial Relations after 1990: a History of Decline and Renewal. In: Trampusch, C. and Mach, A. (eds), Switzerland in Europe. Continuity and Change in the Swiss Political Economy, London: Routledge: Ovaska, T. and Takashima, R. (2010). Does a Rising Tide Lift All the Boats? Explaining the National Inequality of Happiness, Journal of Economic Issues 44 (1): Oishi, S., Kesebir, S. and Diener, E. (2011). Income Inequality and Happiness, Psychological Science 22(9): Senik, C. (2009). Income Distribution and Subjective Happiness: A Survey, OECD Social, Employment and Migration Working Papers, No. 96, OECD Publishing. Stevenson, B. and Wolfers, J. (2008). Happiness Inequality in the United States, Journal of Legal Studies 37: S33-S79. Streeck, W. (2009), Re-Forming Capitalism, Oxford: Oxford University Press. Voorpostel, M., Tillmann, R, Lebert, F.,, Kuhn, U., Lipps, O., Ryser, V.-A., Schmid, F., Antal, E., and Wernli, B. (2014). Swiss Household Panel User Guide , Wave 15, Lausanne: FORS. Veenhoven, R. (2005). Return of inequality in modern society? Test by Dispersion of Life- Satisfaction Across Time and Nations, Journal of Happiness Studies 6: Visser, J ICTWSS: Database on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts in 34 countries between 1960 and Accessed at: Visser, J. (2013). Wage bargaining institutions from crisis to crisis, European Commission Economic Papers 488, Brussels. Wage Bargaining Institutions Wagner, G., Frick, J. and Schupp, J. (2007). The German Socio-Economic Panel Study (SOEP) Scope, Evolution and Enhancements. Schmollers Jahrbuch, 127:

18 Figures Figure 1: the evolution of income inequality in West Germany and Switzerland (LOWESS) Fig. 1.1 Hourly earnings by income quartile, West Germany st quartile 3rd quartile 2nd quartile 4th quartile Fig. 1.2 Hourly earnings by income quartile, Switzerland st quartile 3rd quartile 2nd quartile 4th quartile Fig. 1.3 Hourly earnings by class, West Germany Low working lower middle skilled working upper middle Fig. 1.4 Hourly earnings by class, Switzerland Low working lower middle skilled working upper middle The figure shows the evolution of the median hourly earnings of each quartile and each class in nominal terms (not corrected for inflation). For better comparison between Germany and Switzerland, the hourly nominal earnings is set at 100 in 1999 for the 1 st quartile (figures above) and the low-skilled working class (figures below). 18

19 Figure 2: the evolution of subjective life satisfaction in West Germany, (LOWESS) Fig. 2.1 Life satisfaction by wage quartile 1st quartile 2nd quartile 7.8 3rd quartile 4th quartile Fig. 2.2 Life satisfaction by class Low working skilled working 7.8 lower middle upper middle Fig. 2.3 Difference in life satisfaction between the 4 th and 1 st quartile Fig. 2.4 Difference in life satisfaction between the upper-middle and low-skilled working class

20 Figure 3: the evolution of subjective life satisfaction in Switzerland, (LOWESS) Fig. 3.1 Life satisfaction by wage quartile 1st quartile 2nd quartile 8.2 3rd quartile 4th quartile Fig. 3.2 Life satisfaction by class Low working skilled working 8.2 lower middle upper middle Figure 3.3 Difference in life satisfaction between the 4 th and 1 st quartile 0.4 Figure 3.4 Difference in life satisfaction the upper-middle and low-skilled working class

21 Figure 4: the evolution of self-reported health in West Germany, (LOWESS) Fig. 4.1 Self-reported health by wage quartile 1st quartile 2nd quartile 2.8 3rd quartile 4th quartile Fig. 4.2 Self-reported health by class Low working skilled working 2.8 lower middle upper middle Fig. 4.3 Difference in self-reported health between the 4 th and 1 st quartile 0.4 Fig. 4.4 Difference in self-reported health betw. the upper-middle and low-skilled working class

22 Tables Table 1: the effect of income quartile or class on life satisfaction (random-effects linear coefficients) Germany (SOEP) Switzerland (SHP) (1) (2) (3) (4) Income quartile (ref : 1st quartile) 2nd quartile 0.135*** 0.057** (0.032) (0.026) 3rd quartile 0.268*** 0.068** (0.034) (0.028) 4th quartile 0.393*** 0.135*** (0.036) (0.029) Class (ref : unskilled working class) Skilled working class 0.099*** (0.038) (0.047) Lower-middle class 0.105** (0.047) (0.048) Upper-middle class 0.250*** (0.054) (0.048) Having a partner 0.241*** 0.252*** 0.374*** 0.372*** (0.015) (0.015) (0.024) (0.024) Interactions quartile*period (ref: 1st quartile*84-89) 4th quartile* (0.042) 4th quartile* (0.044) 4th quartile* (0.043) 4th quartile* (0.044) (0.035) 4th quartile* (0.046) (0.038) Interactions class*period (ref: low-skilled*84-89) Upper-middle* (0.055) Upper-middle * (0.062) Upper-middle * (0.062) Upper-middle * ** (0.064) (0.054) Upper-middle * (0.066) (0.061) Constant 8.608*** 8.311*** 8.670*** 8.597*** (0.092) (0.093) (0.115) (0.120) Observations 192, ,609 52,860 51,766 Number of individuals 29,539 29,284 9,933 9,835 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Models controlled for gender, nationality, region, age, age squared, and 5-year periods (not shown). 22

23 Appendix Figures Figure A.1: the evolution of subjective health in Switzerland, (LOWESS) Figure A.1.1 Subjective health by wage quartile 1st quartile 2nd quartile 3.3 3rd quartile 4th quartile Figure A.1.2 Subjective health by class Low working skilled working 3.3 lower middle upper middle Figure A.1.3 Difference in self-reported health between the 4 th and 1 st quartile 0.20 Fig. A.1.4 Differ. in self-reported health betw. the upper-middle and low-skilled working class