The decline of the working-class vote, the reconfiguration of the welfare support coalition and consequences for the welfare state
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
556970
research-article2015
ESP0010.1177/0958928714556970Journal of European Social PolicyGingrich and Häusermann
Journal Of
European
Article Social Policy
Journal of European Social Policy
The decline of the working-class 2015, Vol. 25(1) 50–75
© The Author(s) 2015
Reprints and permissions:
vote, the reconfiguration of the sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0958928714556970
welfare support coalition and esp.sagepub.com
consequences for the welfare state
Jane Gingrich
University of Oxford, UK
Silja Häusermann
University of Zurich, Switzerland
Abstract
The central political claim of Three Worlds of Welfare Capitalism is that class actors, through the instruments
of the democratic process, can modify capitalism. Where working-class mobilization is strong, left parties
have sufficient electoral support in the political arena to alter markets politically in ways that decommodify
and thereby empower workers. The decline of traditional class voting, however, has profoundly changed
this dynamic of welfare politics. We show that the political support coalition for welfare states has been
reconfigured through two processes. On the one hand, the Left may have lost support among the traditional
working class, but it has substituted this decline by attracting substantial electoral support among specific
parts of the expanding middle class. On the other hand, the welfare support coalition has been stabilized
through increasing support for the welfare state among right-wing political parties. We discuss the possible
consequences of this ‘middle-class shift’ in the welfare support coalition in terms of policy consequences.
Keywords
Class voting, electoral realignment, partisan politics, social investment, welfare state
Introduction
The central political claim of Three Worlds of in the political arena grows, left political parties
Welfare Capitalism is that class actors, through the look to ‘decommodify’ labour, transforming labour
instruments of the democratic process, can modify
capitalism. For Esping-Andersen, where unions
Corresponding author:
and other actors organizing the labour market are
Jane Gingrich, University of Oxford, Magdalen College, High
powerful, they provide crucial electoral support to Street, Oxford OX1 4AU, UK.
their representatives in the political arena. As power Email: jane.gingrich@politics.ox.ac.ukGingrich and Häusermann 51
markets to serve workers’ aims. In this framework, the welfare state have simultaneously stabilized the
working-class power begets working-class power, welfare state in the face of declining working-class
as left parties both draw on the support of workers support for the Left, but also changed it. The post-
and, by means of policies, sustain and enlarge it. industrial welfare support coalition is predominantly
Class, however, as the basis for both political anchored in the middle class, which tends to prefer
mobilization and ensuing coalitions, has changed social investment and activation policies over tradi-
profoundly. When Esping-Andersen published Three tional redistributive policies.
Worlds in 1990, the industrial class structure had These claims contribute to the literature in two
already entered its economic twilight. Today, the key ways. First, despite widespread acceptance in the
size of the manual working class is dramatically literature on political parties and party systems that
smaller than in the post-war period, with service sec- traditional class voting is in decline, much of the wel-
tor employment outnumbering manufacturing in all fare state and comparative political economy research
European countries (Oesch, 2006: 31). At the same has largely neglected these dynamics and their conse-
time, social groups that cut across traditional class quences. Most work on the partisan politics of the
boundaries have emerged (Oesch, 2013). In short, welfare state continues to conceptualize the prefer-
the socio-economic structure that defined the growth ences and behaviour of left parties through the lens of
of the welfare state has dramatically changed. a traditional working-class constituency (for a critical
What are the implications of these trends for the assessment of this literature, see Häusermann et al.,
political coalitions around the welfare state in a post- 2013). Even work explicitly examining changes in
industrial setting and – consequently – for the wel- the electoral base of the Left, such as David Rueda’s
fare state itself? Can advanced welfare states survive (2007) analysis of insider–outsider divides within the
without a clear electoral foundation in the working Social Democratic electorate, theorizes changes
class? In Three Worlds, Esping-Andersen himself within the working class rather than the consequences
suggests an affirmative answer, arguing that the of its broader decline. Second, we argue that the Left
institutions of the welfare state would supplant, in has found a new electoral constituency in parts of the
some ways, their political origins. For Esping- middle class that substitutes for declining working-
Andersen (1985), more universal social programmes, class support. While much work has analysed the
such as those in the Scandinavian countries, would changed electoral behaviour of the (new) middle
sustain a coalition of blue- and white-collar workers classes (e.g. Kitschelt, 1994; Kriesi, 1998; Oesch,
around social programmes, thereby ‘endogenously’ 2008), the consequences of these shifts for welfare
stabilizing welfare support. politics remain clearly under-analysed. We show
In this article, we build on these arguments, argu- that while the ‘middle-class shift’ in the welfare
ing that there has been a substantial reconfiguration in support coalition has cemented new electoral foun-
the underpinning support coalition for the welfare dations for the welfare state, it has also led to a
state; however, these changes are more general and change in policy prioritization, away from tradi-
less regime-based than Esping-Andersen suggested. tional ‘pro-worker’ policies. In short, understanding
We build on a large body of research showing that the contemporary politics of the welfare state still
pro-welfare left parties have lost support among the requires attention to class, but it equally requires
working class. However, far from spelling an overall explicit attention to the changed class character both
decline of the pro-welfare coalition, we make two within the Left and the Right.
arguments. First, left parties have attracted substantial
electoral support among specific parts of the expand- Changing coalitions, changing
ing middle class, substituting for a shrinking working-
welfare states?
class base. Second, right-wing political parties have
moved towards more political support for the welfare In the late 1980s, when Esping-Andersen was writ-
state. This reconfiguration of support within the Left ing Three Worlds, the question of how de-industrial-
and the reconfiguration of the cross-party support for ization and the concomitant rise of new post-industrial52 Journal of European Social Policy 25(1)
social groups might reshape welfare support coali- however, he argues that the existing social structure
tions was far from clear. The literature on electoral does not fully determine either the interests or the
dealignment argued that class was becoming increas- strength of class actors. In Three Worlds, Esping-
ingly irrelevant for electoral choice; not only had Andersen (1990) argues that ‘we cannot assume
traditional class groups shrunk but also electoral that a numerical increase in voters, unionization, or
volatility had been increasing since the 1970s seats will translate into more welfare-statism’ (p.
(Dalton et al., 1984; Franklin et al., 1992). Indeed, 17). Instead, the electoral coalitions at the basis of
some analysts even predicted an eclipse of tradi- left power – and thus of comprehensive welfare
tional class voting in favour of more issue-oriented states – are political products, constructed by par-
voting (Dalton, 2008; Katz and Mair, 1995, 2009). ties and the coalitions they strike. In building sup-
While this work did not elaborate what such radical port for the welfare state, unions and Social
partisan dealignment would imply for welfare state Democratic parties needed to both mobilize inter-
politics, a clear implication was that as the working- nally and create coalitions with other groups. We
class vote would no longer serve as a reliable politi- contend that these assertions still hold: a reconfigu-
cal foundation for pro-welfare parties, dealignment ration of the pro-welfare support coalition implies a
could undermine support for the welfare state itself. reconfiguration of the welfare state itself.
However, more recent electoral research has con- Building on these insights inspired by both
tested this early characterization of an entirely vola- Esping-Andersen and the realignment literature, we
tile electorate. A newer body of work argues that contend that it remains fruitful to conceptualize the
class, conceived in terms of individuals’ occupation, politics of the welfare state in terms of societal class
continues to matter. This work, drawing on either coalitions or – more generally – class groups;3 in
better sources of longitudinal data (Knutsen, 2006) doing so, we develop our argument in three steps:
or – crucially – new conceptualizations of class and first, structural change has led to the expansion of a
class voting (Elff, 2007; Evans, 1999), shows that specific segment of the middle class with pro-wel-
class continues to predict political preferences and fare attitudes, as well as a rightward shift of the
vote choices (even if these preferences and the working-class vote. Both left- and right-wing parties
choices may have become different ones). Put differ- have mobilized these changing electoral potentials.
ently, there has been electoral class realignment, not Hence, these shifts have neither doomed the welfare
dealignment (Kitschelt and Rehm, in press). Hout state nor left it with a ‘classless’ base; rather, left par-
et al. (1995) argue, for instance, even where ‘tradi- ties have gained a new class support base, and right-
tional class-voting’ (i.e. the blue-collar vote for the wing parties have moved towards more pro-welfare
Left) has declined, ‘total class voting’ (i.e. the pre- positions. Second, the extent of this electoral recon-
dictive power of occupational class on vote choice) figuration of the welfare state support coalition var-
remains strong.1 This shift in the literature from a ies between different institutional regimes. And
focus on traditional to total class voting coincides third, it is likely to have considerable consequences
with a definitional shift in the understanding of class, for the welfare state, because the pro-welfare middle
one which we adopt. Contrary to earlier works on classes show particularly strong support for policies
class formation and mobilization, this work under- of activation and social investment, rather than tradi-
stands ‘class’ in terms of socio-structural groups tional income protection and decommodification
defined by a particular occupational task structure, policies. In the following, we discuss these three
which contributes to shape their resources, latent claims in turn.
interests and preferences (Oesch, 2006).2 First, structural change has driven the transforma-
What are the implications of these socio-struc- tion of European electorates. Deindustrialization,
tural shifts for the welfare state? Esping-Andersen’s tertiarization of the workforce and especially the
work suggests that the answer is political. For educational expansion have all led to a decline of
Esping-Andersen, the preferences of class actors employment in the skilled and low-skilled industries
are partly derived from their economic position; and massive ‘occupational upgrading’, that is, theGingrich and Häusermann 53
expansion of occupational performance in the high- the occupations of the members of the ‘new’ middle
skilled service sector (Oesch, 2013, in press). While classes are often closely related to the welfare state
early scholarship on the expansion of an educated itself. Thereby, welfare states have to some extent
(service sector) middle class, namely, by Goldthorpe created their own support coalition among the
(1982), speculated that it would be a conservative growing share of people whose employment is
social force, he was soon proven at least partly directly linked to the state, that is, both public sec-
wrong. Rather, the educated middle classes, particu- tor workers and those employed in the welfare ser-
larly in interpersonal service sector occupations vices such as health and education more generally
(Kitschelt, 1994), have become the main progressive (Heath et al., 1991).
force in European politics. Much work traces the Political parties have played a role in both driv-
way these shifts in the post-industrial economic ing and mobilizing these electoral shifts and trans-
structure have affected voter preferences. A promi- formations. As Esping-Andersen (1985) reminds us,
nent line of work investigating these changes empha- individuals with shared attitudes towards the state
sizes the rise of non-economic issues that rival for do not automatically form into organized political
voters’ attention. Attention to non-economic issues actors. While changes in the economic structure
is not new: scholars of electoral behaviour have long have both put a variety of non-economic issues on
noted the crucial role of religiously motivated con- the agenda and created a new cadre of middle-class
servative voting within parts of the working class voters employed in services with more economi-
(Bartolini, 2000; Lipset and Rokkan, 1967). What cally left-wing preferences, in order for these to
the ‘new politics’ literature identifies, however, is a manifest politically, parties needed to act on them.
transformed cultural conflict between post-material- We argue that parties of the Right and the Left have
ist and materialist values (Inglehart, 1984), libertar- shaped the way these economic changes have
ian–authoritarian values (Kitschelt, 1994) or emerged politically in crucial ways, simultaneously
universalism–particularism (Bornschier, 2010; stabilizing support for the welfare state and modify-
Häusermann and Kriesi, in press). Although these ing its underpinning support coalition.
differing terminologies reflect varying temporal and Parties of the Left first responded to these shifts
substantial conceptualizations of a ‘second’ ideolog- by targeting middle-class voters. Kitschelt (1994),
ical dimension, they all point to a new line of con- Müller (1999) and Kriesi (1998) were the first to
flict that cuts across the traditional class cleavage. systematically examine realignment in Western
They share the argument that the new middle classes European party systems. Collectively, they argue
have become the main electoral potential of the Left that voters in interpersonal service occupations have
for reasons of cultural, rather than economic-distrib- become the core electorate of the New Left, some-
utive attitudes. However, it is important to note that thing we investigate further below. While this con-
although these social groups prioritize issues of cul- stituency fostered in particular the rise of green
tural liberalism, they also hold clearly pro-welfare parties, it has also increasingly transformed the elec-
preferences (Häusermann and Kriesi, in press; toral profile of the traditional left parties.
Kitschelt and Rehm, 2014). On the right, Pierson (1996, 2001) influentially
Scholars debate the reasons for these pro-welfare theorized that the sheer electoral importance of the
attitudes of the new middle class. Some work empha- welfare state implied that mainstream parties would
sizes the interpersonal profile of new middle-class converge on a welfare state stabilizing position.
occupations as fostering egalitarian and solidaristic Moreover, culturally conservative parties have been
values (Kitschelt and Rehm, 2014); others stress that attracting a substantial number of working-class
the oftentimes atypical and more vulnerable occupa- votes over the past decades. These groups have
tional situation of these workers, many of whom are explicitly mobilized working-class voters with
women, may explain their interest in state interven- regard to conservative and authoritarian attitudes,
tion (Häusermann, 2010; Häusermann et al., 2014; arguably as a counter-reaction to the aforementioned
Marx and Picot, 2013), and yet others point out that moves by progressive movements of the 1980s54 Journal of European Social Policy 25(1)
(Bornschier, 2010). As more working-class voters middle-class for working-class support has not been
have moved to non-left parties for predominantly neutral. As argued above, new middle-class voters
cultural reasons, however, they potentially alter the support state intervention generally and redistribu-
economic preference profiles of these parties, too, tive policies specifically. In this sense, they do sup-
creating welfare state support among even the far- port income protection policies that aim at what
right (see Betz and Meret, 2012, or Mudde, 2007, on Esping-Andersen has famously called decommodifi-
welfare chauvinism). cation. However, there is a shift in emphasis on other
The political coalition underpinning the welfare policies. The recent welfare state literature has
state, then, has undergone two changes; the prime extensively analysed the emergence of ‘new social
defenders of the welfare state, left-wing parties, now policies’ (Bonoli, 2005), ‘social investment policies’
rely mainly on voters from the educated middle (Hemerijck, 2013; Morel et al., 2012) and the spread
classes, and at the same time, the rightward move- of ‘labour market activation’ policies (Bonoli, 2010)
ment of workers has created a more heterogeneous on the welfare reform agendas of European welfare
set of partisan actors supporting the welfare state. states, well beyond the Nordic countries. These poli-
Second, cross-sectional variation is important. cies deviate from traditional social policy instru-
While the shift towards growing service sector ments, because they aim at fostering labour market
employment and left-mobilization of the new middle participation rather than compensating for income
classes is occurring across all advanced welfare loss (Gingrich and Ansell, in press) and they pro-
states, its translation into a large welfare state sup- mote social inclusion through work and human capi-
port coalition does not occur in an institutional vac- tal investment, rather than transfers (Cantillon and
uum. The existing institutional structure of the Van Lancker, 2013). The main instruments are child-
welfare state itself crucially shapes the options for care services, education and active labour market
parties. On the one hand, the shift to a post-industrial policies. In this sense, they have ‘commodifying’,
employment structure is partly a product of existing rather than decommodifying objectives. Although
regime structures. The tertiarization of employment, there is not yet a consensus in the literature about the
the extent of public sector jobs and the overall size of precise profile and delimitation of social investment
the service class vary across welfare regimes. Indeed, and activation policies, existing evidence suggests
welfare regimes contribute to shifts in the underlying that the middle class is both the main beneficiary and
group structure. On the other hand, welfare regimes the main supporter of such a supply-side-oriented
also provide varying resources to left parties. Left welfare state (e.g. Van Lancker, 2013, on the distrib-
parties who long relied on extensive cross-class coa- utive effect of childcare policies). Geering and
litions have had more success in attracting the new Häusermann (2013) have looked at policy support
middle classes, reproducing their existing strength on the basis of European Social Survey (ESS) 2008
through new social groups. More precisely, we argue and International Social Survey Programme (ISSP)
that the reconfiguration of the welfare support coali- 2006 data, which provide information on support for
tion is most pronounced in the Social Democratic different types of social policies. They find that the
regimes – where both partly endogenous shifts to a ‘new middle class’ (i.e. high-skilled employees in
post-industrial economic structure and the existing interpersonal service occupations) has the highest
left mobilizing capacity are extensive – followed by predicted probability of supporting public childcare
the Christian Democratic countries, with less change and education, considerably higher than the indus-
in the Southern and Liberal regimes, where the trial working class. Conversely, support for provid-
expansion of the middle class has been less pro- ing the unemployed with a decent living standard
nounced (in the Southern case) or where the struc- and support for income redistribution are highest
tural power of the Left to mobilize new actors is among the industrial working class.
more limited (the Liberal case). Taking these preference profiles into account
Third, what are the implications of these shifts for implies that when building welfare state support
the welfare state? Within the Left, the substitution of around a new core of new middle-class voters, leftGingrich and Häusermann 55
parties in post-industrial contexts are under pressure meaning that Sweden, Finland and Austria are not
to move away from policies supporting traditional included until 1995 and Switzerland not at all.
income replacement and decommodification and Norway is included, but only from 1993. Moreover,
towards new middle-class reform priorities. At the not all countries participated in all six waves of the
same time, welfare politics is coalitional politics. ESS. We discuss the implications of these data limi-
Thereby, the coalition space around broader welfare tations where relevant.
reform draws on new combinations of actors, such as To measure vote choice over time, we combine
middle-class and employer-based coalitions in two sets of questions. The EB asks respondents both
favour of labour market activation or middle- and a retrospective vote choice question and a vote inten-
working-class coalitions in favour of encompassing tion question. The ESS asks respondents a retrospec-
reform strategies that involve both activation and tive vote choice question. Although the EB
compensation. retrospective vote choice question is more similar to
In sum, we argue that there is a new class profile at the ESS vote choice question, it was asked only
the core of left parties and a new space for cross-class infrequently. In the EB, the correlation between the
alliances around the welfare state, which has real prospective and retrospective question is quite high
implications for its future politics and distributive (r = 0.81, p < 0.01), leading us to use the vote inten-
policies. In this article, we show this argument in a tion item in the EB. This yields a time series of vote
number of steps. In the section ‘The decline of elec- choice from 1973 to 2012 with limited gaps.
toral support of the working class for the Left’, we We use the combined vote intention measure to
demonstrate that the working-class vote for the Left create a dichotomous ‘Left party choice’ variable,
has indeed declined almost uniformly across advanced limited to social democratic, green, socialist and
European welfare states. The section ‘Sources of wel- communist parties (“the Left”) as measured by the
fare state support’ turns to the two new stabilizing original EB ZEUS coding schema. A dichotomous
support bases: the rise of middle-class voting for the measure of voting for the main social democratic
Left and party system reconfiguration around the wel- party yields similar patterns when used as an alterna-
fare state. In both sections, we show that while these tive dependent variable (see Appendix 1 for coding
trends are common across countries, the existing choices).
regimes shape their implications in varying ways. In In order to measure class across the different sur-
the section ‘Implications of this reconfiguration of the veys, we create four dummy variables, based on
electoral support coalition for the welfare state’, we respondents’ occupation and education levels
argue that this reconfiguration has – albeit not unam- (Appendix 2 outlines the coding choices and their
biguous – policy implications, namely, leading to a reliability), dividing respondents into employers
less pro-worker welfare state. (small and large), middle classes, manual workers
and routine workers (low-skilled non-manual). We
The decline of electoral support exclude retired voters and the non-employed from
our analyses. While the ESS data allow for a more
of the working class for the Left varied operationalization of the middle classes, the
To investigate the electoral coalitions underpinning EB data do not contain continuous information
pro-welfare parties over the long term, we turn to regarding industry, sectors or work logics.
data for 15 countries from the 1970s to present, com- We first examine the evolution of working-class
bining the 1972–2002 Eurobarometer (EB) survey voting over time among employed voters. The
trend file with six waves of the European Social decline of traditional class voting, that is, working-
Survey (2002–2012). Together, these data allow us class support for the Left, is largely uncontroversial
to examine the evolution of class voting over a long (Knutsen, 2006), leading us to only briefly demon-
time period for a number of countries.4 The panel, strate these trends. To assess the class basis of vot-
however, is somewhat unbalanced. The EB was only ing, we first look at a classic measure of vote choice,
conducted in European Union member states, a modified Alford index. This index involves a56 Journal of European Social Policy 25(1)
simple difference between the percentage of the the difference between working-class voters and
workers voting for the Left and all other employed middle-class voters in their propensity to vote Left.
voters, using the above outlined definition of left- Except the Southern countries, this outcome is simi-
wing parties. lar in all regimes: there has been a reduction in the
We examine these results at both the country and size of the coefficient on working class over time.
regime levels. We follow Esping-Andersen’s regime At the beginning of the observation period, in the
categorization, adding a fourth Southern Category. early 1980s, working-class voters were clearly more
The Social Democratic regime includes Denmark, likely to vote Left than middle-class voters, espe-
Sweden, Norway and Finland; the Christian cially in Liberal and Scandinavian countries. While
Democratic regime includes Austria, Belgium, the working-class vote was always more split in the
France, Germany and the Netherlands; the Liberal continental regimes and also in Southern Europe
regime includes the United Kingdom and Ireland (Bartolini, 2000), workers were still clearly more
and finally, the Southern countries include Greece, likely to vote for the Left than the middle classes.
Spain, Portugal and Italy. However, by the late 2000s, this difference has com-
Figure 1 shows the results (using a 3-year moving pletely withered in Scandinavia and even turned neg-
average to compensate for small numbers of respond- ative (albeit not always significantly so) in liberal and
ents). The clear story that emerges from Figure 1 is continental welfare regimes, indicating that in some
one of declining class-based differences, with only cases middle-class voters are now more likely to vote
limited regime-based differences. In all welfare for the Left than working-class voters. Multivariate
states, even those with a traditionally strong work- analysis confirms the above results: time has a nega-
ing-class base of the Left, the difference between tive effect – meaning a reduction in the coefficient on
workers’ and non-workers’ support for the Left has working class – across the sample. This erosion is not
shrunk. less extensive in the Social Democratic regimes than
Because of concerns that descriptive data can be the Liberal or Christian Democratic, with only the
misleading in assessing class differences (Evans, Southern regime demonstrating a different, more sta-
1999), we further run a two-stage regression model ble pattern (Appendix 3).
to evaluate class voting. In the first stage, we run a
series of country-year level logistic regressions (this
Sources of welfare state support
process yields 407 separate country-year regressions
since 1980). In each regression, we regress the The previous section showed that the working-class
dichotomous left vote choice variable on a range of vote for communist, socialist, green and social dem-
variables, including a dummy for being a member of ocratic parties has clearly declined in both relative
the working class, a dummy for being an employer and absolute terms, especially in Social Democratic,
and a dummy for being a routine worker (middle- Christian Democratic and Liberal regimes, with a
class respondents are the reference group), a coun- somewhat more stable development in the countries
try-year standardized measure of income, age in of Southern Europe. It is true that the working class
years, gender and education. The first stage model is and workers’ parties always needed allies to support
kept relatively parsimonious because of missing data and develop the welfare state – such as the agrarian
on other relevant variables.5 parties in Scandinavia or the Christian Democrats in
Comparing the coefficient on working-class sta- continental Europe. Nevertheless, the political
tus over time (Figure 1b, again using a 3-year mov- organization of the working class was the crucial
ing average) reveals a similar pattern to the political foundation of most welfare states (Esping-
descriptive data. In the first stage, a positive value on Andersen, 1985, 1990; Stephens, 1979). Hence,
working-class status suggests that working-class what does this decline of working-class mobilization
voters are more likely than middle-class voters to for the Left imply for the political support of the wel-
vote for left parties. Over time, a move towards a fare state? What does its nearly uniform nature imply
zero or negative coefficient suggests an erosion of for differences among welfare states?Gingrich and Häusermann 57 Figure 1. Declining working-class vote: (a) Alford index and (b) coefficient on working class for vote Left. SD: Social Democratic; CD: Christian Democratic. When we turn from class differences in the pro- (see, for example, Gingrich, 2011; Häusermann, pensity to vote for the Left to overall Left support, 2010; Hemerijck, 2013; Palier, 2010; Pierson, 2001) we see two quite different patterns. Figure 2 shows – the overall size of the state in spending terms has overall support for left parties across regimes, as remained predominantly stable. well as support among the working class and middle How then do we account for the combination of a class. Despite declining working-class support, there near-uniform decline in working-class support for the has been neither a generalized decline of Left vote Left with an overall stability in support for the Left shares nor a cross-regime convergence in left and the welfare state? As explained above, we argue strength. The Left remains strongest in the South, that there has been a dual substitution in the support with similar levels of strength in the Social and coalition for the welfare state: at the individual and Christian Democratic regimes and lower support in partisan levels. First, while left parties have mas- the Liberal counties of the United Kingdom and sively lost working-class votes, they have gained new Ireland. Assuming that these parties of the Left constituencies among specific parts of the educated, remain a pro-welfare force (something we confirm pro-welfare middle class. Second, under pressure in Figure 4), Figure 2 suggests that the decline in from both electoral shifts, as well as institutional traditional class voting has not led to decline in polit- feedback, traditionally right-wing parties have moved ical support for the welfare state. towards more pro-welfare positions. We examine Furthermore, not only has the support for the Left these two sources of welfare support sequentially. remained remarkably stable, but so too have welfare In order to investigate the first ‘substitution’, states themselves. Although almost all welfare states within the electoral constituency of the Left, we have been extensively reformed – retrenched in return to the EB–ESS combined data. As Figure 2 some instances, expanded and recalibrated in others shows, the middle-class vote for the Left has
58 Journal of European Social Policy 25(1)
Percentages of Respondents Voting for a Left Party
.8
.8
.6
.6
.4
.4
.2
.2
0
0
1980 1990 2000 2010 1980 1990 2000 2010
Social Democratic Regimes Christian Democratic Regimes
.8
.8
.6
.6
.4
.4
.2
.2
0
0
1980 1990 2000 2010 1980 1990 2000 2010
Liberal Regimes Southern Regimes
% of Middle Class Vote % of Working Class Vote
% of Overall Vote
Figure 2. Percentage of workers, middle classes, routine workers and the employed electorate voting Left.
increased over time relative to the working-class Figure 3 shows, in line with our theoretical argu-
vote. In the Social Democratic welfare states, the ment, a clear story of electoral substitution. With the
middle-class vote for the cumulative Left is now exception of the Southern countries, the middle
almost exactly equal to the working-class vote, at classes are now a very substantially more important
slightly above 40 percent. Since the mid-2000s, the part of the Left’s employed base than workers. Left
share of middle-class voters choosing the Left in parties became predominantly middle-class based
Christian Democratic and Liberal regimes has actu- around 1990 in the Nordic, Continental and Southern
ally exceeded the working-class vote. Southern European countries. In 2012, across the countries
Europe, by contrast, shows a more stable pattern, included in the sample, there were three middle-class
with the working-class vote for the Left being clearly voters casting a ballot for a left party for every worker
above average across the entire time period and the doing so.6 Together, Figures 2 and 3 show that the
middle-class vote remaining stable (at the Left vote proportion of the middle-class voting for the Left has
share average). grown, and the expanding size of this group means
When we combine this information on the shift in that it has become the core base for the Left. While
the behaviour of these groups with the massive these shifts are occurring everywhere (even in the
changes in the structural size of the groups, the pic- Southern regimes where they are less pronounced),
ture that emerges testifies to a radical reshaping of there are regime differences. Both the propensity of
the electoral base of the Left. Figure 3 shows how middle-class voters to vote Left and the importance
the relative shares of working-class voters and mid- of the middle class for left parties are higher in the
dle-class voters in the total electorate of the Left Social Democratic and Christian Democratic regimes
have developed over time. Middle-class voters have than the Liberal and Southern regimes.
clearly become the largest share in the Left electoral Who are these new middle-class left voters? The
base in all regimes. literature suggests that they are largely drawn fromGingrich and Häusermann 59
Class Composition of Employed Left Voters
1
1
.8
.8
% of Left Base
% of Left Base
.6
.6
.4
.4
.2
.2
0
0
1980 1990 2000 2010 1980 1990 2000 2010
Social Democratic Regimes Christian Democratic Regimes
1
1
.8
.8
% of Left Base
% of Left Base
.6
.6
.4
.4
.2
.2
0
0
1980 1990 2000 2010 1980 1990 2000 2010
Liberal Regimes Southern Regimes
Working class voters Middle class voters
Figure 3. Changes in left parties’ electoral base.
a clearly identifiable group of professionals and pro-welfare parties’ employed base (16 percent of the
semi-professionals in interpersonal services; by total base, when we include retirees and those out of
contrast, professionals in managerial and technical the labour force), while workers constitute 21 percent
occupations continue to vote predominantly for the of the employed base (10 percent of the total base).
right (Geering and Häusermann, 2013; Häusermann When we look to further unpack the new service
and Kriesi, in press; Kitschelt and Rehm, 2014; groups, distinguishing between private and public
Oesch, 2008). Our EB–ESS combined data do not employees, we again face data constraints in terms
allow us to operationalize variation within the mid- of longitudinal analysis. However, two EB surveys,
dle class, but the ESS data from the 2000s bear out 1988 and 1994, and the ESS in the mid-2000 meas-
these claims. In the ESS sample, 51.2 percent of a ure employment sector. We use these items to create
combined group of socio-cultural professionals, a measure of public employment, including all gov-
semi-professionals and skilled service workers ernment and public sector employees and employees
voted for a left-wing party, compared with only of state-owned enterprises. Here, we again find
47 percent of higher and associate manual and rou- interesting regime variation regarding both the size
tine workers. By contrast, only 40 percent of high- of these groups and their electoral choice. The Social
skilled managers, associate managers and office Democratic regimes have clearly the highest share of
workers voted for the Left. We see, then, a remarka- public sector workers with around 35 percent in both
ble shift towards socio-cultural professional and time periods, followed by Christian Democratic and
skilled service workers as a core base for the Left. Liberal regimes, each with between 25 percent and
From the perspective of these parties, and across the 30 percent of public sector workers, and finally the
six ESS surveys, socio-cultural professionals and Southern regimes with a little below 20 percent on
skilled service groups amount to 28 percent of average. We find similar differences when looking at60 Journal of European Social Policy 25(1)
0 5 10 15 20
0 5 10 15 20
welfare expansion
welfare expansion
1970 1980 1990 2000 2010 1970 1980 1990 2000 2010
Social Democratic Regimes Christian Democratic Regimes
socialist/soc−dem green socialist/soc−dem green
agrarian conservative christ−dem. & cons. right−pop
right−pop liberal liberal
20
0 5 10 15 20
welfare expansion
welfare expansion
15
10
5
1970 1980 1990 2000 2010
0
1970 1980 1990 2000 2010 Southern Regimes
Liberal Regimes
soc/social−dem green
soc/social−dem green christ−dem conservatives
conservative liberal right−pop liberal
Figure 4. Party positions on welfare expansion over time, Manifesto data.
the share of respondents employed by the welfare We hypothesized above in terms of a second substi-
state more directly. We cumulate the share of work- tution effect that as workers move to the right, this
ers in public administration, education, health and could foster support among right parties for the wel-
social work for the mid-2000s (ESS data, NACE fare state, whereas the pro-redistributive attitudes of
sector coding) and find about 27 percent in new socio-cultural professionals would solidify wel-
Scandinavia, 23 percent in Liberal and Christian fare support on the Left.
Democratic regimes and only 13 percent in Southern In order to trace and categorize party positions
European countries. More importantly, our data over time, we use the Comparative Manifesto Project
show that support for the Left among both public database and aggregate political parties into party
sector workers and welfare state employees is clearly families (Appendix 4). The Comparative Manifesto
higher than average in all regimes, with a particu- Project provides systematic over-time coding of
larly strong difference in Scandinavia. Again, the party positions in election manifestos. We examine
Southern regimes are different since support for the the average position of parties of a party family on
Left from public and welfare state workers is not sig- the item ‘welfare state expansion’ (per 504).
nificantly different from the other employees. In line Although manifesto data arguably measure both sali-
with other work, these data confirm that the welfare ency and ideological positioning, they nonetheless
state shapes the occupational structure of societies, provide a continuous time series from the 1970s to
directly feeding back into its own support (Esping- 2010 that allows us to tap underlying party support
Andersen, 1993). for the welfare state. Figure 4 shows the average
As parties’ electoral base changes, the question position of the relevant party families on the issue of
arises of whether the parties themselves continue to welfare state expansion over time. High values indi-
represent clear stances vis-à-vis the welfare state. cate high levels of welfare state support.Gingrich and Häusermann 61
Two observations are important regarding cross-class support coalition for the welfare state. In
Figure 4: first, on aggregate, the position of left par- the Christian Democratic regimes, the left parties are
ties has not shifted to the right, despite the massive now even more strongly supported by the middle
inflow of middle-class voters to the Left. In Social class than by the working class, but again their
Democratic and Southern regimes, left parties have favourable position towards welfare state expansion
even shifted to the left, and the level of welfare state has remained stable. Here too, the Christian
support remained more or less stable in Christian Democratic and right-wing populist parties have
Democratic regimes. The Liberal regimes, however, shifted towards more welfare-supporting positions.
do not follow this trend; here, left parties have indeed It seems that the continental European countries
shifted towards less pro-welfare positions than since exhibit a pattern of reconfiguration that comes clos-
the early 1980s. est to the kind of status quo convergence Pierson had
The second important finding in Figure 4 is that hypothesized.
right-wing parties have indeed shifted towards more In Liberal countries, by contrast, the electoral
pro-welfare positions in most countries. This is par- shifts of workers to the right and middle-class voters
ticularly true for the conservative and liberal parties to the left have rather led to a more welfare-critical
in Scandinavia and for the Christian Democrats in convergence. The Left’s support for welfare state
continental and Southern Europe.7 Right-wing popu- expansion has declined over time and almost con-
list parties, where they exist, have also tended to verged with the conservatives on a comparatively
shift to more pro-welfare positions over time, espe- low level, which may be due to the smaller endoge-
cially in Christian Democratic regimes, but the nous institutional effects of the more residual liberal
development is less clear and certainly far from welfare state, and also to the majoritarian dynamic of
strong pro-welfare positions (as some probably pre- party competition in these countries. Finally, the
mature arguments about the rise of welfare chauvin- Southern countries show a more stable picture than
ism may suggest). the others: there is little electoral reconfiguration,
While the move of moderate right parties towards and party positions have also remained more stable,
more welfare-supporting positions may also be due although some right-wing parties have moved
to institutional feedback mechanisms, as hypothe- slightly towards more pro-welfare positions. This
sized by Pierson (1996, 2001), there is also a rather stability may result from the less programmatically
close correspondence of shifts in the electorate and structured political space in Southern countries,
party positions. Figure 5 shows the working-class where parties have long adopted more particularistic
vote for moderate and populist right parties over policies (see the contribution by Manow, 2015). As a
time. In Scandinavia and Continental Europe, over result, the working-class support for the Left remains
time support for moderate right parties among work- high and stable in these countries, but equally, the
ers has grown and, more recently, for right-wing Left has been more limited in its ability to attract
populist parties, as well. In Liberal and Southern additional constituencies. This picture suggests an
regimes, the patterns are more stable, with fewer interpretation of the Southern regimes as increas-
right-wing populist parties (except for Italy, but they ingly traditional insider welfare states.
do not mobilize the working class strongly).
What do these data imply for the reconfiguration
of welfare support coalitions across different Implications of this
regimes? In Scandinavia, the Left is now equally reconfiguration of the electoral
strongly supported by the working and parts of the support coalition for the
middle classes. Nevertheless, the positions of the
main left parties have become decidedly more pro-
welfare state
welfare over time. More importantly even, the mod- Does this dual substitution matter for the welfare
erate right-wing parties have also moved towards state itself? As developed in section ‘Changing coa-
pro-welfare positions so that there is a very broad litions, changing welfare states?’ of this article, a62 Journal of European Social Policy 25(1)
% of Voters Supporting Right Parties
1
1
.8
.8
.6
.6
.4
.4
.2
.2
0
0
1980 1990 2000 2010 1980 1990 2000 2010
Social Democratic Regimes Christian Democratic Regimes
1
1
.8
.8
.6
.6
.4
.4
.2
.2
0
0
1980 1990 2000 2010 1980 1990 2000 2010
Liberal Regimes Southern Regimes
% of Working Class Vote for Populist % of Working Class Vote for Moderate Right
Overall Moderate Right Vote
Figure 5. Support for right-wing parties.
changed electoral base might imply new policy pri- 1999) or ‘social investment’ (Gingrich and Ansell, in
orities for the Left and thus transformed welfare press; Hemerijck, 2013; Morel et al., 2012) – have
politics, even if on aggregate, we see stable support made the point that other, typically post-industrial
for left parties. In this section, we examine the evo- risk groups face different (or at least additional) risks
lution of labour market policy and family policy requiring distinct policies. These new social policies
across European welfare states to investigate whether comprise most clearly issues of work-care reconcili-
we see evidence for such a shift. ation. Left voters from the ‘new middle classes’ are
In the traditional class-power model, labour mar- highly sensitive to these policies, because for them
ket policy assumes a crucial role. For Esping- the type and extent of labour market participation
Andersen (1990) and other power-resource scholars, and family obligations are a key choice they have to
one of the central goals for left-wing parties involved make in the interest of their family’s welfare. In this
insuring against the risks of unemployment. Recent sense, family policy services can be seen as a typical
work on occupational risk and the insurance func- example of a more ‘middle-class’ welfare state
tions of the welfare state continues to focus on strategy.
unemployment-related policies, which are particu- Our expectation thus is that where left-wing par-
larly crucial to those with specific skills or occupa- ties have a relatively stronger working-class base,
tional risks (e.g. Bonoli, 2010; Gingrich and Ansell, they will tend to pursue more pro-worker policies
2012; Rehm, 2009; Rueda, 2007). (i.e. more generous unemployment insurance
By contrast, scholars focusing on the social poli- replacement rates). Where parties are competing
cies key to the post-industrial societies – notably the more directly for the middle-class vote, we expect
literature on ‘new social risks’ (Armingeon and more investment in family policy and services,
Bonoli, 2006; Bonoli, 2005; Esping-Andersen, reflecting the interests of a broadened constituency.Gingrich and Häusermann 63
In order to see whether the priorities of parties on 5-point scale of partisan control of cabinet (1 = right-
the Left shift as the composition of their base shifts wing hegemony to 5 = left-wing hegemony) with the
towards a more middle-class electorate, we construct modified Alford index discussed in section ‘The
a dataset combining the descriptive data on the com- decline of electoral support of the working class for
position of the voting base from the EB–ESS com- the Left’. We further include a measure of the over-
bined dataset with policy outcomes data. We focus all percentage of the working class that supports the
on two policies: unemployment replacement rates as Left. In the analysis of unemployment replacement
a typical welfare benefit in the direct interest of rates, we include controls for overall union density
working-class voters and parental leave replacement (Visser, 2011), the unemployment rate and gross
rates as a typical hedge against ‘new social risks’ and domestic product (GDP) growth (Armingeon et al.,
‘social investment’-policy instrument, which 2012). In the analysis of leave replacement rates, we
responds most directly to the needs and demands of include a measure of female labour force participa-
the educated middle classes. tion (OECD, 2013) and the percentage of the popu-
Data on replacement rates for unemployment and lation under 65 years and GDP growth (Armingeon
parental leave come from two sources. First, we et al., 2012)
draw on the Comparative Welfare Entitlements Table 1 shows the results. In line with expecta-
Dataset Version 2, which provides replacement rate tions, the interactive effect of the Alford index and
data for all 16 countries in our sample from 1970 to cabinet control is highly statistically significant for
2011 (Scruggs et al., 2014). The replacement rates unemployment replacement. Substantively, these
data have the advantage of providing a measure of results are quite strong: dominant Left cabinets with
generosity based on existing policy, rather than a predominantly working-class electorate are associ-
spending data, which is affected by current macro- ated with predicted unemployment replacement
economic conditions. Unemployment replacement rates that are more than two percentage points higher
rates for single ‘average production workers’ are a than those associated with left cabinets under the
standard measure of the generosity of the unem- control of parties that have lost their working-class
ployment insurance system. To capture a similar profile. These effects are yearly, so the cumulative
measure for family policy, we examine the com- effects are substantial.9
bined replacement for maternity and parental leave Columns 3 and 4 of Table 1, however, show that
for the first 26 weeks of leave.8 This measure pro- there is no significant interactive effect between cab-
vides a gauge of the generosity of parental leave inet control and the Alford index for parental leave
benefits (Gauthier, 2011). replacement rates.10
In order to provide some preliminary and illustra- Figure 6 shows the interaction visually using in-
tive evidence of the effects of the left parties’ con- sample predicted probabilities based on the first and
stituency profile on policy priorities, we run two third specifications. It shows that left governments
types of time-series specifications. First, we run a with high levels of concentrated working-class sup-
pooled regression with a lagged dependent variable, port appear to expand unemployment replacement
year dummies, and with error terms adjusted for one- rates more; by contrast, left governments with a
period autocorrelation and country-specific heter- strong middle-class voter base are significantly less
oskedasticity. Second, we run a country fixed-effects generous in terms of unemployment replacement
regression, also with a lagged dependent variable, rates. However, contrary to expectation, we find no
year dummies, and country-clustered standard errors. differences in the propensity to support parental
In both cases, we lag all independent variables by leave policies based on working-class concentration:
1 year. We linearly impute data for missing years. where left-wing parties have a strong middle-class
Our key interest lies in whether left parties basis (i.e. whether the Alford index is low), the level
behave systematically differently based on the of generosity in leave replacement rates is not sig-
degree of support from the working class. To meas- nificantly distinct to where its base is more concen-
ure this, we interact Armingeon et al.’s (2012) trated in the working class.64 Journal of European Social Policy 25(1)
Table 1. Determinants of unemployment and family policy replacement rates, in-kind spending.
(1) (2) (3) (4)
RR UE RR UE RR leave RR leave
Lagged DV 0.973* (0.0092) 0.914* (0.0356) 0.992* (0.0109) 0.863* (0.0303)
Union density −0.0001 (0.0001) −0.0007 (0.0004)
Alford index −0.121* (0.0367) −0.0697 (0.0509) −0.0336 (0.0363) −0.0111 (0.107)
Cabinet composition −0.0031* (0.0012) −0.0024* (0.001) 0.0005 (0.0013) −0.0007 (0.0012)
Alford*Cabinet 0.0384* (0.0098) 0.0340* (0.0101)
WC support for Left 0.00745 (0.0137) −0.0236 (0.0215) 0.0099 (0.0180) −0.0090 (0.0487)
GDP growth −0.00116 (0.0009) −0.0009 (0.0008) 0.0011 (0.0015) −0.0009 (0.0011)
Unemployment −0.0004 (0.0004) −0.0004 (0.0005)
Female LFP −0.0002 (0.0002) 0.0032* (0.0013)
Non-elderly population share 0.0019 (0.0036) −0.0007 (0.0036)
Constant 0.0221 (0.0133) 0.0951* (0.0322) −0.152 (0.126) −0.0013 (0.325)
Year dummies Yes Yes Yes Yes
Observations 377 377 382 382
R2 0.984 0.894 0.964 0.831
*significant at the 0.05 level.
RR: replacement rate; UE: unemployment insurance; DV: dependent variable; WC: working class; GDP: gross domestic control;
LFP: labour force participation.
Figure 6. Predicted values of unemployment and parental leave replacement rates for different levels of working-
class concentration in Left electoral support.Gingrich and Häusermann 65
While we do not find a mirror effect for the two has been pushed to more welfare-supporting posi-
policy strategies, we nonetheless think that the tions. In Continental Europe, there has been a narrow-
absence of an effect regarding family policy may ing of partisan differences around a pro-welfare
actually provide some evidence that declining work- equilibrium. In the Liberal regimes, we see a right-
ing-class support does shape policy priorities for the ward shift, as few left-wing competitors remain.
Left. Left governments with a strong working-class Finally, in the residual and particularistic welfare
profile do not lead to lower parental leave replace- regimes of Southern Europe, we see more stability
ment rates (as their voters also benefit from them), over time. Indeed, although declining class voting,
but left governments with a strong middle-class base which cuts across regimes, would seem to belie their
support this policy equally strongly. analytic utility, we confirm many of Esping-
Andersen’s early suspicions. Regimes have bred their
support base, and they have done so in distinct ways.
Conclusion However, these shifts do not imply stasis in the
The electorate in 2014 looks, in many ways, quite state itself; the new electoral coalition space has con-
differently to how it did in 1990 when Three Worlds sequences. Where the working class is less aligned
was published and most certainly to the period of to the Left, it allows less generous unemployment
welfare state expansion from the 1940s to the 1970s policies, but not less generous social investment pol-
at the heart of Three Worlds’ analysis. Traditional icies. Both dynamic suggest a key transformation in
working-class voters now make up a much smaller the state alongside its electoral constituencies.
share of the electorate, and indeed, the concept of
class itself has changed along with the socio-eco- Acknowledgements
nomic structure. Moreover, in this article, we have We would like to thank Patrick Emmenegger, Jon Kvist,
shown that the working-class vote for the Left, in Paul Marx, Klaus Petersen and the reviewers for helpful
both relative and absolute terms, has declined, a comments.
decline that has occurred across regime boundaries.
This erosion would seemingly portend quite radi- Funding
cal changes for the welfare state and for scholarship This research received no specific grant from any funding
on it. For scholars of power resources, in particular agency in the public, commercial or not-for-profit sectors.
Esping-Andersen, the electoral foundations of work-
ing-class power were crucial to the longevity of the Notes
welfare state. And yet, despite these shifts, the wel- 1. Many recent contributions have confirmed the
fare state remains in place. We are hardly the first to ongoing structuring power of occupational location
note this situation; indeed, Paul Pierson’s (1996, and occupational risk on political preferences and
2001) influential theorization of the ‘new politics’ choice (e.g. Bornschier, 2010; Häusermann, 2010;
pointed to precisely a replacement of old cleavages Häusermann and Walter, 2010; Kitschelt and Rehm,
with generic support for the welfare state. While the in press; Marx, 2014; Marx and Picot, 2013; Oesch,
findings here are, to some extent, supportive of these 2008; Rehm, 2011; Walter, 2010).
claims, they are also substantially distinct in empha- 2. We also deliberately use ‘too broad’ notions of work-
sis. We find a real decline in working-class support ing and middle class, which do not denote homogene-
ous occupational groups anymore (Oesch, 2006). We
for left parties – not simply a generalized levelling
do so in order to relate our argument to the earlier and
up of welfare support. This decline occurs partly broader literature on welfare state politics.
from a pro-welfare move by right-wing parties, but 3. One could argue that electoral realignment is rel-
also partly from a leftward move among middle- evant for welfare politics only if welfare issues have
class voters. remained salient for voters’ party choice. Except for
The result is shifting welfare state coalitions that the working-class vote for the populist right (which
look quite different from those of the past. In Social is predominantly culturally motivated), this assump-
Democratic countries, the entire political spectrum tion indeed seems to hold. At the individual level,You can also read
