556976 research-article2014

ESP0010.1177/0958928714556976Journal of European Social PolicyEsping-Andersen

Journal Of European Social Policy

Article

Welfare regimes and social stratification

Journal of European Social Policy 2015, Vol. 25(1) 124­–134 © The Author(s) 2014 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0958928714556976 esp.sagepub.com

Gøsta Esping-Andersen Universitat Pompeu Fabra, Spain

Abstract This study examines the conditions under which welfare state policies contribute to an equalization of the opportunity structure, focusing in particular on Scandinavia. Using data on inter-generational mobility and educational attainment, I find a clear equalizing effect in Scanidnavia that does not obtain elsewhere. The effect, however, is assymetric in the sense that it is almost exclusively a bottom-up equalization. Keywords Welfare regimes, educational attainment, social mobility, equality of opportunity

Preface When I was told of the plan to bring out a special issue of the Journal of European Social Policy to revisit my Three Worlds on its 25th birthday, I felt more honoured than at any other moment in my life. But then elation gave place to panic: what meaningful contribution could I possibly deliver? Re-examine my regime classification in light of 25 years of change? Defend myself against its many critics? Such options seemed utterly boring to me and then, one day, I saw the light: here was my great opportunity to attack the one great question that I never really managed to address in any serious way. The Three Worlds had two broad aims. The first was to identify the causes behind welfare state diversity; the second, to identify whether this diversity had any influential effect on the quality of citizens’ life chances. Examining the employment effects of distinct welfare state models, I claimed (at the end of Chapter 8) that they were the midwives of qualitatively different ‘post-industrial’ stratification

scenarios. On hindsight, these claims appear rather grandiose considering the paucity of relevant data I managed to marshal.1 So I shall exploit the unique opportunity that this issue of Journal of European Social Policy (JESP) grants me to revisit the stratification question once again. In order to trace the effects of welfare states on stratification, I obviously must adopt a comparative design, both diachronically (pre–post welfare statehood) and across nations. The analyses to follow will focus on Scandinavia (comparatively, however) due to its much more deliberate efforts to promote equality via social policy. I shall hone in on one dimension of stratification, namely the inter-generational transmission of social class. Corresponding author: Gøsta Esping-Andersen, Department of Political and Social Sciences, Universitat Pompeu Fabra, C. Ramon Trias Fargas 27, 08005 Barcelona, Spain. Email: [email protected]

Esping-Andersen To pre-empt my conclusions, I do find compelling evidence that the social democratic regime has effectively equalized the opportunity structure whereas this is not the case in either the conservative or liberal models I examine. Interestingly, this has almost exclusively been a ‘bottom-up’ achievement: enhancing upward mobility chances for workingclass offspring, while hardly touching the advantages bestowed upon the privileged classes. As to the main drivers, I argue that Scandinavia’s equalizing achievements are primarily the consequence of policies pursued for other egalitarian objectives, in prima facia democratization of the education system and female emancipation. Although welfare reforms have, historically speaking, routinely been launched in the name of equality, the core aim of all welfare states was social protection and income maintenance, not to reshape the class structure. Only in the closing decades of the 20th century can we identify a policy shift in favour of equalizing the opportunity structure – on one hand, via education reforms and efforts to diminish child poverty; on the other hand, via policies aimed at promoting female employment and greater gender equality. And such a shift was indisputably far stronger in the social democratic regime than elsewhere. Any effort to identify the causal impact of welfare states on stratification processes is made exceedingly difficult due to potential endogeneity. As I argued at length in Three Worlds, my three rival welfare regimes were the outcome of distinct types of cross-class coalitions. Different patterns of social stratification were historically the midwives of different welfare state conceptions. In the case of Scandinavia, for example, it is likely that strong social democracy, its unique welfare model and its egalitarian thrust are all the joint product of a common historical legacy: the real explanation is buried in the conditions that favoured both social democratic rule and the ‘peoples’ home’ policy model.

Identifying the sources of endogeneity There is a good case to be made that the Nordic countries were pregnant with social democracy long before its actual appearance on the political scene.

125 The agrarian question had been settled already in the mid-1800s and, as historians show, a ‘red–green’ coalition came far more naturally than elsewhere, not only because a large segment of the rural population faced precariousness, but also because nascent industry was relatively small scale, dispersed geographically and often located in rural areas (Baldwin, 1985; Misgeld et al., 1988). The universalistic design of social policies is, as Baldwin (1985) rightly emphasizes, a logical outcome of Scandinavian history and not really a social democratic achievement. But one can only stretch the endogeneity thesis so far. Would these countries exhibit the same degree of intra- or inter-generational equality that we see today had the welfare state looked different or had the social democrats not ruled for half a century? One way to answer this question is to compare the Nordic countries with others in a ‘pre-social democratic’ era. To this effect, we could focus on two dimensions. One, were the Scandinavian welfare states already more advanced than others, say in the 1930s or, for that matter, in 1960? If so, it would be hard to attribute their achievement solely to social democracy. As it turns out, the answer is ‘no’. In terms of social expenditures, the Nordic countries were international laggards until the 1960s – even lagging behind the United States (Esping-Andersen, 1990; Lindert, 2004). Two, were social class differentials less accentuated in the Nordic countries than elsewhere, pre-social democracy? Again, the answer tilts in favour of a ‘no’. Sweden boasted comparatively very high levels of earnings inequality and wealth concentration at the end of the 19th century (Bentzel, 1952; Gustafsson and Johansson, 2003; Korpi, 1983; Morrison, 2000). Historical data on income distributions do not portray the Scandinavian countries as more egalitarian in the pre-welfare state era. Measured either via top–bottom quintile ratios or Gini coefficients, income inequality matched that of Britain and was even greater than in Germany (Brandolini and Smeeding, 2009; Morrison, 2000). To exemplify, the top decile’s income share was 39.5 percent in prewar Sweden, compared to 36 percent in France. Moving forward to the 2000s, the Nordic countries boast uniquely low inequality levels. All the evidence suggests that this occurred after the 1960s, and

126 all the evidence suggests also that equalization was primarily achieved by raising the income shares of the bottom quintiles. The relative income position of the top quintile hardly changed at all. In fact, in the mid-1960s, Sweden was actually less redistributive than Germany (Brandolini and Smeeding, 2009).

Equalizing opportunities In this article, I focus on inter-generational mobility. One approach is to compare the class bias in educational attainment across nations and cohorts. Were origin effects as strong in Scandinavia as elsewhere for the pre-welfare state cohorts, that is, for those born prior to any genuine levelling accomplishments? And do we find divergence for subsequent cohorts? As we shall see, the Nordic countries looked very similar to the rest of the world in the pre-welfare state era in terms of social origin effects on educational attainment. They do, however, boast a significant equalization effect within the younger ‘welfare state’ cohorts. From the 1960s onwards, the Scandinavian social democrats were pressed to deliver real equality. Starting with the famous education reforms associated with the Olaf Palme era in Sweden, all three social democracies strove to eradicate class barriers throughout the educational pyramid. This was done by eliminating tracking (through upper-secondary levels), constructing a comprehensive school system, and by eliminating financial barriers throughout. As I shall argue, it is possibly a second reformulation of equality that – although not actually aimed at class inequalities – came decisively to democratize the opportunity structure. Primarily in the name of gender equality, the social democrats began in the 1960s to prioritize family policy and support for women’s employment. Within a couple of decades, they erected an essentially universal childcare system based on – and this requires emphasis – extraordinarily high quality norms in all respects. In parallel fashion, they introduced very generous parental leaves and child allowances. The result was, of course, more gender equality in terms of career opportunities. And this has decisive second-order effects for inequality since maternal employment is a truly effective guarantee against child poverty. In my

Journal of European Social Policy 25(1) previous work, I have estimated that the risk of child poverty falls by a factor of 3–4 when the mother works (Esping-Andersen, 2009; see also Rainwater and Smeeding, 2003). Additionally, participation in a universally high quality pre-school system is possibly far more decisive than education reforms in terms of equalizing the opportunity structure. Comparative mobility research routinely concludes that Sweden is an exception to the constant flux thesis (Breen et al., 2009; Erikson and Goldthorpe, 1992; Shavit and Blossfeld, 1992). Since these studies included only Sweden, the question has remained open as to whether Denmark and Norway, too, are exceptionally mobile. Most explanations of Swedish exceptionalism have stressed the educational reforms that unfolded from the 1960s onward – and in Denmark and Norway some years later. As I show below, the class bias in educational attainment has indeed diminished in all three countries. But whether this, in turn, also fostered more inter-generational mobility in terms of class positions is less clear (Erikson and Jonsson, 1996). In fact, the expansion of higher education may very well produce the opposite effect, as happened in Britain (Blanden et al., 2005). The empirical evidence leans towards a positive mobility effect. Jaeger and Holm (2007) show a clear effect for Denmark. For Norway, Bratberg et al. (2005) show that education reforms helped to promote (slightly) more inter-generational income mobility. Both Holmlund (2006) and Holzer (2006) conclude the same for Sweden, but both studies include important riders. Holmlund argues that the positive mobility effect was primarily the indirect – and unanticipated – consequence of declining assortative marriage brought about by the comprehensive schools. Holzer shows that the mobility dividend obtains for all but children of very low educated parents. We are beginning to understand that education reforms may be a necessary but clearly not sufficient precondition for effective life chance equalization. In fact, one of the major insights in recent scholarship is the decisive causal influence of the pre-school years for children’s educational attainment and, more broadly, life chances. This means that family conditions over-determine whatever impact education policies may have. If this is the

127

Esping-Andersen

gini

uk

elasticity

us

dk no rw ay f in la nd ca na da sw ed en ge rm an y fra nc e

0,6 0,5 0,4 0,3 0,2 0,1 0

Income Inequality and Intergenerational Income Mobility

Figure 1.  Income inequality and inter-generational income elasticities.

LIS: Luxembourg Income Study. Source: Gini coefficients are from LIS data; inter-generational income elasticities derive from Corak (2003).

case – and all available evidence suggests this – it is clearly erroneous to make any hard and fast distinction between the two principal kinds of egalitarian objectives in terms of strategy. In fact, ‘equality here and now’ is very much a precondition for equality of opportunities (and vice versa). This is recognized in recent economic models of inter-generational mobility, where the levels of inequality in the parental generation strongly influence the degree of mobility within the child generation (Corak, 2003; Solon, 1999). To illustrate, I have plotted inter-generational income mobility elasticities against countries’ Gini coefficients (see Figure 1). It is immediately evident that the Nordic countries – Denmark especially – promote far more mobility than others (except, surprisingly, Canada). And the differences are huge: the Danish parent– child income correlation is more than three times smaller than the American, British or French. But these summary measures can be misleading if, as is now well established, mobility patterns are non-linear: mobility tends to be quite extensive in the middle regions of the distribution, but there is generally far less up- or downward mobility from the very bottom or the top. The reason has to do with marginal effects of policies. The marginal value of a universal child allowance is far greater for lowincome families; the same goes for the effect of maternal employment. And as James Heckman and his colleagues have demonstrated, the same logic applies to early childhood programmes (Heckman and Krueger, 2003).2

In fact, the latter point helps explain why even the most ambitiously egalitarian education system alone is unlikely to weaken much the impact of social origin on opportunities. As is well established, class differentials in school preparedness are already manifest when children first enter the education system, and schools are not well equipped to remedy the problem. Later remedial programmes are costly and not particularly effective (Heckman and Lochner, 2000). The basic logic is depicted in Heckman’s learning-begets-learning model. Those who start strong will, over time, distance themselves from the rest; those who have a poor start will fall ever more behind. And the key point here is that conditions during early childhood are decisive. While the latter argument enjoys broad scholarly consensus, it is still not clear which conditions matter most. American research has for decades emphasized the adverse effects of lone motherhood and poverty (Biblarz and Raftery, 1999; McLanahan, 2004; Mayer, 1997). But as Mayer (1997) emphasizes, the poverty effect is potentially spurious if, as is likely, the real cause lies in those parental characteristics that, in the first place, explain why they are poor. The latter view lies at the heart of the Heckman thesis, namely that the primary cause of disadvantage in early childhood stems from inadequate cognitive and behavioural stimulus. If living conditions during childhood are decisive, there is a ready-made explanation for why Scandinavia excels in terms of equality of opportunity. As we shall see below, the eradication of child poverty is without doubt one of its crowning achievements. If instead it is cognitive stimulus that really matters, any levelling trends are more likely to be spurred by the universal high-quality pre-school institutions.

Equalizing educational attainment The International Adult Literacy Survey (IALS) data permit us to compare social origin effects on educational attainment across cohorts and countries. In Table 1, I compare Denmark, Norway and Sweden against Germany, the United Kingdom and the United States. I focus on three birth cohorts: born,

128

Journal of European Social Policy 25(1)

Table 1.  The odds of attaining upper-secondary education among children of low educated fathers. Comparing three cohorts (Logistic odds ratios). United States

United Denmark Norway Kingdom

1970s 0.115*** 0.185*** 0.449** cohort 1950s 0.097*** 0.153*** 0.248*** cohort 1940s 0.133*** 0.162*** 0.213*** cohort

Sweden

Germany

0.661*

0.320**

0.094***

0.447**

0.164*** 0.067***

0.205*** 0.091*** 0.098***

IALS: International Adult Literacy Survey. Data source: IALS. Cohort 1 is born after 1970; cohort 2, 1955–1964; cohort 3, in the 1940s. Controls for cognitive abilities, sex and immigrant status. * = 0.05; ** = 0.01; *** = 0.001 or better.

respectively, in the 1940s, 1950s to early 1960s and after 1970. The latter is the cohort for whom we should expect any major welfare state-cum-education reform effects. In Table 1, I present odds ratios for the likelihood of sons of unskilled fathers completing upper-level secondary education. Reference group for the odds ratios are sons of fathers with upper-secondary or higher education. The data show that for the oldest (pre-welfare state) cohort, social origin effects were slightly weaker in Denmark and Norway than in other countries. But in Sweden, the origin effects were similar to the German and actually stronger than in either the United Kingdom or United States. The results suggest that any equalization of educational attainment began with the middle cohort (born 1955–1964) but only really took effect for those born after 1970 (the odds double in both Denmark and Sweden). We find no such equalizing trend at all in Germany, the United Kingdom or in the United States. In other words, the 1940s ‘prewelfare state’ cohort faced roughly similarly inegalitarian opportunity structures in all countries. This is no longer the case for the 1970s ‘welfare state generation’: the Swedes are three times, and the Danes four times, as likely to attain higher education as are their German or American equivalents. Whether this was produced by the education reforms or by other factors, such as more income equality and less poverty, is of course an open question. Still, the data do add up to a decisive welfare state effect that is unlikely to be biased by endogeneity.

Table 2.  Child poverty is especially decisive.

Denmark Norway Sweden France Germany Italy Spain The Netherlands Canada United Kingdom United States

1

1970s

mid-2000s

NA 4.8 6.4 7.2 4.4 NA 12.7 2.7 4.4 6.0 19.3

3.9 4.9 4.0 7.9 10.7 18.4 17.3 9.1 15.1 14.0 21.2

LIS: Luxembourg Income Study. Data from LIS, using the 50 percent of median adjusted income poverty line.

Asymmetric equalization As will be recalled, equality of condition and of opportunities are functionally inter-dependent. Examining the patterns more closely, we see that equalization is rather one sided: lifting the bottom without touching the top. Considering that child poverty in particular is known to have seriously adverse consequences, this is obviously a key indicator. Table 2 presents comparative data on the incidence of child poverty, comparing present day societies (ca. 2005) with the past (ca. mid-1970s). This gives us a clearer picture of achievements as regards inequality of condition. And if we hone in on an exceptionally vulnerable group, namely lone-mother families, the differences become all the more accentuated. Using Luxembourg Income Study (LIS) data, the Nordic lone-mother poverty rate has remained low since the 1970s; in comparison, it has jumped from 22 to 33 percent in Spain, and remains stubbornly around 50 percent in the United States. What perhaps is most salient is the ability of the Nordic countries to resist the trend towards heightened child poverty that we find across so many Organisation for Economic Co-operation and Development (OECD) countries, perhaps most spectacularly in the Netherlands and United Kingdom. All this suggests considerable success in raising the relative position of the bottom. Now, is income

Esping-Andersen compression and low poverty the product of the welfare state? At first glance, not really. If we ignore pensions, the share of welfare state income transfers that benefit the bottom income quintile is 32 to 33 percent in all three Nordic countries which compares unfavourably with a 36 percent average for all OECD countries. Similarly, the Swedish top-quintile receives 11 percent of all, which is identical to the OECD average. Scandinavia’s more compressed income distribution has probably less to do with redistribution than with the greater work income within bottom-end households. And this, in turn, can best be ascribed to the success in minimizing unemployment and, more importantly, in maximizing labour supply of women generally and of mothers in particular. In other words, Scandinavia’s egalitarian outcomes are – arguably – primarily the second-order consequence of labour market policies, collective bargaining and of the unusually concerted genderequalization policies. This can be illustrated via a simple simulation exercise. Using the European Community Household Panel (ECHP) data, if we simulate Danish levels of female employment for Spain, the Spanish Gini would decline by roughly 12  percentage points (Esping-Andersen, 2009). Put differently if, as in most countries, female employment is concentrated within top-quintile households, the effect is inegalitarian. The trick is to maximize female labour supply at the bottom of the social pyramid.3

Inter-generational mobility The asymmetric, bottoms-up approach to equal opportunities should, as already Table 1 suggested, become most evident when we examine inter-generational mobility. The advantage of examining trends in class mobility is that we can implicitly ‘control’ for endogeneity. We can identify mobility with three kinds of measures. One is to examine the degree to which offspring’s income correlates with fathers’ (as in Figure 1). A second is to focus on origin effects on educational attainment (as in Table 1). And the third is to estimate inter-generational correlations of class position. As mentioned, researchers who estimate intergenerational income correlations have now firmly

129 established the presence of non-linearities (Bjorklund and Jantti, 2009; Couch and Lillard, 2004; Jantti et al., 2006). The core finding is that parent–child correlations are especially strong at the very top and bottom of the income distribution, and comparatively weak in the middle. Jantti et al.’s (2006) study is especially revealing since it shows a similar lack of downward mobility from the very top across many countries, whereas there are truly significant country differences in terms of upward mobility chances for those who come from the bottom. Children from bottom-quintile families in Scandinavia are more than twice as likely to be upwardly mobile as in the United States. And, yet, there is no greater degree of downward mobility from the very top in Scandinavia than elsewhere. To illustrate, Jantti et al. (2006) find that 25 percent of Danish (and Swedish) sons from the lowest quintile end up in the same quintile as adults. This is only 5 percentage points more than what we would have expected from pure random assignment. And it compares very favourably with the United States, where 42 percent persist in the same bottom quintile. But inherited privilege at the very top is basically identical in Scandinavia and the United States (in Denmark as in the United States, 36 percent of sons from the top quintile end up, as adults, in the same top bracket). One alternative is to examine how parental class position affects sons’ income attainment across different birth cohorts – as did Ermisch and Francesconi (2004). But in order to identify whether any observed change in mobility is policy induced – and if so by what policies – we need to identify to what degree such change is the product of direct effects (origin– offspring correlations) or indirect effects (mediated via educational attainment). To this aim, we require data that allow us to compare cohorts born before and after welfare state maturation. The shortcoming of virtually all social mobility studies is that their youngest cohorts are in fact quite old. Breen et al.’s (2009) study utilizes data that are far more recent than those used by Erikson and Goldthorpe (1992) or Shavit and Blossfeld (1992). And, yet, their youngest cohort was born in 1959–1966, which implies that childhood experiences evolved prior to the ‘egalitarianpush’ welfare state phase. Neither the expansion of

130 early childcare nor the major educational reforms would have had any real effect on this birth cohort. The European Union Statistics on Income and Living Conditions (EU-SILC) data (from 2006) include an inter-generational module that permits us to compare pre- and post-‘welfare state’ cohorts. I will focus on two: those born in the period 1945– 1957 and those in 1968–1977. In a first step, I examine indirect effects: how social class origins influence the probability that sons attain tertiary-level education. This, of course, presumes having already passed the upper-secondary level. A bi-variate probit estimation is therefore the most suitable in this case. The second step is to examine direct effects: how class origins influence class destiny. To identify origins, I use the Erikson, Goldthorpe and Portocarera (EGP) 5-class scheme and hone in on mobility from the top (salariat origins) and from the bottom (unskilled origins). In each case, the reference is the three ‘middle’ EGP-class origins.4 The measurement of class destination is wrought with trade-offs. Were we to use the EGP-class scheme also for destinies, we face the problem that the youngest cohorts (1968–1977) are often still at the early stages of their careers. This means that they are less likely to be found in higher level managerial or leadership positions simply for reasons of age. For these reasons, I prefer to identify ‘destination’ in terms of earnings quintiles, since the latter are measured exclusively in intra-cohort terms. For the sake of simplicity, I present only mobility matrices that show, respectively, quintile destinations for sons from salariat and from unskilled origins.5

Step one: Indirect origin effects (origin class → tertiary education) To identify indirect effects, I focus on tertiarylevel education. I use the EU-SILC data and hone in on the two extremes: originating from the lowest (i.e. unskilled) and highest (i.e. salariat) classes. The ‘middle’ EGP classes are, collectively, the reference for estimating the coefficients (see Table 3). The bi-variate probit models estimate tertiary and upper-secondary-level attainment simultaneously. Table 3 shows only coefficients for tertiary level.

Journal of European Social Policy 25(1) Table 3.  Origin effects on attaining tertiary-level education (boys): Bi-variate probit estimation. Origins

Denmark

Lowest EGP class  45–57 −0.34** cohort  64–77 −0.11 cohort Salariat  45–57 cohort  64–77 cohort

Norway

France

Italy

Spain

−0.42***

−0.33***

−0.36***

−0.45***

−0.07

−0.39***

−0.17**

−0.25***

0.66***

0.49**

0.92***

0.63***

0.86***

0.68**

0.47**

0.51***

0.69***

0.56***

EGP: Erikson, Goldthorpe and Portocarera. Controls for mother’s education, sibling size, economic hardship during childhood, single-mother family and immigrant status. Reference classes for analyses are the three ‘middle’ EGP classes. * = 0.05; ** = 0.01; *** = 0.001 or better.

Here the evidence in favour of a ‘welfare state’ effect appears quite strong. If we begin by examining the old, ‘pre-welfare state’ cohort, the Nordic countries display strong social origin effects that do not deviate much from other countries – except that salariat origins had extraordinarily strong effects in France and Spain. The pattern changes quite evidently for the young cohorts almost exclusively because, in Scandinavia, unskilled origins now no longer have any statistically significant adverse effect on tertiary education attainment. The disadvantage has evidently abated also in Italy and Spain; in France, it persists. Turning to sons from the salariat class, we note that the relative advantages associated with privileged origins have not abated in either Denmark or Norway while, indeed, they did in both France and Spain – but we also note the extraordinarily strong class bias in these countries for the old cohorts. The upshot is that privileged origins now exert a stronger effect in Denmark than in either France or Spain. A basically similar conclusion emerges if we instead focus on the impact of financial hardship in childhood (not shown). In the old, ‘pre-welfare state’ Nordic cohorts, poverty had a highly negative influence on schooling that completely disappears later. In Spain, Italy and France, it continues to exert a strong negative influence also for the young cohorts.

131

Esping-Andersen

Step two: Direct origin effects (origin class → income attainment)

Table 4.  Inter-generational income mobility.

As is now well established, the influence of social origins tends to be far stronger in terms of shaping educational outcomes than in predicting final destinations. This is, at first glance, already evident when one compares the r-square (r2) values for origin– education and origin–destiny regressions. This is additionally confirmed by more formal decomposition techniques, such as the ‘Sobel-test’ (Sobel, 1982). In another study based on the same data and similar modelling procedures, we apply this technique and find that the relative weight of direct and indirect effects depends very much on the overall strength of the associations (Esping-Andersen and Wagner, 2013). In France, where origin effects are overall quite strong, the indirect effect overwhelms the direct effect. We find basically the same pattern for Italy and Spain. But the opposite is the case in the two Nordic countries where, as we already know, overall origin effects are comparably weak. In Denmark, as in Norway, the direct effect is far stronger than the indirect effect. To identify the direct effect, I adopt a simple inter-generational mobility matrix approach and examine how sons of, respectively, the lowest EGP (A) and salariat origins (B) are distributed across the five earning quintiles (see Table 4). The associations are not adjusted for possibly relevant controls and are, therefore, quite similar to those calculated in the Jantti et al.’s (2006) study – except that I use EGP-class origins rather than parental income. Table 4 reveals yet again the asymmetric pattern of Scandinavian social levelling. If we first examine sons of the lowest EGP class, the two Nordic countries have achieved a levelling not observed elsewhere. For the old, pre-welfare state cohort, the probability of landing in the bottom quintile was quite similar in all countries. For the young cohort, we find a significant decline in Scandinavia (most dramatically in Norway), while the distribution is unchanged in France, Italy and Spain. In comparison, Spain stands out as a society in which there is exceptionally little upward mobility from the

Income attainment of sons from lowest EGP origin Old cohort  Bottom 23 24 23 22   3 middle 63 60 63 62  Top 14 16 14 16 Young cohort  Bottom 21 18 23 22   3 middle 64 67 64 60  Top 15 15 13 18

Son’s income Denmark Norway France Italy Spain quintile

Income attainment of sons from salariat origin Old cohort  Bottom 19 15 14 12   3 middle 54 51 46 49  Top 25 34 40 39 Young cohort  Bottom 16 19 16 17   3 middle 53 57 46 52  Top 31 24 38 31

24 63 13 24 62 14

 9 41 50 11 50 39

EGP: Erikson, Goldthorpe and Portocarera.

bottom, a finding that echoes the earlier work of Julio Carabaña (1999). As we would expect, almost all mobility here is between the bottom and the three middle quintiles. Turning to sons of the salariat, we uncover a rather different scenario. For the old cohorts, the likelihood of ending up in the top quintile was far stronger in France, Italy and, especially, in Spain. Here, pre-welfare state Scandinavia (and Denmark in particular) does appear substantially more egalitarian than elsewhere. For the young cohorts, we see a clear equalizing trend throughout, most spectacularly in Spain and Norway. But we also see that privileged origins exert a stronger bias in Denmark. In fact, here the likelihood of ending up in the top quintile has risen by 24 percentage points in comparison to the old cohort. Both Denmark and Norway have been quite effective in equalizing opportunities for those who originate at the bottom of the social pyramid. Norway has undoubtedly also accomplished something similar for those from the top. But, ironically, the advantages

132 of privileged origins have strengthened in Denmark while they have clearly weakened elsewhere, producing in a sense cross-national convergence in terms of the salariat classes.

Concluding By and large, the data do suggest that the Scandinavian welfare regime has been, comparatively speaking, substantially more effective in equalizing the opportunity structure. And it has accomplished this feat primarily by enhancing the mobility prospects for those with humble social origins. In contrast, the privileges of being born into the top social strata appear pretty much unaltered. The magnitudes that separate the bottom from the top have been narrowed significantly, and this is surely a major accomplishment. How do we explain the socially skewed road to equality of opportunity? I can think of three, not necessarily mutually exclusive, logics that lurk behind these patterns. First, it may be attributable to political constraints. The perennial need for political coalition building compelled the social democrats to pursue a selective egalitarian strategy. The red–green alliance that underpinned the early welfare state project was, from the 1960s onward, replaced by a broad wage earner coalition, the core of which were manual workers and public employees. The wage earner coalition became the basis upon which social democrats could effectively pursue full employment policies and, especially, embark upon gender equalization. Indeed, the extraordinarily female-biased public sector in Scandinavia undoubtedly helps explain why the policy thrust has been so much in favour of gender equalization. In other words, the wage earner coalition logically promoted policies that would reap second-order egalitarian effects, be it in terms of early childhood programmes, democratizing access to education, active labour market policy or supporting working mothers. Hence, comprehensive schools and the abolition of tracking meant that workingclass kids attained upper-secondary-level education on par with the privileged; and, hence, employment intensity came to differ far less across households than in other countries; and, hence, the most vulnerable, like lone mothers, could pretty much support

Journal of European Social Policy 25(1) themselves. But why would these same coalition logics preclude any inroads into the relative privileges of the top social strata? Did social democracy need to pamper to the rich? Maybe they did not. A second possible explanation is that the top of the social pyramid always finds ways – and has the means – to counter any serious egalitarian moves by leapfrogging, so to speak. This is in fact an age-old explanation for why welfare state expansion never seemed to diminish the distance between the rich and the rest (Townsend, 1979). This would have been a ready-made explanation were it that the rich opted out of the system via private schools and universities, or via massive tax evasion. Such exit strategies are difficult to identify in the Nordic countries. This brings me to a third – and no less speculative – explanation. The top can reproduce social distances because, in large part, we are dealing with advantages that cannot be equalized via policy. And I believe that this is more the case today than it was in the past – the reason being that privilege today is increasingly associated with human (and cultural) capital. We know from mortality research that the social gap in life expectancy is very large – and widening: a professional will, on average, live 5–6 years longer than a manual worker. The way that pension systems are constructed implies that the net result is hugely regressive. Marriage behaviour is also polarizing, and on several fronts. Marital homogamy is on the rise, especially among the highly educated. This means that the top combines two high earners; the bottom, two low earners (at best), the net result obviously being that total household income differentials widen. Partnerships tend also to be far more stable among the highly educated; divorce and lone parenthood is increasingly concentrated among the low skilled. But perhaps most important are the non-tangible differences in how parents invest in their children’s life chances. The gap in parenting, both in terms of time invested and quality, is widening across the social classes. In particular, we observe that very highly educated parents hyper-invest in their children’s life chances. For Denmark, to exemplify, Jens Bonke and I have estimated that the total amount of parenting time is 50 percent greater in a highly educated couple compared to those with only

Esping-Andersen secondary schooling (Bonke and Esping-Andersen, 2011). The Nordic welfare states have undoubtedly lowered the barriers to educational attainment and social mobility for working-class kids. What exactly explains this success is not easy to pinpoint. But it is at least evident that there is not one single smoking gun. To a degree, it is undoubtedly the product of deliberate policy aims, such as was the case with education reforms. And in large part, it is the (largely unplanned) second-order consequence of policies aimed at altogether different objectives, in particular the pursuit of female employment. Funding This research was made possible by the author’s ERC Advanced Research Grant, and by Spanish Ministry of Economy Research Grant (CS02013-43461-R).

Notes 1.

At the time of writing the book, there were preciously little comparative data available on relevant stratification dimensions. 2. In a Denmark–United States comparison, I and my colleagues find that the educational boost of early childcare is highly concentrated among the most disadvantaged children (Esping-Andersen et al., 2012). 3. The Danish lone-mother employment rate is 81 percent (Esping-Andersen, 2009). 4. EGP refers to the Erikson, Goldthorpe and Portocarera classification scheme (see Erikson and Goldthorpe, 1992. The Statistics on Income and Living Conditions (SILC) data furnish only 2-digit International Standard Classification of Occupations (ISCO) coding for fathers’ occupation which prohibits constructing the more elaborate EGP classification. 5. The results from simple mobility matrices are highly consistent with those obtained from more complex multivariate estimations. See Esping-Andersen and Wagner (2013).

References Baldwin, P. (1985) The Politics of Social Solidarity. Cambridge: Cambridge University Press. Bentzel, R. (1952) Indkomstfordelingen I Sverige. Stockholm: Prisma. Biblarz, T. and Raftery, A. (1999) ‘Family Structure, Educational Attainment and Socioeconomic Success’, American Journal of Sociology 105(2): 321–65.

133 Bjorklund, A. and Jantti, M. (2009) ‘Intergenerational Income Mobility and the Role of Family Background’, in W. Salverda, B. Nolan and T. Smeeding (eds) Handbook of Economic Inequality, pp. 491–521. Oxford: Oxford University Press. Blanden, J., Gregg, P. and Machin, S. (2005) ‘Educational Inequality and Intergenerational Mobility’, in S. Machin and A. Vignoles (eds) What’s the Good of Education? pp. 99–114. Princeton, NJ: Princeton University Press. Bonke, J. and Esping-Andersen, G. (2011) ‘Family Investments in Children’, European Sociological Review 25(1): 1–14. Brandolini, S. and Smeeding, T. (2009) ‘Income Inequality in Richer and OECD Countries’, in W. Salverda, B. Nolan and T. Smeeding (eds) Handbook of Economic Inequality, pp. 71–100. Oxford: Oxford University Press. Bratberg, E., Nilsen, A. and Vaage, K. (2005) ‘Intergenerational Earnings Mobility in Norway’, Scandinavian Journal of Economics 107(3): 419–35. Breen, R., Luijkx, R., Muller, W. and Pollak, R. (2009) ‘Non-Persistent Inequality in Educational Attainment’, American Journal of Sociology 114(5): 1475–521. Carabaña, J. (1999) Dos Estudios Sobre Movilidad Intergeneracional. Madrid: Fundacion Argentaria. Corak, M. (ed.) (2003) Income Mobility between Generations. Cambridge: Cambridge University Press. Couch, K. and Lillard, D. (2004) ‘Non-Linear Patterns of Intergenerational Mobility on Germany and the United States’, in M. Corak (ed.) Generational Income Mobility, pp. 190–206. Cambridge: Cambridge University Press. Erikson, R. and Goldthorpe, J. (1992) The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon Press. Erikson, R. and Jonsson, J. (1996) Can Education Be Equalized? The Swedish Case in Comparative Perspective. Boulder, CO: Westview Press. Ermisch, J. and Francesconi, R. (2004) ‘Intergenerational Mobility in Britain. New Evidence from the British Household Panel Survey’, in M. Corak (ed.) Generational Income Mobility, pp. 147–89. Cambridge: Cambridge University Press. Esping-Andersen, G. (1990) The Three Worlds of Welfare Capitalism. Cambridge: Polity Press. Esping-Andersen, G. (2009) The Incomplete Revolution. Cambridge: Polity Press. Esping-Andersen, G. and Wagner, S. (2013) ‘Asymmetries in the Opportunity Structure: Inter-Generational

134 Mobility Trends in Scandinavia and Continental Europe’, Research in Social Stratification and Mobility 30: 473–87. Esping-Andersen, G., Garfinkel, I., Han, K., Magnuson, K., Wagner, S. and Waldfogel, J. (2012) ‘Child Care and School Performance in Denmark and the US’, Children and Youth Services Review 34(3): 576–89. Gustafsson, B. and Johansson, M. (2003) ‘Steps towards Equality’, European Review of Economic History 7(1): 191–211. Heckman, J. and Krueger, A. (2003) Inequality in America. Cambridge, MA: MIT Press. Heckman, J. and Lochner, L. (2000) ‘Rethinking Education and Training Policy’, in S. Danziger and Waldfogel, J. (eds) Securing the Future, pp. 47–86. New York: Russell Sage. Holmlund, H. (2006) ‘Intergenerational Mobility and Assortative Mating Effects of an Educational Reform’, Socialforskningsinstituttet Working Paper, 4. Stockholm: Stockholm University. Holzer, S. (2006) ‘The Expansion of Higher Education in Sweden and Equal Opportunities’, Center for Labor Market Research Working Paper. Vaxjo: School of Management, Vaxjo University. Jaeger, M. and Holm, A. (2007) ‘Does Parents’ Economic, Cultural and Social Capital Explain the Social Class Effects on Educational Attainment in Scandinavia?’, Social Science Research 36(2): 719–44. Jantti, M., Bratsberg, M., Roed, K., Raaum, R., Naylor, R., Osterbacka, E., et al. (2006) ‘American Exceptionalism in a New Light: A Comparison of Intergenerational Earnings Mobility in the Nordic Countries, the United

Journal of European Social Policy 25(1) Kingdom, and the United States’, IZA Discussion Paper, no.1938, January. Bonn: IZA-Bonn. Korpi, W. (1983) The Democratic Class Struggle. London: Routledge. Lindert, P. (2004) Growing Public. Cambridge: Cambridge University Press. McLanahan, S. (2004) ‘Diverging Destinies: How Children Fare under the 2nd Demographic Transition’, Demography 41(3): 607–27. Mayer, S. (1997) What Money Can’t Buy. Cambridge, MA: Harvard University Press. Misgeld, K., Molin, A. and Aamark, K. (eds) (1988) Socialdemokratins Samhalle. Stockholm: Tiden. Morrison, C. (2000) ‘Historical Perspectives on Income Distribution: The Case of Europe’, in A. Atkinson and F. Bourguignon (eds) Handbook of Income Distribution, pp. 217–61. Amsterdam: North Holland. Rainwater, L. and Smeeding, T. (2003) Poor Kids in a Rich Country. New York: Russell Sage. Shavit, Y. and Blossfeld, H.P. (1992) Persistent Inequality. Boulder, CO: Westview Press. Sobel, M. (1982) ‘Asymptotic Confidence Intervals for Indirect Effects in Structural Equation Models’, in S. Leinhardt (ed.) Sociological Methodology, pp. 290–312. Washington, DC: American Sociological Association. Solon, G. (1999) ‘Intergenerational Mobility in the Labor Market’, in O. Ashenfelter and D. Card (eds) Handbook of Labor Economics, Vol. 3A, pp. 1761– 800. Amsterdam: Elsevier Science. Townsend, P. (1979) Poverty in the United Kingdom. Harmondsworth: Penguin Books.