2.2.1 The Dual Geographic Transformation of the Knowledge Economy

Defining the ‘knowledge economy’ is complex, but it generally involves a shift to both higher-skilled forms of production involving new technological innovations (e.g. ICT) and high-skilled forms of service provision (e.g. finance, business services) (Boix Reference Boix2019; Hope and Martelli Reference Hope and Martelli2019; Iversen and Soskice Reference Iversen and Soskice2019). The growth of the knowledge economy has meant both a shift in the underlying economic structure and, through mass educational expansion and changes in the nature of the work, dramatic changes in the class structure (Oesch Reference Oesch2008b, Reference Oesch2013).

These economic and class transformations have not been geographically flat. Indeed, changes in the nature demands for skill, and the subsequent distribution of types of work (and their associated cultural preferences), have taken on an increasingly varied geographic character, both across and within countries.

On aggregate, much high-skilled employment growth in the last decades has occurred in both the public sector and in what Eckert et al. (Reference Eckert, Ganapati and Walsh2019) label ‘skilled tradeable’ services. While the former often occurs across geographic regions, the latter is less evenly geographically distributed both across Europe’s higher-level geographic regions and within these broad regions across cities and suburbs – leading to what I label the ‘dual economic transformation’ of Europe’s regions.

To take broad economic regions first. People live and work in areas that make or do different things. For instance, in the United States, the city of Houston, is part of the broader economic area of Houston–Woodland–Sugarland, an economic area that is historically more dependent on petroleum natural resource extraction. By contrast, the city of Detroit, located in Detroit–Warren–Dearborn area, is historically an industrial producer of automobiles and other manufactured goods. San Jose, part of the Silicon Valley area of California, has, for the last decades, been a leader in new technology firms. The broader economic structure of Houston, or Detroit, or Silicon Valley then, depends in part on competitiveness of the US energy, automobile and technology sectors (Boix Reference Boix2019).

The same is true in Europe – both within and across countries. Historically, there are distinct regional economies that generally have different economic strengths even within a given country (e.g. Herrigel Reference Herrigel2000). In the face of post-industrial economic change, some regions of Europe have moved more extensively towards employment in the knowledge economy than others.

To show this outcome, I follow much of the literature in European regional political economy and look at NUTS-2 regions. NUTS-2 regions are high level units, with between 500,000 and 3,000,000 inhabitants. NUTS units are both the basis of the distribution of parts of European structural funds (Becker et al. Reference Becker, Egger and von Ehrlich2010) and other work shows that they are meaningful economically and political relevant units (Colantone and Stanig Reference Colantone and Stanig2018b; Rodríguez-Pose Reference Rodríguez-Pose2018).

Figure 2.1 develops a measure of the regional structure of the knowledge economy across European NUTS-2 regions, using three indicators: the share of the 25–64 year-old population with a higher degree (defined as ISCED 5–8), the number of patents per 1,000 population and the share of employment in the NUTS-2 region in finance and business services (all data taken from Eurostat). Each indicator is averaged across decades, and rescaled to run 0-1, with the regions at the 99th percentile and above scored as 1. Each of the three components counts equally in the overall index, which is itself rescaled to run 0-1. This regional index correlates at the national level with the World Bank’s 2005 Knowledge Economy Index at 0.82.Footnote ¹ Figure 2.1 shows the distribution of regions across and within European countries (unfortunately, data on Switzerland is not available).

Figure 2.1 Knowledge economy index

What Figure 2.1 shows is that in many countries, there are regions that are extremely knowledge intensive – including in parts of UK, the Scandinavian countries, France, the Netherlands and Belgium. But that within these same countries, there are also much less knowledge-intensive regions. The growth of knowledge-intensive regions is, in some cases, associated with greater overall regional divergence, but in other regional divergence remains flat (Rosés and Wolf Reference Rosés and Wolf2019). However, where large and regionally redistributive welfare states have limited growing regional divergence – such as in the Scandinavian countries or the Netherlands – Figure 2.1 shows that many of the underlying structural trends are still present. By contrast, there are fewer knowledge-intensive regions in the European south and lower variation in the more industrial Germany and Austria.

However, there is also variation within regions. To return to the example of Houston, Detroit and San Jose. Within these broader economic areas, some people live in the centre of the core city, some live in a rich inner suburb or an (often poorer) exurb, and some live in an outlying rural area that is not directly linked to the city.

New economic sectors – finance, and parts of the knowledge economy are linked in particular to urban conurbations, with capital cities experiencing particular growth in these sectors (Odendahl et al. Reference Odendahl, Springford, Johnson and Murray2019). This sorting of high-skilled work into urban areas has both fuelled inequality among the high skilled – with workers in high-skilled jobs in high-skilled firms located, often, in high-skilled areas – being particular winners (Song et al. Reference Song, Price, Guvenen, Bloom and Von Wachter2019); at the same time, the sorting of high-skilled work into urban areas has had further knock on effects on other forms of inequality, such as housing wealth.

By contrast, both less knowledge-intensive industrial regions and new outlying suburban areas bordering knowledge-intensive areas may have higher concentrations of poverty or economic duress; the latter often including areas housing the lower-paid worker who service major cities but without the associated gains of urban growth (Kneebone and Garr Reference Kneebone and Garr2010). These areas have been hard hit by job losses and weaker rates of economic growth than the knowledge-intensive urban core (Rickard Reference Rickard2020).

Figure 2.2 draws a on different way of thinking about geography, based on the relationship between local units and cities. Many economists studying local economic effects examine areas defined ‘commuter zones’ or ‘travel to work areas’ rather than high level regions, as these areas are more closely linked to common economic experiences (Autor et al. Reference Autor, Dorn and Hanson2013). Figure 2.2 draws on an original dataset of highly localized data of educational attainment (Gingrich, McArthur and Cuibus Reference Gingrich, McArthur and Cuibus2023), aggregated to urban areas defined by commuter zone (broken apart by core city and suburbs) vis-à-vis those not attached to a commuter zone (rural and towns). The precise measures are discussed in the data section later.

Figure 2.2 Urban and rural areas

What Figure 2.2 shows is that in both knowledge and less knowledge-intensive regions, there are differences in the educational concentration of the population across cities and rural areas. In the Anglo-American economies and, to a lesser extent, the Scandinavian economies, economic growth in the last decades has generically benefited higher-skilled workers and particularly higher-skilled workers in either urban core areas or areas with high pre-existing levels of human capital (Moretti Reference Moretti2012). In cultural terms, Maxwell (Reference Maxwell2019) finds that major cities have become magnets for higher-skilled and culturally liberal voters, and substantial cultural sorting has occurred. Where the knowledge economy is less developed, either due to ongoing industrial or agricultural production, the gaps between the city and countryside are less stark, both because there are fewer centralizing urban pressures (i.e. de-concentrated forms of growth remain viable) and the gaps in economic and cultural experiences between the city and the countryside are less stark. Nonetheless, here too we see differences.

In other words, the geography of voters’ experiences differ both across types of economic growth models (knowledge-economy intensive or not) and their proximity to economic centres.

2.2.2 Changing Political Alignments

Do these shifts matter? Early work on the knowledge economy and party system change argued that changes in the class structure and nature of knowledge economy growth both created new pressures for social democratic parties. I argue later, that as the above-mentioned geographic shifts emerged, they magnified these difficulties.

The rise of knowledge-intensive work put pressure on social democrat’s traditional strategy of mobilization via broad based economic policy. Early work argued that the transition to the knowledge economy put increasing constraints on fiscal Keynesian demand side policies, limiting social democratic consumption spending, for example for pensions, health, and unemployment, while also making ‘supply side’ investment policies in skills to elevate the bottom half of the income distribution more imperative (Boix Reference Boix1998). At the same time, the rise of a new category of educated professionals – particularly in the sociocultural fields (education, culture, and health/wellness) – created a broad constituency calling for a societal liberalization in addition to demands for economic security and redistribution (Kitschelt Reference Kitschelt1994).

Social democratic parties that continued to primarily appeal to working-class voters on economic grounds were thus likely to face electoral decline. However, for modernizing social democratic parties, weaving together a coalition of their core base of voters with the new professional strata was increasingly difficult in the changing competitive space.

Subsequent work has vindicated many of these early claims. The changing class composition of the electorate has prompted a dramatic class realignment on the Left and Right (Kriesi et al. Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2006, Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2008; Kitschelt and Rehm Reference Kitschelt and Rehm2014, Reference Kitschelt, Rehm, Beramendi, Häusermann, Kitschelt and Kriesi2015; Gingrich and Häusermann Reference Gingrich and Häusermann2015; Häusermann and Kriesi Reference Häusermann, Kriesi, Beramendi, Häusermann, Kitschelt and Kriesi2015; Oesch and Rennwald Reference Oesch and Rennwald2018); the distinctiveness of traditional social democratic appeals – and policy – has declined (Huber and Stephens 2015; Raess and Pontusson Reference Raess and Pontusson2015), while moderation on new issues, particularly so-called ‘social investment’ has been critical to gaining new voters (Beramendi et al. Reference Beramendi, Häusermann, Kitschelt, Kriesi, Beramendi, Häusermann, Kitschelt and Kriesi2015; Abou-Chadi and Wagner Reference Abou-Chadi and Wagner2019) it has often been blocked by traditional labour market ‘insiders’ (Rueda Reference Rueda2005, Reference Rueda2007; Häusermann and Schwander Reference Schwander2012; Schwander Reference Schwander2012; Lindvall and Rueda Reference Lindvall and Rueda2014). The result has been an ongoing decline of social democratic parties and a competitive space that is increasingly fragmented not only across parties but also the dimensions of political competition.

However, as Boix (Reference Boix2019) and Iversen and Soskice (Reference Iversen and Soskice2019) argue, as these generic changes in skills and cultural values took on a geographic character that reinforced these broad shifts, sharpening the trade-offs for social democratic parties.

The growth of knowledge-intensive cities has the potential to further fragment the interests of working- and middle-class citizens across high- and low-productivity areas, making a singular policy appeal – whether on the demand or the supply side – increasingly difficult. This fragmentation follows from the geographic challenges that knowledge economy growth creates.

In one regard, regional economic divides are nothing new. In the postwar period, large parties, both on the Left and the Right, had to address deep economic disparities – Northern and Southern Italy, rural Wales and Manchester, and Stockholm and Northern Sweden, were all historically highly economically diverse. Both social and Christian democrats looked to appeal across regions with broad nationally based policies – with the expansion of the non-geographically targeted welfare state a method for doing so.

However, with the advent of knowledge society, there may be greater trade-offs. On the economic side, supply-side investments are likely to be less effective in lower productivity areas, absent some support for stimulating local growth. A redistributive supply-side policy (or ‘social investment’) is unlikely to meet the needs of these regions, where there are few high-paying employers demanding skills and where traditional industry is in decline. These deteriorating regions, therefore, demand social consumption policies – in the form of early retirement packages, higher pensions and greater unemployment payments – that fuel heterogeneity and disparity of wage earner demands across regions with different economic conditions. While working-class voters and professionals in knowledge-intensive cities often demand basic redistribution, they face specific pressure due to housing costs, childcare and transportation, making them more supportive of investment (Häusermann et al. Reference Häusermann, Pinggera, Ares and Enggist2019). Those in rural areas and outer suburbs, by contrast, face more pressure in terms of anaemic job creation, infrastructure investment and population aging. Creating a single economic package to address both – although not impossible (this was the attempt of US President Biden’s new revived industrial policy via the CHIPS and Science Act and Inflation Reduction Act) – often runs up against voters’ generic concerns about high taxation and social spending.

The same is likely to be the case for societal and cultural concerns with ecology, restrictions on automobile traffic, sociocultural pluralization and proliferation of art and entertainment. While urban areas and industrial districts – which Section 2.2.3 shows were often the traditional strongholds for the Left – were historically wealthy, the nature of production in these places has shifted away from traditional heavy industry with skilled crafts and manufacturing jobs, converting them into knowledge-intensive service job areas populated by professionals with tertiary education. As Rodden (Reference Rodden2019) shows, these shifts in the class structure produce a new cadre of urban voters increasingly willing to vote for the Left, but whose economic and sociocultural priorities may differ from those of the older outer suburban and industrial voters on questions of urban development and transportation, immigration, gender and climate issues. Taken together, these shifts mean that social democratic parties have a hard time to develop a unified strategy that appeals to all of these groups.

Put differently, there is a spatial disparity of the potential social democratic electorate between the new knowledge-intensive metropolitan areas and the more peripheral suburban and rural areas, but there is also a social and political disparity within each of these spaces between different potential constituencies which social democratic parties can no longer reach with the same appeals. There are unifying national issues left – such as support for an encompassing and generous national public pension system – but regional and group divides on other issues may overwhelm these bridging strategies.

2.2.3 New Geographic Competition

How geography shapes social democratic party strategies depends on the structure of party competition. This competition, in turn, is likely to vary across electoral system. Where electoral systems organize representation based on small geographical single-member constituencies with plurality or majority electoral formula, the strategic dilemmas of social democracy are often particularly intense.

In majoritarian countries, such as the US, Canada, UK and Australia, the electoral system magnifies geographic changes. While majoritarian systems insulate social democratic parties from new competitors, these electoral rules are associated both with both sharper geographic divides and a clearer transmission belt between geographic divisions and political conflict. As Iversen and Soskice (Reference Iversen and Soskice2019) argue, in (largely Anglo) countries with single-member district plurality systems parties have engaged little in redistributive policies that make regions reaping the benefits of the knowledge economy share them with those that do not. Majoritarian systems then, offer social democratic parties a particular dilemma. While the first-past-the-post system has protected them from shedding large proportions of votes to smaller radical left and green/left-libertarian parties, growing regional economic variation divides the social democratic base, making win-win urban-non-urban working-class coalitions hard to achieve. This pressure is compounded by urban malapportionment. As Rodden (Reference Rodden2019) argues, urban voters often deliver inefficient, over-sized electoral majorities to left parties, but few majorities in more suburban, peripheral districts. The result is that moderate left parties here have an increasingly urban base of victorious districts won with appeals conducive to attract professionals. But this makes it hard to win mixed suburban and rural districts with more working-class voters, with whom social democratic appeals that rally urban professionals will not resonate strongly.

Proportional systems would seem to avoid these problems. In these countries, the overall level of regional variation (and individual earning inequality) is much lower, with both larger welfare states, and more dispersed models of skill investment (through vocational training) continuing to redistribute resources and opportunities to more peripheral areas and the electoral system does not mechanically enhance urban–rural geographic disparities. Indeed, in such systems redistributive coalitions are more effectively viable (Iversen and Soskice Reference Iversen and Soskice2019), which may also cement more even interregional transfers (Beramendi Reference Beramendi2012). Moreover, proportional systems offer no (dis)advantage to concentrated support, hence they are not producing under-representation of left-party votes due to malapportionment across districts (Rodden Reference Rodden2019).

While systems of proportional representation enable parties to craft more cross-regional coalitions, and thereby reduce their electoral dilemma to serve different social constituencies, they also increase this dilemma in a different fashion. Because they pose low entry barriers to new contenders, such novel parties may seize upon locally concentrated political demands (De Vries and Hobolt Reference De Vries and Hobolt2020). After decades of cultural and economic sorting, parties representing a cross-region compromise face challenges from parties that are more directly targeted to the interests of particular areas. Social democratic parties can be outflanked in growing cities on the left by green-left parties catering primarily to sociocultural professionals, while in suburbs by parties of the Moderate Right or the Radical Right that pick up on traditional social democratic core voters’ lack of enthusiasm for libertarian societal innovations.

The result is that maintaining broad electoral support requires Social Democrats to compete with different demands in distinctive competitive spaces, both more or less knowledge-intensive areas and more or less rural and urban areas. Even when social democratic parties are broadly popular, they increasingly face distinct regional competitors that are more locally popular. The result is a loss on multiple fronts.

Table 2.1 lays out four different configurations. The competitive threat from the Green Left is likely to be strongest in knowledge-intensive cities (particularly those without a strong industrial past), which compositionally have more new professionals. In rural areas and outer suburbs of knowledge-intensive areas, particularly those with an industrial past, there will be more old-fashioned left voters disposed to follow Moderate Right or the Radical Right populist appeals to counter the rising influence of Green Left parties. Social democrats will be engulfed here by competition from all partisan directions.

Less knowledge-intensive urban and rural areas follow different patterns. Here, many wealthy rural semi-industrial areas remain, where moderate right parties continue to have strong support. There are many fewer large urban non knowledge-intensive areas (e.g. Naples) but many mid-sized cities fall into this category (e.g. the French mid-sized city of Dijon scores as much less knowledge intensive than Grenoble). In these areas, radical left parties, where they exist, have often mobilized in both cities and suburbs, picking up on young educated voters frustrated with their lack of economic opportunities.

In other words, the geographic shifts are part of the broader well-theorized class realignment around parties. Moving towards the urban strata of knowledge society professionals may alienate working-class wage earners from social democracy. The rise of knowledge society has the potential to magnify the tensions between both within-region as well as cross-regional divisions over economic and societal issue preferences, making it hard for social democratic parties to address these tensions with a single unifying appeal.

Table 2.1 summarizes the likely patterns of party system configuration, but glosses over complexities and overlaps. Consider industrially declining urban areas surrounded by a thriving knowledge economy. These may mitigate party alignments attributed to either strong or weak urban knowledge economies. The cells in Table 2.1 thus suggest relative (not absolute) patterns of competition, compared to other urban regional types.

In conjunction, then, the claim here is that in knowledge-intensive areas social democratic parties face a growing divergence between the competitive space across urban and non-urban voting base, with challenger parties targeting these particular geographic strongholds in different ways. In less knowledge-intensive areas, social democratic parties face the competitive challenge of retaining support in adjusting urban areas where Radical Left and Radical Right populists may be growing in support, while competing with more traditional parties in the rural periphery. Thus, even in proportional systems, Social Democrats face a challenge in holding together cross-geographic coalitions under the umbrella of a single mainstream left appeal.

2.2.4 Regimes of Vulnerability?

What do the above-mentioned claims imply for social democratic parties’ overall electoral vulnerability? It is difficult to generalize to the country level, given that there are many moving parts across competitive systems, however, a few implications from the preceding discussion emerge.

Historically, these parties mobilized voters in particular places. If social democratic parties face different competitive threats across economic structures, then all else equal, the extent of these threats to overall performance will depend on how vested they are in particular geographic economic structures. Structural changes thus may have different aggregate implications for Social Democrats depending on their historic strongholds of mobilization.

For instance, social democratic parties that historically mobilized in cities, will face more threat from the Green Left in knowledge economies than the Radical Right. But, where they historically mobilized in the countryside, they might face more relative threat from the Radical Right. Indeed, while Abou Chadi and colleagues’ chapter on vote switching suggest that the electoral threat of social democratic voters switching to the Radical Right is generally overstated, it may be that there are specific contexts, such as Sweden where mobilization was historically more rural and where knowledge economy divides are growing, the Radical Right does pose a threat to social democratic parties. In less knowledge-intensive economies, social democratic parties may face fewer competitive threats where they are strong in rural areas, however, those with stronger traditional supporting in urban areas or suburbs are likely to face more threat from new radical left parties. These features suggest that the optimal competitive strategies for social democratic parties vary across place: trying to outcompete the Green Left may be less crucial to them in Sweden than in the Netherlands, for instance.

These structural shifts are not fully determinative. Where social democratic parties mobilize cities or rural areas effectively, they may prevent the rise of competitors on the supply side. Latent vulnerability is not always manifest.

2.3.1 Where Did They Lose?

The first claim mentioned earlier is that social democratic parties, over time, lose in nearly all geographic areas, without picking up new areas. To examine the question of losses descriptively, Figure 2.3 plots the vote shares for the main social democratic party in the early 1980s against their vote shares in post-2010 for the most disaggregated unit available in sixteen countries (data at this level of disaggregation is only available from early the 1990s in France and unified Germany). The small number of units for which data in the 1980s is missing are assigned the average in the NUTS-3 region. For each country, the solid black line represents the relationship between the early 1980s vote share and post-2010 vote share by local unit, and the dashed black line is a 45° line (i.e. the early 1980s vote share plotted against itself). In nearly all cases, we see either a uniform cross-regional decline (as in Sweden), or an accentuated loss of support in historically strong regions (i.e. the dashed and black lines cross much closer towards the origin). In France and Germany, for instance, the traditional strongholds have shifted substantially away from the PS and SPD, respectively. In Italy, there has a been an even dramatic decline for the Moderate Left – driven by the rise of new populist radical right and radical left parties in both cases – even the traditionally strong ‘red belt’. In Greece, the historically strong PASOK, now KINAL, shows an even more extreme collapse in regional support. Thus, while there remains a high correlation in historic vote share and current vote share, social democratic parties have almost universally lost traditional strongholds without gaining new ones – in other words, we are not seeing geographic realignment in their support.

Figure 2.3 Voting in the early 1980s and 2010s

The patterns discussed earlier suggest, descriptively, that social democratic decline is in large part driven by both these parties’ inability to maintain voters in traditional strongholds and to build new forms of regional support. In many ways, this outcome is not surprising. Benedetto et al. (Reference Benedetto, Hix and Mastrorocco2020) find an on aggregate relationship between declining industrial shares and long run social democratic support, a finding that resonates with the above-mentioned long line of work at the individual level theorizing the relationship between changing class structures and a political dilemma of social democracy (Kitschelt Reference Kitschelt1994). The industrial structure – and its associated organization of economic and political life in trade unions – that gave birth to social democratic parties has changed, and so too have traditionally strong social democratic regions (Rodden Reference Rodden2019).

However, when viewed from another angle, the patterns are somewhat surprising. Social democrat’s traditional strongholds varied substantially across place from the wealthier industrial and urban regions of Germany, Italy and Switzerland to the poorer and more sparsely populated regions of Northern Scandinavia and Finland. Moreover, the transformation of these regional economies has been profound – and yet vote shares in many European countries correlate at above 0.80 in the 2010s to the 1980s. The continuity of geographic support then is as striking as the general overall losses in the post-financial crisis era. In other words, social democratic parties have lost everywhere, including their past strongholds, without picking up new areas. This trend is relatively universal across countries.

2.3.2 Whom Did They Lose To?

To whom have social democratic parties lost voters? In order to examine these trends, I begin descriptively, looking at urban–rural divisions – using the highly localized data aggregated to functional urban areas – across the included European countries. Figure 2.5 shows that across Europe, mainstream social democratic parties historically had a more urban base, with more substantial support in major and small cities, and to a lesser extent suburbs, than in rural areas and towns. Over time, this support has collapsed across all areas, with marginally more resilience in smaller cities than other areas. The Moderate Right shows the reverse geographic pattern, with stronger support in rural areas and towns and suburbs, with historically less relative support in major and smaller cities. It too has experienced across the board decline, albeit less dramatic decline in the post-2010 period. When we look at competitor parties on the Left and Right however, a starker geographic pattern emerges. Green and left-libertarian parties and radical left parties have a strong urban core, particularly in Europe’s major cities. Right populists, by contrast, while gaining in all geographic categories, have gained much less in major and small cities than outlying areas and suburbs. On aggregate then, there are stark urban–rural divides in new parties, with suburbs often looking more ‘rural’ than their associated cities.

Figure 2.4 Party family vote share by urban type

Figure 2.5 Regional patterns of competition, post 2010

There are some key exceptions to these trends. In the post-2010 period, the non-mainstream Left grew substantially in support in Spain and Greece, compressing geographic differences; and, both M5S and Lega in Italy have a strong urban base, including in the major cities and inner suburbs. In both the Netherlands and France, ‘major cities’ do not differ that much from the rest of the population because they are internally heterogenous – Amsterdam and Paris are very different from Rotterdam and Marseille. However, as Figure 2.4 shows, on aggregate, green/left-libertarian and radical left parties have picked up disproportionately in urban areas, whereas populist radical right parties have gained outside them.

Does the structure of the knowledge economy shape these urban–rural partisan divides? To provide a suggestive test of the claims, in Table 2.2 I run a series of linear regressions, regressing vote share for Social Democrats, the Moderate Right, Green Left, Radical Left, and Radical Right at the LAU level, on the NUTS-2 knowledge economy index (Figure 2.1) interacted with the three-part urban classification. Each model includes a control for logged population in the LAU, and NUTS-3 historic industrial structure in employment (1980 baseline), employment share in agriculture (averaged over the decade) and GDP per capita. All context data is from Eurostat-ARDECO (2018). In each model, I include election-specific fixed effects – meaning that individual areas are being compared only to other areas in their country for a given election. I further cluster the standard errors by election and all analyses are weighted by population in the unit. Table 2.2 shows the results, with Figure 2.5 showing the results graphically.

We see that social democratic parties, on average, have a lower vote share in knowledge-intensive regions, and the gap between cities and other areas are larger in these regions. When the UK (with its majoritarian system) is excluded, the patterns are similar but the standard errors on the city*KE interaction are larger. The Moderate Right, by contrast, does worse in knowledge-intensive urban regions (compared to rural areas and suburbs), an effect that is somewhat reduced in size when the UK is excluded, but remains statistically significant. Green and left-libertarian parties do substantially better in urban areas in knowledge-intensive regions compared to both other types of areas and urban areas in non-knowledge-intensive regions. These results are not just from large cities, green and left parties have the strongest base in mid-sized cities. Here, the exclusion of the UK magnifies these effects (particularly as, in this volume, the SNP, with a more mixed geographic base is considered a green left party). Radical left parties only exist in several countries, and here suburbs differ across knowledge and non-knowledge-intensive regions, with a higher urban gradient but no differences across knowledge-intensive areas. Finally, the knowledge index is positively associated with radical right voting, and this effect is driven by rural areas and suburbs, which are much more likely to support radical right parties in knowledge-intensive regions. The Appendix replicates these results, here looking at the patterns of interaction with the 1980s industrial structure, rather than the knowledge economy measure. We see somewhat congruent patterns, with social democratic parties performing better in industrial cities than other areas, with the Moderate Right performing worse in industrial cities. Green and left-libertarian parties do best in less industrial cities, with radical left parties performing less well in industrial suburbs and rural areas. Finally, the Radical Right performs better in all types of industrial areas, with a weaker gap between cities and suburbs and rural areas here (additional analysis shows that this is particularly driven by mid-sized cities).

In short, largely in line with Table 2.1, the competitive space around party mobilization varies across both urban types and economic regions, creating distinct patterns of competition. The heightening of structural pressures created new geographic pressures, combined with tipping point of the financial crisis, pushed Europe’s regions into different directions. In cities, social democratic parties must compete with green and left-libertarian or radical left parties for both middle- and working-class citizens, in the country side, with the Radical and Moderate Right. The results here suggest ongoing difficulties in creating a cross-regional left coalition.

2.3.3 Historic Mobilization Meets New Pressures

What do these dynamics mean for aggregate differences across countries? Understanding how the regional dynamics translate into different national strategic threats, requires tracing the relative threat different competitors face.

Historically, social democratic parties had different areas of strength across countries. To measure historic patterns then across all the European countries, I return to the LAU dataset, which includes election results from the 1980s (and from 1993 for France and 1994 for Germany). Here I average voting through all elections 1980s measuring disproportionality in social democratic parties’ share across towns and rural areas, suburbs and cities. Disproportionality is measured as the proportion of a party’s total vote that comes from a type of geographic area relative to the population share of that area. For instance, about 54 per cent of Sweden’s population lived in rural areas and towns in the 1980s, but about 58 per cent of SAP voters were town dwellers (35 per cent of Swedes were city dwellers, compared to 31 per cent of SAP voters). This situation creates some mild rural disproportionality in Sweden. By contrast, in Austria, 46 per cent of SPO voters lived in towns, compared to 49.5 per cent of the population, creating a more urban base. The difference between these numbers is the rural disproportionality measure. For Italy, I count the Communist Party as the main social democratic party in the 1980s. Figure 2.6 plots this disproportionality measure on the x-axis, against total vote share for the party through the decade on the y-axis. While the differences are not large, we nonetheless see consistently different patterns of mobilization across place.

Figure 2.6 Mobilization in the 1980s

To find out how these historic structures intersect with economic change, I conduct one final analysis. Here I combine the knowledge index and the urban categorization, splitting the top and bottom half of those on the knowledge index, and further splitting urban areas from suburbs and rural areas. The result is four groups, urban more-knowledge-intensive local areas (Urban-KE), urban less-knowledge-intensive areas (Urban Non-KE), and rural and suburban more-knowledge-intensive areas (Rural-KE), and rural and suburban less-knowledge-intensive areas (Rural-Non KE). I then interact this categorization with the share social democratic parties received in the local area in the 1980s, with urban KE regions as the baseline. Each model uses country-year fixed effects and country-year clustered standard errors, and all results are weighted by the LAU vote share. This approach examines whether the places that social democratic parties traditionally mobilized have been more or less vulnerable to particular types of competitors, giving us hints at the aggregate threat they face.

Table 2.3 shows the results, with Figure 2.7 demonstrating the results graphically. We see that not surprisingly, social democratic parties do better in all types of places that they historically mobilized, with no major differences across the urban regional types. They do marginally better in urban areas that they were historically weak in, suggesting that on aggregate they have picked up some urban support outside of their core areas. When the UK is excluded, we see a small, but significant positive effect on vote share in rural strongholds, suggesting that outside of the UK, they have been marginally more effective at holding their rural and suburban strongholds. However, the effect size is small and the standard errors large, suggesting heterogeneous patterns.

Figure 2.7 Historic mobilization and regional type

The Moderate Right shows the reverse pattern to social democratic parties. These parties perform best in the regions where social democratic parties were traditionally weak, especially rural areas, and are weak in historic social democratic urban knowledge strongholds. Green and left-libertarian parties are strongest in urban knowledge centres both in traditionally strong and weak social democratic areas. Only rural and suburban knowledge-intensive areas are more prone to radical left voting if they were social democratic strongholds, for other areas there is no strong association. Finally, while the previous analysis suggested rural knowledge-intensive regions were key to radical right parties’ strength, we see that this strength is attenuated when the Left had a stronger presence in the region.

These results suggest, in line with Section 2.2, that the aggregate threat to social democratic parties may vary based on historic mobilizing structures but not always in straightforward ways. Social democratic parties do, however, face different challenges in their historic areas based on their structure, with less relative challenge from the Moderate Right (although absolute moderate right voting remains the highest competitive group in all areas), strong challenge from the Green Left, and a more mixed pattern with the Radical Right.

3.3.1 Who Left?

First, we show which groups have left social democratic parties since the 2000s. Specifically, we examine how LR self-placement, education and union membership are associated with voting for a social democratic party, conditional on whether respondents voted social democratic in the previous election. We thus separate voters into three categories: stay (social democratic vote at this and the last election), attract (social democratic vote at this but not the last election) and leave (social democratic vote at the last but not this election). All graphs convey two sets of information: How relevant a group is for the pool of social democratic voters, taking into account the group size within the whole electorate (Figure 3.2(a)) and how stay/attract/leave are distributed within that group of voters (Figure 3.2(b)).

Figure 3.2 Vote switching from social democratic parties conditional on LR ideology

Source: national election studies (see Footnote footnote 3).

Figure 3.2 shows which voters decided to turn their back on social democratic parties conditional on their LR ideology. Figure 3.2(a) shows the proportion of all voters in the stay/attract/leave pool. We can see that centrist and left voters make up about the same share of the overall pool of social democratic supporters in the eight countries under investigation. Voters who identify as right-wing constitute a barely relevant group for social democratic parties. Overall, social democratic parties lost most voters among people who identify as centrist.

In Figure 3.2(b), we condition on LR ideology, so this panel presents the proportion for stay/attract/leave within each group. This panel shows that the more right-wing a person is the more likely they left social democratic parties. Since right-wing voters constitute a small support group for social democratic parties overall, the consequences of their departure were arguably small. However, the large share of centrist voters that left social democratic parties plays a crucial role in their decline. This provides initial empirical evidence against the economic narrative. If backlash against centrist policies was the main driving factor behind the electoral decline of social democratic parties, why would we see that it is mostly centrist voters who leave these parties?

Figure 3.3 shows patterns of support for social democratic parties based on education. We can see (in (a)) that social democratic parties have significantly lost voters at all levels of education, with people in the middle category making up the most important group for social democratic support. The dilemma of Social Democracy that Kitschelt already outlined in 1994 is strongly visible in this graph. Social democratic parties’ electoral decline cannot be reduced to one of the three groups. The restructuring of the political space has made them less attractive to those with lower as well as those with higher education.

Figure 3.3 Vote switching from social democratic parties conditional on education

Note: Below UpSec = below upper secondary; UpSec/Voc = upper secondary and vocational.

Source: national election studies (see Footnote footnote 3).

Interestingly, Figure 3.3(b) shows for the highly educated group that both attract and leave are higher, indicating high levels of competition over educated voters. At the same time, Social Democrats have proportionally seen somewhat stronger losses among the higher educated, so their support has been in slightly higher decline among this group. This indicates that narratives that focus on the decline of Social Democracy purely through a lens of eroding working-class support fail to identify a main driver of their declining vote shares. It also means that social democratic parties face dwindling support in a group that has grown in post-industrial societies and will continue to do so.

In Figure 3.4, we can see the different patterns of changing social democratic support for people who are union members versus those who are not. As Figure 3.4 shows, union members remain a crucial source of support for social democratic parties. We can also see that union members showed a smaller tendency to leave social democratic parties than other voters. While losses are pronounced among both groups, unions thus seem to continue to act as a stabilizing factor in the social democratic electorate.

Figure 3.4 Vote switching from social democratic parties conditional on union membership

Source: national election studies (see Footnote footnote 3).

3.3.2 Where Did They Go?

In order to understand the electoral decline of social democratic parties, a crucial factor is not only which voters left, but even more importantly which parties they chose to support instead. Knowing which competitors former social democratic supporters left the party for gives us important insights into these voters’ reasons for leaving the Social Democrats. Both narratives discussed in this chapter provide us with some expectations about which parties former social democratic voters should support. The economic narrative is based on the idea that supporters of social democratic parties turn their back on them because these parties have become too neoliberal. Based on this narrative, we should thus expect voters to choose parties that are economically more left-wing than social democratic parties. A classic example of this would be former supporters of the German SPD who, as a response to the Hartz IV labour market reforms, defect from the party in favour of the left-wing populist Die Linke. The narrative that social democratic parties have lost electoral support because their cultural positions are too progressive is strongly linked to the idea that voters, especially in the working class, have switched to the Radical Right. Empirically investigating where social democratic voters went thus provides us with important evidence about the empirical plausibility of these narratives.

Figure 3.5 presents the distribution of party family support for voters who left social democratic parties during the last two decades. The percentages can thus be interpreted as shares of all voters who left social democratic parties for another party in this decade. We exclude those voters switching to abstention.

Figure 3.5 Vote switching from social democratic to other parties

Source: national election studies (see Footnote footnote 3).

The findings in Figure 3.5 provide important evidence about the dynamics of social democratic electoral decline. First, only a small share of voters – just above 10 per cent – went to parties that belong to the radical right party family. This finding stands in stark contrast to much public debate about the decline of Social Democracy that largely focuses on this group. Second, we find that many voters who left social democratic parties since 2000 switched towards moderate right parties, overall almost 30 per cent of defectors. Competition with the Moderate Right seems to be a decisive factor for the decreasing vote shares of social democratic parties. Third, social democratic parties lost most of their voters by far to progressive green and left-libertarian parties. A closer look at the data reveals that this is even more pronounced in the last ten years that include the watershed elections in France and the Netherlands in 2017. In sum, what stands out in this figure is that nearly 60 per cent of former social democratic voters either switched to a moderate right or a green and left-libertarian party.

Figure 3.6 shows the same vote switching analysis, but now conditional on voters’ level of education. Among voters with low levels of education, the highest share left for a moderate right party. The higher the education of a voter, the more likely it is that they left for a green and left-libertarian party. For voters with tertiary education, we can see that about as many people decided to switch to a green and left-libertarian party as to all other parties combined. Even among people with low levels of education, we only find very small shares of vote switching to the Radical Right.

Figure 3.6 Vote switching from social democratic to other parties, conditional on education

Note: Below UpSec = below upper secondary; UpSec/Voc = upper secondary and vocational.

Source: national election studies (see Footnote footnote 3).

These findings speak against the empirical expectations that follow from the economic narrative. The fact that in the aftermath of the Third Way period and during a time that saw a large-scale economic crisis, social democratic parties lost many more of their voters to the Moderate Right than to the Radical Left is strongly at odds with an idea of a direct backlash against Third Way policies. About a fifth of former social democratic voters opt for the Radical Left, who oppose social democratic parties mainly on economic issues.

The predominant cultural narrative that claims that social democratic parties are too progressive on second dimension issues finds even less support. First, only a small share of people who previously supported social democratic parties decided to vote for a radical right party. Second, during a period that saw increased attention to second dimension issues and particularly to immigration, social democratic parties lost most of their voters to parties that are more and not less progressive than they are. Losses come disproportionally from educated voters. None of these findings support the dominant narratives of the decline of social democratic parties.

In multi-party competition, social democratic parties of course do not unidirectionally lose voters to another party family but can also win voters back. Just focusing on losses might thus provide an incomplete picture of social democratic decline. In Figure 3.7, we thus show both social democratic gains and losses to other party families. Overall, we can see that social democratic parties have a negative balance with all other party families – there is no party family that social democratic parties gain more voters from than they lose to. The gains and losses patterns support our previous findings. The net losses to green and left-libertarian parties strongly outweigh all other net losses. For moderate right parties, we find a more balanced picture of gains and losses, but in absolute terms the net difference shows a strongly negative effect for social democratic parties. These findings also contradict the idea that party competition largely happens within and not between ideological blocks.

Figure 3.7 Vote switching between social democratic and other parties

Source: national election studies (see Footnote footnote 3).

Overall, our findings demonstrate that it is necessary to move away from a narrow focus on working-class and low-education voters and to incorporate the educated middle class into approaches studying the electoral fate of social democratic parties. We argue that a focus on this group at the centre of tripolar competition between moderate left, moderate right and green and left-libertarian parties is necessary to understand the electoral trajectory of social democratic parties and leads to an account of their decline that strongly differs from the economic and cultural narrative presented in this chapter.

These findings are based on data from countries with proportional representation electoral systems. One might indeed wonder how these dynamics play out in single-member district systems that provide stronger incentives against defecting to smaller green and left-libertarian parties. Additional analyses for the United Kingdom, however, show a pattern that is remarkably similar to the one presented in this chapter. Labour has only lost very few voters to United Kingdom Independence Party (UKIP). Losses to the Liberal Democrats play an important role and, in the UK, outweigh losses to the Greens. Losses to abstention are especially high in the UK. This indicates that in SMD systems, voters that are dissatisfied with the mainstream left might be more likely to abstain than to switch to a more progressive party. It is noteworthy that in SMD systems, smaller changes in votes might be more harmful to the mainstream left than in PR systems. Even small shifts in votes can translate into larger shifts in seats and vote losses to the Liberal Democrats or the Greens may even cause seat losses to the Conservative party.

4.3.1 Data

The questions guiding this chapter pose quite high demands in terms of the data required to address them. To be able to assign individuals to their own and their parental social class, we need detailed information on their employment status and occupation, as well as their parents’ occupation when respondents were young. Therefore, we rely on data from ESS, merged with ESS-DEVO data (Ganzeboom Reference Ganzeboom2013) that provides a recoding of the answers to the open-ended question asking about parental occupation into four-digit ISCO codes. This is only available for ESS rounds 1–5, hence restricting our analyses to the period 2002–10. Our analytic sample consists of respondents aged 18 and over (eligible to vote) with information on their current occupation and parental occupation when respondents were young. We restrict the analysis to respondents from West European countries (nineteen in total), because occupational transformations and the structure of party competition differs for post-communist democracies. These countries are grouped into Northwestern Europe, Southern Europe, and Anglo-Saxon countries. In a second step of the analyses, when we zoom in on whether current patterns of class voting may be generating new ‘realigned legacies’ of left-wing support among the offspring of SCP, we restrict the analyses to the Northwestern European sample. Scandinavian and Continental European countries display stronger patterns of electoral realignment and, hence, represent a better sample to address the future of intergenerational transmission of left-wing support. In spite of this focus, additional analyses on all countries indicate that the trends here identified are largely consistent across Western Europe.

4.3.2 Variables and Operationalization

Our dependent variable captures whether the respondent voted for one of the following party families in the last national election: social democratic, green and left-libertarian, radical left, moderate right, and radical right, or other party families (including regionalist, agrarian and religious parties). For the recoding of party choice into these party families, we rely on the classification provided by the Chapel Hill Expert Survey. To characterize the full choice set, we also include a category for abstention, for respondents who indicate not having voted in the last national election. This means that our dependent variable for party choice is categorical, with seven non-ordered response categories. To facilitate interpretation, we restrict the presentation of the results to those outcome categories that are meaningful in the respective analyses (i.e. we frequently exclude abstention and ‘other parties’), but all categories are included in the model estimations.

The main independent variables are respondents’ class of origin and class of destination. Four-digit ISCO codes of occupations are used to classify occupations in different social classes. To minimize the number of observations without an assigned social class, we code class of the respondent based on their occupation and, when this is missing or the respondent is not in employment (as their main activity) we rely on partner’s class location, when the respondent is cohabiting with them. To code parental class, we rely on father’s occupation when the respondent was fourteen years old. If this information is missing (because the father was absent, not working, or because there is no information on his occupation), we rely on mother’s occupation. For the operationalization of social class, we use a simplified version of the Oesch eight-class scheme (Oesch Reference Oesch2006) – presented in Table 4.1, where the shaded cells indicate the six (aggregated) class categories used in the analyses.

We combine some of the classes into fewer categories because our main interest is in the class constituencies of the Left: the production and service workers – aggregated into ‘workers’ – and the SCPs. The old middle class (OMC) is operationalized by aggregating managers, self-employed professionals and large business owners and serves as a point of comparison in the analysis, as it represents a typical moderate right-wing constituency. Technical professionals (TCP) are kept separately as some research indicates that they are more left-leaning than the OMC (but less so than the SCP) (Kitschelt and Rehm Reference Kitschelt and Rehm2014). Office clerks (CLERKS) and small business owners (SBO) are kept as separate classes in the analyses without a specific partisan profile. The aggregation of certain classes simplifies the high number of potential intergenerational transitions between social classes, when we focus on different combinations of class of origin and destination, and their support for different party families. A respondent is socially mobile when their class of origin is different than their class of destination. To investigate our expectations regarding patterns of socialization and social mobility, specific combinations of class of origin and class of destination are used as categories of independent variables in the second set of analyses.

The analyses distinguish four generations of respondents: the silent generation (born 1928–45), baby boomers (1946–64), gen X (1965–80), and millennials (1981–96). While the silent generation and baby boomers have been socialized during periods of industrial alignments, the two younger generations have been socialized during post-industrialization periods. The median birth year in our sample is 1959 (mean: 1957); the 99th percentile is 1988. This means that most respondents are baby boomers, and that for the millennials, only relatively older respondents of that generation are included.

Control variables are included for gender, age, and marital status. To investigate whether part of the intergenerational transmission of left-wing support could act through a higher likelihood of being unionized, additional analyses (not shown) have been estimated that control for trade union membership of the respondents. The results remain unchanged with this additional control variable; so the figures here presented are based on the more parsimonious specification.

4.3.3 Analyses

Modelling the probability of voting for different parties or abstaining requires fitting a discrete choice model for non-ordered responses. Because not all respondents face the same choice set across the different countries under consideration (not all party families are represented in all countries), we implement a multinomial conditional logistic regression model that allows for varying individual alternative choice sets (Thurner Reference Thurner2000; Weber Reference Weber2011). To facilitate the interpretation of the results, and to avoid having to interpret log-odds relative to a reference category, we present the results from these analyses either as average predicted probabilities or as average marginal effects (AMEs) on probabilities. Standard errors are adjusted at the party-system (country) level to account for the different choice sets.

We estimate two sets of analyses. First, we analyse differences in party support by parental class of origin (and class of respondent, for comparison), using the OMC as a reference category. We further study differences in these patterns by estimating the model separately for the three geographical regions in Western Europe and four generations of respondents. These models are relatively restrictive because they include both class of origin and class of respondents in the same model, to estimate its impact net of each other. Second, we study how specific patterns of intergenerational social mobility (combinations of class of origin and class of destination) explain differences in party family support. We thereby focus specifically on transitions into the SCP and transitions out of the SCP (the latter for the two youngest generations of respondents only, because socialization in the SCP is less frequent among older generations). As mentioned earlier, this second set of analyses is restricted to North Western Europe, as the class–party realignment that this chapter builds on has started earlier there than in other regions within Western Europe.

4.4.1 Party Family Support by Class of Origin and Class of Destination

Before presenting differences in party family support by class of origin, we present descriptive analyses of the class composition of the electorate, as the regression models we present next do not account for the size of the relative classes. Figure 4.1 displays the class composition of the electorate of different party families, for class of origin, and class of destination, respectively. The results indicate substantive differences in the class composition of the electorate of different party families (as already documented extensively by existing research), but, more importantly, that this heterogeneity is also manifested by class of origin (parental class). Differences between the class composition by class of origin and class of destination display the trends of occupational upgrading and tertiarization of the post-industrial employment structure, as well as the realignment of social classes and parties.

Figure 4.1 Composition of electorates of party families by respondents and parental class

Source: Authors’ calculations using ESS 2002–2010.

First, focusing on the composition of the electorates by class of destination (offspring class), it shows that social democratic parties rely to a large extent (45%) on support from production and service workers. The middle-class electoral base of the Social Democrats is also substantial with one-third (35%, the total of SCP, TCP, and OMC), but the green and left-libertarian parties rely much more on middle-class support (total of 57%, vs. 24% from workers), especially from SCP (27%). This relatively large middle-class social democratic support is in line with the changes in the electorate of Social Democracy documented in earlier work (Gingrich and Häusermann Reference Gingrich and Häusermann2015) and discussed in the introduction of this volume. Compared to the other party families, the Social Democrats rely more on working-class support than moderate right parties, or radical left parties. However, we see an even greater weight of workers’ support among the radical right and abstention categories (close to, or over 50%).

Second, we ask to what extent the support base of the Social Democrats stems from voters with working-class backgrounds in the parental class. When we contrast the foregoing figures with the composition of the electorate by class of origin rather than class of destination, 52 per cent of social democratic party support stems from offspring from workers (mainly production workers). Compared to other categories, the proportion of the social democratic electorate coming from a working-class background is similar to that of abstention, and larger than the one of the Radical Right. Radical left and green and left-libertarian parties also rely on substantial support from offspring from workers (respectively, 46% and 33%), but less so than the Social Democrats. Especially the green and left-libertarian parties have a relatively larger support base from middle-class offspring than the Social Democrats (42% compared to 21%).

Taken together, one of the most important conclusions is that social democratic parties draw over-proportionally on support from the offspring of the working class. By comparison, in their electoral base by class of destination, the support base among workers is lower, and among the middle class is higher. By class of destination, workers form a larger part of the support base of the Radical Right (or abstention). These differences can be attributed to compositional effects: due to the transformations of the occupational structure, class sizes differ in the parental and offspring class, which account for the differences in the composition of the electorate by class of origin compared to class of destination. This is most clearly illustrated by the large fraction of offspring of workers across all parties, compared to that by class of destination.

Put simply: many contemporary middle-class voters have a working-class background, which explains differences in the class base of parties between the two analyses. Class–party realignments and intergenerational social mobility thus go together in explaining change in the class base of parties in general, but the Social Democrats in particular. For the Social Democrats, middle-class voters currently comprise one-third of their electorate. The fraction of the electorate with a middle-class background is only likely to increase in the future, as the offspring of these middle-class voters are more likely to hold middle-class than working-class positions. What does that mean for the support for the Social Democrats in the long run? Are they able to retain the support of middle-class voters without the ‘legacy’ of the working class? We will focus on this question in the multivariate analyses.

Next, we present results of our regression analyses predicting party family support by class of destination and class of origin, respectively net of each other, with added control variables. In Figure 4.2, AMEs of support (reference: OMC) for the five party families of interest are displayed for our categories of social class of origin and destination. First, focusing on class of destination (respondent’s class), the results indicate some patterns of class voting documented by earlier research, showing that social class is associated with support for different party families. However, not all classes differ in support for the five party families of interest. Regarding the social democratic party family, workers, SCP, and TCP show similarly higher support levels than the OMC. Concerning green and left-libertarian parties, the new middle classes – particularly the SCP – show higher levels of support, while the working classes show much lower support than the OMC. This is the exact opposite pattern of what we observe for the Radical Right: relatively higher support from workers and lower support from the SCP. Support for the Moderate Right also shows a clear class pattern: the highest support comes from the OMC, as all other class categories show lower support levels.

Figure 4.2 AMEs of respondents and parental class on support for party families (reference category: OMC)

Source: Authors’ calculations using ESS 2002–2010.

Second, when comparing support levels by class of destination with those by class of origin, some interesting differences and similarities in social democratic support become apparent, while support for other party families does not show many noticeable differences. Levels of social democratic support are quite similar across class of origin and destination when it comes to working-class voters and voters with working-class roots. However, social democratic support from offspring from SCP is not different from offspring from OMC, while the difference with working-class offspring is quite large. On the other hand, offspring of SCP and current SCP show similarly high support for green and left-libertarian parties. Offspring of SCP are also relatively more likely to support the Radical Left, a party family with a typically strong economically left-wing agenda – even to a greater extent than the offspring of workers. Given the trend of occupational upgrading, these findings may point at the fact that the relatively high social democratic support among contemporary SCP is mainly a legacy of their working-class background – and will decrease in next generations. Patterns of socialization in the SCP parental class of origin seem, however, to be related to left-wing support, but rather for green and left-libertarian and radical left alternatives. We take up the question of the SCP legacy further in Section 4.4.2, after we focus on differences in the effect of class of origin across regions and between generations of respondents.

In Figure 4.3, we present the AMEs of class of origin separately for the three regions of Western Europe. Northwestern Europe is the largest group of countries in our sample and therefore displays lower standard errors than the other two regions. The patterns across the three regions are relatively similar and mirror largely those from Figure 4.2, but there are some relevant differences. First, as expected, in Northwestern Europe we observe a relatively strong legacy of socialization in a working-class environment for left-wing support, particularly the Social Democrats. In the other two regions, we observe this as well, but the contrasts with an OMC background are smaller. Second, like concluded from Figure 4.2, having SCP roots does not have a similar effect as having working-class roots does (they do not show different levels of social democratic support than those with OMC roots, across all regions). However, in Northwestern Europe, socialization in the SCP does predict higher green and left-libertarian support, compared to socialization in the OMC. The fact that this difference is not present in the other two regions demonstrates the earlier transformation to post-industrial class–party realignments in Northwestern Europe, which is the reason why subsequent analyses focus on this region only.

Figure 4.3 AMEs of parental class (reference category: OMC) on support for party families, across geographical regions

Source: Authors’ calculations using ESS 2002–2010.

The results presented in Figures 4.2 and 4.3 suggest that parental class of origin generates a legacy in party leaning that persists over time and continues to influence offspring’s electoral behaviour, even net of respondents’ own class location. Moreover, some of the differences in the impact of parental class of origin and respondents’ class seem to indicate that current class–party alignments differ from those prevalent in the past. This is to be expected given the fast pace of post-industrial realignment. As different studies have manifested, the core electorates of left-wing parties have substantially changed over the past few decades (Gingrich and Häusermann Reference Gingrich and Häusermann2015; Benedetto et al. Reference Benedetto, Hix and Mastrorocco2020), and what were once their natural electoral preserve – workers – are now increasingly contested by radical right parties (Oesch and Rennwald Reference Oesch and Rennwald2018). This changing alignment between social class and parties could also affect the socialization into a specific partisan leaning. To illustrate this with an example, the kind of working-class left-wing legacy that was built in the 1960s – and the impact of socialization into this milieu – might take on very different nature in the early 2000s, as workers have increasingly realigned with the Radical Right. To explore this possibility, we disaggregate the effect of parental class of origin by different generations: from silent generation respondents to millennials.

Figure 4.4 presents AMEs of parental class of origin on support for different party families by generations. All coefficients indicate deviations from the reference category, offspring from the OMC. This figure presents some varying patterns of how the impact of class of origin differs across generations, which are exemplary for trends of realignment in which the class base of particularly parties on the left has changed. Focusing on parties in the left field and their core electorates, we observe that among voters from the silent generation there is a clear social democratic working-class legacy. Voters with working-class origins are almost 10 per cent more likely to vote for this party family than OMC offspring, while a similar impact is not visible for SCP offspring. On the contrary, the latter appear less likely to vote social democratic than voters with OMC roots. At the same time, a weak left-wing legacy in terms of support for radical left or green and left-libertarian parties is apparent for the offspring of SCP already in the silent generation.

Figure 4.4 AMEs of parental class (reference category: OMC) on support for party families, across generations, for Northwestern Europe

Source: Authors’ calculations using ESS 2002–2010.

The picture, however, is quite different if we focus on gen X or millennial respondents, for whom patterns of class–party realignment are reflected in the impact of class of origin of new strongholds of the Left. In these younger generations, we find evidence for a left-wing legacy among the offspring of SCP, for all three party families in the left field. This points at offspring of SCPs having been socialized into generalized support for left-wing parties, from the Radical Left and the Green Left, to the Social Democrats. For millennials, some of these differences are not statistically significant, but this could be partly attributed to the lower frequency of observations in this generation. The fact that SCP offspring is not only more supportive of the New Left but also of the Radical Left compared to OMC offspring, displays a trend of increasing support for the Left in general, with a corresponding decline in support for the Moderate Right. This marks an increasing distinction of socialization in the OMC compared to the new middle class, the SCP. As some of the left-wing legacy becomes stronger for the offspring of SCP as we move to younger generations, we observe a parallel decline in these allegiances among workers’ offspring. In fact, among millennials, offspring of workers are as likely to vote social democratic as offspring of SCP. Interestingly, among the children of workers, we observe an increasing contrast in the lower likelihood to support green and left-libertarian parties in comparison to the offspring of the OMC.

While the left-wing impact of having a working-class origin appears to have diminished for younger generations, we do not observe a parallel increase in favour of a radical-right legacy. Moreover, the lower probability of voting for the Moderate Right among workers’ offspring (in comparison to the OMC) appears quite stable over generations. Hence, it appears that working-class origins have somewhat ‘de-aligned’ from the Left, but there has not been a parallel ‘realignment’ in support for the Right. Among SCP offspring, we do, however, find evidence of left-wing realignment, because, as explained earlier, support for the left field is higher among younger generations with this background. Among the offspring of TCP, we find little evidence of changing alignments. Their electoral preferences resemble those of the OMC across all generations.

Focusing on long-term legacies of left-wing support, these analyses clearly highlight that the kind of social democratic legacies apparent for the offspring of workers in older generations are not replicated for any other classes or generations. While SCP could be building a new left-wing legacy among its offspring (rather for green and left-libertarian and radical left alternatives), this is in no way comparable to the strong parental effects visible for silent generation respondents with working-class roots. This speaks to a certain unique imprint of having a working-class origin among older generations that is not replicated in current realigned patterns of class-party linkages.

4.4.2 Party Family Support by Patterns of Intergenerational Social Mobility

Previous analyses have shown relevant differences in party support across parental class of origin and respondents’ own class of destination but did not consider these two in conjunction with each other. Therefore, we now model specific combinations of class of origin and destination. To more precisely investigate to what extent the left-wing support of current SCP is contingent on their working-class background and explore whether a new model of SCP support is sustainable in the long run, we estimate regression models by different patterns of intergenerational social mobility. For the purpose of our analysis, we specifically focus on transitions into and out of the SCP class. In Figure 4.5, we present results for the former. The graphs display predicted levels of party family support.

Figure 4.5 Average predicted levels of support for party families by patterns of intergenerational mobility for Northwestern Europe (focus: into SCP)

Source: Authors’ calculations using ESS 2002–2010.

When comparing the social democratic support from SCP with and without a working-class background, it shows that SCP with working-class origins are not only more likely to support the Social Democrats, they also show a lower likelihood of voting for the Moderate Right. Hence, a working-class background is associated to a more distinct left-wing allegiance. Only immobile workers are less likely to support the Moderate Right compared to SCP with working-class roots – showing a certain pattern of so-called acculturation to the SCP class of destination (de Graaf et al. Reference De Graaf, Nieuwbeerta and Heath1995). Interestingly, SCP with a working-class background and intergenerationally immobile workers show similar levels of social democratic party support. In what concerns other types of left-wing allegiances, a working-class background aligns SCP more with radical left parties, and less so with green and left-libertarian parties. Hence, this working-class legacy appears more strongly related to party families pertaining to the left bloc on the economic dimension (rather than on the societal dimension).

When considering the different categories from the immobile OMC to the immobile workers, each category shows a decreasing level of support for the Moderate Right, and increasing support for the Social Democrats. While among immobile workers support for the Social Democrats is higher than support for the Moderate Right, this pattern is reversed for SCP with a non-working-class origin. Processes of adaptation to the class of destination for SCP mean that the Moderate Right is a direct competitor from the Social Democrats in the support from this class. In the long run, these results may point at decreasing social democratic support in the future, when the legacy of a parental working class is reduced over time by patterns of occupational upgrading. Other left-wing parties and the Moderate Right, on the other hand, hold relatively higher levels of support from SCP without working-class roots, a class that is likely to become larger over time.

In Figure 4.6, we explore the potential legacy of the SCP – the class displaying among the highest levels of social democratic and other left-wing party support – towards the next generation, by asking how being socialized in this class predicts levels of party support. To approximate what future patterns of intergenerational transmission of preferences might entail, we reduce the sample to the two youngest generations in the sample (Gen X and Millennials). These results need to be interpreted as a tentative answer to our question, since relatively few individuals have been socialized in the SCP (class of origin), compared to the current size of this class (as class of destination), and therefore the size of the sample is reduced (which also explains the larger confidence intervals). The findings indicate that those socialized in the SCP that moved to a different middle-class position (OMC) – a common transition of intergenerational social mobility in the post-industrial era – are not very likely to support the Social Democrats, displaying similar levels of social democratic support as the immobile OMC, and lower than the immobile SCP. These findings, together with the high support for the Moderate Right, mainly point at a process of acculturation to the class of destination.

Figure 4.6 Average predicted levels of support for party families by patterns of intergenerational mobility among Gen X or Millennial respondents for Northwestern Europe (focus: out of SCP)

Source: Authors’ calculations using ESS 2002–2010.

If we focus on other parties within the left-wing bloc, we see some indications of a slightly stronger left-wing legacy among SCPs offspring, but not to the extent of the left-wing alignment we observe among intergenerationally immobile SCPs. The immobile SCPs display highest support levels of all classes for the green and left-libertarian parties and the Radical Left. The probability to vote for these parties is lower for SCP-origin voters in the OMC but remains slightly larger than those displayed by the intergenerationally immobile OMC. We observe similar patterns among workers with SCP roots: they show higher green and left-libertarian and radical left support than immobile workers. These results are an indication of modest SCP-origin socialization effects, when offspring holds a different class. If being socialized in the SCP has any left-leaning legacy, it seems like its allegiance lies more with the Radical Left and green and left-libertarian parties, than with the Social Democrats.

Taken together, these findings indicate that, as the offspring of workers are likely to become less numerous over time, the same goes for the legacy of this class in retaining social democratic (and other left-wing party) support, predicting likely lower levels of social democratic support among future generations.

5.4.1 Case Selection

Our case selection is pragmatic; the countries we study are the only ones in Western Europe conducting long-running and large-panel studies. However, we believe that the three countries provide interesting variation to study patterns of social-democratic decline. The German SPD, the British Labour Party, and the Swiss SP provide an excellent snapshot of quite different social democratic party organizations. While all three of them originate from the classical Rokkanian cleavage mobilizing around capital (owner) versus workers, the SP never had the electoral size of its sister parties – which is mostly due to the party and electoral system in Switzerland. However, its importance for policymaking and governing is comparable. All three of them are struggling (in different degrees) to maintain their vote shares and face severe electoral challenges particularly on the second dimension, most notably by green and left-libertarian parties. In the UK and Switzerland, socially liberal and more center-oriented parties add an additional element of electoral competition from the LibDems and the Green-Liberal party, respectively. Also in all three countries, there has emerged a notable competitor on the Radical Right, which allows to empirically examine the relevance of the often-claimed alleged voter transitions from Left to (Radical) Right. Finally, different institutional setups and distinct programmatic profiles in the three countries under consideration allow us to some extent to assess which voter transitions from social democratic parties to competitors are conditional on the electoral system and the specific ideological orientation of the party.

5.4.2 The Panel Data, Our Coding Decisions

The analyses in this chapter will focus on description only. What we first want to understand are the individual voting trajectories across time as such. In a second step, we then correlate these long-term voting trajectories with key individual-level characteristics believed to be relevant for social democrats’ voting decisions in the last forty years. Our study is based on three high-quality individual-level panels that maximize representation of the general population at the national level. Table 5.1 gives an overview of the panels we included, the time span we analyze, and the number of social democrats included in our analysis.

The number of respondents we analyze varies across cases, but it is sufficiently large across all three countries as can be seen in the last column of Table 5.1. In the UK, we essentially rely on two panels – the British Household Panel Survey (BHPS) and the Understanding Society: the UK Household Longitudinal Survey (UKHLS) (University of Essex 2023). The UKHLS continues the data collection efforts of the BHPS in most regards, and we made sure to only include respondents in the UKHLS which were already part of the BHPS’ original data collection efforts. In Germany, we rely on the German Socio-Economic Panel (SOEP) and in Switzerland on the Swiss Household Panel (SHP) (Voorpostel et al. Reference Voorpostel, Tillmann, Lebert, Kuhn, Lipps, Ryser, Antal, Monsch, Dasoki, Klaas and Wernli2020).

To conduct our analyses, we need a consistent definition of social democratic “core” voters across time and space. We decided to use a conservative strategy and define a social democratic voter as one who repeatedly votes for the Social Democrats. In all three countries, thus, we use the survey years falling into the first election cycle and define social democratic voters as those who report to vote for the Social Democrats twice in this first election cycle.Footnote ³ We then use this sample of respondents and follow their trajectory across all election years contained in each panel. We worked with different definitions not reported in this chapter and can confirm that the major patterns presented later do not hinge on our specific coding decision.

Beyond the broader trends across time, we also seek to understand the lifetime cycle of all voters. To do so, we report the first and their last voting behavior recorded for all of our panelists. We are fully aware that this is a tremendous simplification of individual voting habits as it ignores any within changes throughout a respondents’ life. Yet, it pictures the beginning and end of voting for parties in a very effective way.Footnote ⁴

While the three panels allow us to understand voting trajectories across a long-time span for a rich number of respondents, it also comes with important limitations. First, not all panel studies report voting across the entire study period. The GSOEP only does so in 2018 and 2014. Instead, we have to rely on a question that is more related to party identification than voting; respondents are asked to name “the party they support” – which is best understood as a mixture of survey questions on voting and party identification. However and importantly, in the case of the social democrats in Germany, they are heavily correlated. In 2014 and 2018, 85% of respondents with a SPD party identification report to have voted for the SPD. Thus, clearly the measure comes with measurement error but given the strong correlation with voting, we are reasonably certain that the cross-time patterns are sufficiently approximated. Also, the patterns we uncover for the SPD using party identification align very much to the patterns we find when looking into reported voting in 2014 and 2018 only. In contrast to the German panel data, BHPS and SHP report voting and, thus, we rely on standard voting questions for both countries.Footnote ⁵

Second, while all three panels are sufficiently large for our analyses, eventually all panels end because respondents either drop out or die. Since we want to follow the voting patterns of “original social democrats,” this means that panel attrition makes the number of respondents included across time shrink. To visually address this, we recalculated the percentages for each year such that it sums to 100%. Yet, almost all patterns we outline later seem to be more general trends. Thus, we can be fairly certain that they are not subject to biases due to panel attrition.

5.5.1 Transitions through Time

The first part of our analyses used a “big brush approach” in which we pooled respondents across time to get a very general understanding of the pattern of voter transitions. The second part takes the temporal dimension more seriously and looks into transitions of all social democrats in our data across each election. For the next set of analyses, we focus on the group of original social democratic voters in the first available wave of each panel data set. We define original core voters as those who have a strong (at least eight out of ten) social democratic party identification in Germany and the UK and those who have voted three times in a row for the Social Democrats in Switzerland (due to missing party ID information). We then plot the voter flows between elections over time. The bars always capture the total number of remaining panel respondents who originally supported the Social Democrats along with their updated party identification or vote intention.

We start the second part of the chapter with the Swiss case (Figure 5.2) because of the particularly large number of effective parties and, hence, potential competitors for the Social Democrats. The Swiss panel started in 1999 with a total of 1,488 respondents who we define as core supporters of the social democratic party. Following this cohort for four years until the next general election, we can see that about a quarter of them abandon the party. The largest share moved to the Green party, which in Switzerland has an almost identical ideological profile but differs somewhat in terms of issue saliency, most notably on environmental issues. A similarly large group of former social democratic voters indicates that they support “other” parties, which in most cases means that they gave their vote to a mixed group of politicians from different parties (“vote for persons, not a party”) rather than submitting the social democratic list to the ballot. Finally, more marginal segments of voters defect in all other directions including a small but nonnegligible group of voters who directly moves to the other end of the ideological spectrum and votes for the radical right Swiss People Party who continued their rapid growth at that time.

Figure 5.2 Transition away from SP, across elections

Four years later, in 2007, the pattern largely resembles one of the previous election with the Greens capturing a particularly relevant share of former social democratic votes. The year 2011 brought the relevant entry of a new competitor, the Green-Liberals, a more centrist environmental party classified into the Moderate Right camp. This new electoral option has certainly attracted parts of the Social Democratic electorate, but its entry hurt the Green party at least as much as the Social Democrats at that time. The year 2015 represents the consolidation of this pattern with the votes of the social democratic defectors almost evenly split across all the possible competitors. All in all, most former social democrats remained within the left bloc and only about 24% eventually moved on to the right bloc.

Much in contrast to the Swiss SP, both the British Labour Party and the German SPD went through a period of radical programmatic renewal – the Third Way. Both party leaders at the time (late 1990s), Tony Blair and Gerhard Schröder, “modernized” their party mainly on the economic dimension by proposing and adopting liberal market policies. To be fair, at least in the case of Schröder, this period was also marked by a period in government with the Green party and, thus, similar shifts on societal dimension issues – particularly the phase out from nuclear energy. So, these programmatic shifts should potentially also show us a quite different pattern of voting transitions than in the Swiss case. Specifically so, because this period is at the core of the public detour effect narrative.

Yet, what we find first for the UK, shown in Figure 5.3, aligns in most regards with the patterns in the Swiss case. Again, voters get demobilized – but mostly in the first two election cycles (1997 and 2001) we analyze. Interestingly, this is the period during which Tony Blair and his Third Way dominated the party. Thus, it appears that at least to some degree, these policies might have driven voters away from Labour.

Figure 5.3 Transition away from Labour, across elections

But what is then much alike to the Swiss case is the defection to the “progressive option” – the LibDems, which are the main actor in the Green/Left-Libertarian party family in the UK. Overall, even less voters moved away from the left bloc and joined the Right: All in all about 20% of former Labour voters moved to the Right with only a small fraction eventually voting for UKIP (3–5%).

In our last case, shown in Figure 5.4, we focus on the German case. In the first part of the analyses, we finished with a rather pessimistic view for the German SPD. Yet, the transitions across elections reveal a similar picture. From the beginning of our analysis, the SPD struggles to mobilize its core, and from the beginning their core defects to the Green party. In 1987, the Greens just entered the German parliament one election cycle ago in 1983. They are a young party, often still perceived as a protest party which is incapable to “deliver” policies. But social democratic voters appear to be attracted by this option much from its beginnings but also in considerably smaller numbers than in Switzerland and the UK. This is very interesting as at the time in many ways environmental concerns were perceived as being incompatible with policies for the working class: How to offer environmental policies and at the same time ensure that factories and companies keep running? Yet, in contrast to the former two countries, the SPD’s case seems exemplary for the abstention argument: Supporters leave the party but then also never return to any political party identification. They seem to abstain from politics altogether and in large numbers.

Figure 5.4 Transition away from the SPD, across elections

Apart from dealignment and the Green defection, the SPD seems to be spared from competition for its core voters by other parties. Former social democrats do not defect to the Right in large numbers, neither to the Linke. This means that the SPD is incapable in mobilizing its core as well as keeping their core from defecting to the Greens. Yet, there is good news: This means that much like the SP, a more progressive platform on the societal dimension could keep voters from defecting to the Greens while competition on the first dimension – such as welfare chauvinism – might not be relevant to keep voters from defecting to the right and left extremes. But we need to keep in mind that the first analyses showed that the SPD barely has any inflows from other parties. If this pattern persists, it may become difficult to save the SPD from descending into political insignificance.

This becomes drastically clear in Figure 5.5. Here we focus just on the last two German elections and rely on reported voting by respondents instead of party identification. We do this in order to show two things.

Figure 5.5 The threat of AfD, a new party entry from the right

First, that the patterns here align well with the findings discussed earlier based on party identification instead of voting. Second, the radical right party AfD entered the German political arena in 2013 as a mainly Eurosceptic party – but signs of radical right policies became relevant shortly after the 2013 elections. This means that the AfD started to offer the policies that allegedly drove former SPD voters to it in between the two elections. Again, we do not find such a flow to the AfD. The SPD loses more of its 2013 voters to the Christian Democrats, the Greens, the Linke, and abstention. For robustness, we also looked into voter inflows for the AfD to further substantiate that former social democrats did not flock toward the Radical Right in meaningful numbers in Germany. In line with our previous findings, by far the largest inflows are from voters who previously reported not having strong links to any party (figure not shown here for space constraints).Footnote ⁶

5.5.2 Who Leaves?

In a final step of the analysis, we wish to better understand the abovementioned party transitions by looking into the sociodemographic underpinnings of distinct switching patterns. More specifically, we ask what kind of voters characteristics correlate with the choice of distinct electoral alternatives to the Social Democrats.

The existing work on the social democratic decline suggests various potentially important individual-level factors that may help us understand defection better. We first look at a standard set of socioeconomic and sociodemographic variables which is typically used to explain vote choice: gender, age, education, and income. In addition, we also examine the role of unemployment (e.g., Kurer Reference Kurer2020; Wiertz and Rodon Reference Wiertz and Rodon2021) and union membership (e.g., Karreth et al. Reference Karreth, Polk and Allen2013; Abou-Chadi and Wagner Reference Abou-Chadi and Wagner2019) in supporting social democratic parties.

Given the above-discussed patterns over time and the evident difficulty of social democratic parties to attract new voters, a second important aspect of vote switching may be related to birth cohort. We differentiate between five different birth cohorts ranging from respondents born during WWI to the so-called Generation Y/Z born after.Footnote ⁷

Finally, perhaps the most frequently investigated factor explaining social democratic support (and the increasing lack thereof) is occupation and class. The well-documented decline of traditional class voting (Kitschelt Reference Kitschelt1994; Oesch and Rennwald Reference Oesch and Rennwald2010; Karreth et al. Reference Karreth, Polk and Allen2013; Rennwald Reference Rennwald2014; Rennwald and Evans Reference Rennwald and Evans2014; Gingrich and Häusermann Reference Gingrich and Häusermann2015) implies that we would expect disproportionate defection among the traditional working-class base. These outflows should be partly compensated by stronger or more enduring support from the new middle class. At the same time, the presence of a strong green or left-libertarian party that also attracts this culturally liberal segment of society is another source of competition (e.g., Rennwald and Evans Reference Rennwald and Evans2014).

In order to examine switching patterns by individual characteristics, we turn to regression models. More specifically we rely on the same sample of respondents discussed earlier – the social democratic core – and then estimate regression models of the following form:

First, we estimate the correlation of these factors for switchers. We define Y_i,t as being ‘1’ whenever a social democrat suggests to have voted for a different party then the Social Democrats. We estimate all regressions using ordinary least squares, cluster our standard errors at the respondent level (i), and use country-level fixed effects (α_c) to control away any country-specific differences.

Figure 5.6 reports the findings of the regression model – we rely on a coefficient plot reporting the point estimates of these regressions as markers along with the shaded 90%, 95%, and 99% confidence intervals. All coefficients are standardized such that a direct comparison of the coefficients is possible. It becomes strikingly visible that the fundamental issue of social democratic parties is to attract younger cohorts: Specifically, the generations born after 1981 are flocking away from Social Democracy in larger numbers. By contrast, the “war generations” – being born before the end of World War II – are sticking with Social Democracy. The results discussed later further substantiate that it is the old core that sticks with Social Democracy and the young core leaving Social Democracy behind.

Figure 5.6 Who switches away from Social Democrats to any other party?

Note: OLS regression models with country fixed effects and standard errors clustered on respondent ID. Variables standardized by dividing by two standard deviations. Cohorts: WWI < 1930; WWII < 1946; boomer < 1965; Gen X < 1981; Gen Y/Z ≥ 1981.

The remaining variables behave as laid out in previous work: Economic hardship tends to drive voters away from Social Democracy, only recent unemployment appears to drive voters toward Social Democrats. But this effect is small and its insecurity is large. By contrast, higher education and union membership keep voters aligned.

But are there differences across party switchers? It is more than likely that the individual factors predict the different vote outcomes differently across parties. To better understand this, we split the voting outcome by party destination in Figure 5.7. Notice that in an effort to ensure readability, the x-axis varies across outcomes, meaning that the size of the coefficients cannot directly be compared across outcomes. One factor remains relevant across all destinations: age cohorts. No matter which new party former social democratic voters choose, the cohort they are born is the major factor correlated with switching away. Thereby, the youngest generations are the ones that are most likely to switch to the fringes. While the generation X seems the one most driven to dealign with party politics altogether.

Figure 5.7 Who switches to whom?

Note: OLS regression models with country fixed effects and standard errors clustered on respondent ID. Variables standardized by dividing by two standard deviations. Cohorts: WWI < 1930; WWII < 1946; boomer < 1965; Gen X < 1981; Gen Y/Z ≥ 1981.

As others have suggested (e.g., Kitschelt Reference Kitschelt1994; Rennwald and Evans Reference Rennwald and Evans2014; Gingrich and Häusermann Reference Gingrich and Häusermann2015), the relevance of occupation varies tremendously across party voting. While manager and professionals stick to the Left, service, elementary and machine workers are disenfranchised and flocking in smaller numbers to the Radical Right. Similarly lower income is a strong factor dealigning social democrats from party politics altogether, and similar effects are visible for having been unemployed in the past.

Altogether, these findings substantiate theoretical arguments elsewhere but also make it clear that in order to understand where Social Democracy is today, we need to understand within individual voting patterns. We do not find stark deviations from this general pattern when looking at the three countries one by one.

6.4.1 Room to (Re-)grow? Mobilization Potential of Social Democratic Parties

Figure 6.1 plots the share of respondents in each country who indicate that they can well imagine (≥6) voting at some point for the social democratic party, that is, the mobilization potential. The figure also indicates (darker area) the vote share in the last national general elections. All numbers are averaged per decade. The ratio between the vote share and the potential can be interpreted as the “electoral yield,” that is, the extent to which the party managed to mobilize and realize its potential. For seven out of ten countries, the yield of social democratic parties ranges on average between 60% and 85% of their potential. Only in France, the Netherlands, and Italy are yields markedly lower, which makes sense in terms of the massive electoral losses the social democratic parties had experienced in the national elections.

Figure 6.1 Social democratic electoral potential (voting propensity ≥ 6) and mobilization (vote) in ten European countries, 2000s and 2010s

The main insight from Figure 6.1 is that social democratic parties in Western Europe are generally a viable electoral option for many more voters than those who actually gave them their vote. On average, around 50% of all voters can imagine voting for the social democratic party at some point in their life. Importantly, the unrealized potential is not simply a function of the electoral performance but on average remains rather stable over time (with the exception of France). However, Figure 6.1 also shows that in all countries, the electoral potential of social democratic parties has rather declined, in almost half of them by more than 10 percentage points.

Regarding robustness, there are two important observations: First, the findings are robust to a cutoff point of 7 instead of 6. Potentials are on average about 10 percentage points lower with the higher cutoff point, but the comparative pattern of (realized and unrealized) potentials remains robust. The observation of declining potentials between the first and second decades of the 2000s is also robust. I conducted a second robustness test by excluding respondents with large and less specific consideration sets of three or more parties. Thereby, I want to see if I overestimate the potential because some (especially younger) respondents may be open to voting for many different parties. When excluding these respondents with larger choice sets, the “unrealized potentials” decline quite sharply (from 20–40% on average to 10–20%) in those countries with very fragmented left party spectrums – especially the Netherlands, France, Germany, Finland, and to some extent Spain. This indicates that the high electoral potentials for social democratic parties indicate a general predisposition of very many voters to choose a party of the left-wing spectrum. However, it seems less indicative of a specific preference for the social democratic party in particular.

Figure 6.2 pools the countries by regions and compares electoral potentials and electoral yields across party families for the two decades. We notice that the electoral potential of social democratic parties is actually on average among the highest, together with the potentials of moderate right and green/left-libertarian parties in Nordic and Continental European Countries. In Southern European countries, the high potential of social democratic parties stands out even more, clearly outnumbering the electoral potential of green/left-libertarian and radical left parties.

Figure 6.2 Electoral potential (voting propensity ≥6) and mobilization (vote) for different party families in ten European countries, 2000s and 2010s

Hence, Figures 6.1 and 6.2 combined show that there indeed seems to be room to (re-)grow for social democratic parties, since actual vote shares do not exhaust their potentials by far. There does not seem to be a sort of a “ceiling” of social democratic voting in particular and left voting in general.

However, the competitive situation of social democratic parties seems challenging, and increasingly so: While the potential of radical left parties seems more narrowly confined, the massive, and massively increasing potentials of green/left-libertarian and radical right parties are striking. In Nordic and Continental European countries, the electoral potential of green and left-libertarian parties equals the one of the Social Democrats, but a much larger share of it is (still) unrealized. At the same time, the electoral potential of radical right parties has almost doubled in this region between the first and the second decades of the current century. In Southern Europe, both green and radical right parties have strongly increased their electoral potential over the same period. Hence, these numbers clearly show that the competitive environment for social democratic parties has markedly intensified both within the left field and at the conservative end of the ideological spectrum.Footnote ¹

6.4.2 Inward and Outward Overlaps with Rival Parties: By Region, Class, and Education

Now that we have established that there is in most countries a substantial unrealized electoral potential, where are these voters to be found? In other words, which rival parties have voters who also include the Social Democrats in their consideration set? And which other parties to social democratic parties include in their consideration sets?

I define as inward resp. outward overlaps those voters of rival resp. social democratic parties who at the same time report a voting propensity for the social democratic party resp. a rival party of 6 or higher (with robustness analyses for a threshold of 7). Importantly, there are two ways to estimate the size of these overlaps. On the one hand, one can ask between which parties the overlaps are largest. This information is important, as is reflects which other party electorates are “closest in reach.” However, if such “close” parties are very small, a valid estimation of the magnitude of the overlaps requires a calibration by party size. Consequently, in order to compare inward and outward overlaps across parties, they need to be calibrated by a common denominator.

I take the first perspective in this section – to show the prevalence of overlaps with other party families – and the second perspective in Section 6.4.3 – to compare potential gains and losses of different programmatic strategies across the entire party spectrum.

Table 6.1 presents inward and outward overlaps for the regions of Nordic and Continental Europe. The table above (inward overlaps) indicates the share of respondents who have actually voted for green/left-libertarian, radical left, moderate right, or radical right parties, or who have abstained, but who at the same time indicate that they can just as well imagine voting for social democratic parties. The table below shows outward overlaps, that is, the shares of respondents who have voted for the social democratic parties but report a high voting propensity for a different party family, as well. Region means are weighted averages.

We see that overlaps are highest with green and radical left parties. About half of left-libertarian voters on average can as well see themselves voting social democratic and vice versa. The inward overlaps are similarly high with radical left parties, but with more variation. Hence, both inward and outward overlaps concentrate within the left field.

These high inward overlaps in the left field are followed by the group of abstentionists, among whom about a third could also imagine voting for the social democratic party. They are lower for moderate right and – very clearly – radical right parties. On average less than a fifth of radical right voters would consider ever voting social democratic and even fewer (7–13%) of social democratic voters can imagine ever voting radical right. There is quite some variation in the levels across countries (with Finland in the 2000s being an extreme case), but the rank ordering of overlaps is very consistent.

Table 6.1 already casts doubt on the widespread assumption that social democratic parties mainly compete with radical right parties over the votes of lower social classes. These doubts are further confirmed by the disaggregation of inward and outward overlaps into different education groups in Figure 6.3 (and different subjective classes in the Appendix of this chapter).Footnote ² In Northwestern Europe, about 70–80% of all respondents on average are in the categories of the medium and highly educated. The overlaps with green and left-libertarian parties, moderate right parties, as well as abstentionists exceed this baseline level by far: With these party families, overlaps clearly concentrate among the middle and highly educated voters. With green/left-libertarian parties, the overlaps even concentrate among the most highly educated category (i.e., respondents who studied on after the age of 20). With the Radical Left and Radical Right, the shares are closer to the distribution in society, but even there, we see that overlapping voter potentials are found mostly among medium- and highly educated voters. This picture is confirmed when looking at subjective social class (see Appendix Figures 6.A1 and 6.A2): Overlaps, especially inward overlaps, exceedingly concentrate among middle-class voters.

Figure 6.3 Inward and outward overlaps by education, Nordic and Continental Europe (AT, DK, FR, FI, DE, NL, and SE)

Note: “Residual” denotes nonresponses and students.

Moreover, it is important to note that the distributions look strikingly similar when it comes to inward and outward overlaps. In other words, it is not the case that social democratic parties “asymmetrically” risk losing working-class voters, while they may win over middle-class votes. Rather, both potential gains and losses concentrate in the middle and upper classes, especially when it comes to those political parties that the overlaps are highest with. This finding is consistent with what Abou-Chadi and Wagner find with regard to vote switching (see their chapter in this volume).

Moreover, Figure 6.3 shows a somewhat gloomy overall picture from the perspective of social democratic parties: Potential gains from other party families have on average declined over time, whereas potential losses tend to have increased. In an increasingly realigned and differentiated party landscape, the social democratic parties, on average, seem to lose the favors of other party voters, even within the left field.

The picture we find for Southern Europe is similar in rank order to Nordic and Continental Europe but at a lower level of overlaps (see Table 6.2). Both inward and outward overlaps are strongest with green/left-libertarian and radical left parties (i.e., within the left field) and on average lower with moderate and radical right parties. Abstentionists rank between the Left and the Right when it comes to inward overlaps. Inward and outward overlaps are on average relatively similar in size, but outward overlaps are overall smaller, especially to the Radical Left (probably indicating that those social democratic voters who wanted to join the new left parties have mostly done so). There is one exception to point out, that is, the strongly increased inward overlaps from radical right parties in the 2010s (stemming mostly from Portugal and Spain), which are not at all matched by symmetrical outward overlaps. It may be too early to tell if these overlaps are sustainable over time.

Hence, the pattern and rank order of overlaps is similar to Northwestern Europe, but it is important to notice that, as expected, levels of overlaps (especially with green and left-libertarian parties) are smaller, in line with the overall relatively recent reconfiguration of the left field and with the antiestablishment appeal of the new parties on the left in Southern Europe.

The share of lower-class voters in the sample is on average higher in the Southern European than in Nordic and Continental European countries, both in terms of education years and subjective class. Still, overlaps with the green and left-libertarian parties (including new left parties such as Podemos or Sinistra e Libertà) overproportionally concentrate among the middle classes and highly educated voters (see Figure 6.4). Beyond that, however, it is important to note that in Southern Europe, overlaps with the Radical Left, the Radical Right, and abstentionists are more concentrated among the lower- to medium educated and working-class voters.

Figure 6.4 Inward and outward overlaps by education, Southern Europe (IT, ES, and PT)

Note: “Residual” denotes nonresponses and students.

In sum, we see across the regions mainly commonalitiesFootnote ³: Overlaps are highest within the left field, both with regard to inward and outward overlaps. Overlaps with the right, and particularly with the Radical Right are on average much lower. In particular, hardly any social democratic voters indicate they are tempted to vote for the Radical Right. Finally, the Figures 6.3, 6.4, 6.A1, and 6.A2 show that the electoral competition social democratic parties face does not concentrate around voters from lower social classes. Rather, where we observe the highest overlaps, competition revolves around middle-class voters with medium and high education. Finally, a last consistent finding across all countries and regions is that – especially within the left field – inward overlaps tend to decline over time, while outward overlaps tend to increase. This corroborates the increased competitive challenges social democratic parties face.

6.4.3 Programmatic Strategies and Electoral Payoffs

In this section, I integrate the comparison of inward and outward overlaps – in terms of both magnitude and respective voting propensities – to arrive at a very approximative evaluation of likely payoffs of the four possible programmatic strategies. The most important change with regard to the previous analyses is that from now on, I present inward and outward overlaps relative to the same denominator – the entire electorate (i.e., all respondents who indicate having voted for a party in the sample or abstained). This calibration is important, since the relative size of the respective potential gains and losses electorates obviously matters for gauging payoffs: Even if a very high share of, for example, green voters can imagine voting social democratic, this implies very different payoffs depending on whether the green party electorate is small or large.

I will subsequently discuss specific comparisons of party overlaps: When discussing the likely payoffs of a New Left strategy, I will compare inward overlaps from the green and left-libertarian parties to outward overlaps toward radical right parties and to moderate right parties. For the Old Left strategy, I compare inward overlaps from the Radical Left to outward overlaps to the Moderate and Radical Right. A Centrist strategy is supposed to appeal mainly to moderate right voters but may entail potential losses to radical left (and possibly green) parties. Finally, a Left National strategy appeals to inward overlap voters from the Radical Right, at the cost of potential losses to green/left-libertarian and possibly radical left parties. The comparison of these overlaps relies on two parameters: first, the magnitude of the groups, indicated as a share of the overall electorate; second, the average voting propensity among the voters in the overlap. These two indicators provide information on the likely balance of potential gains and losses, as well as on the likeliness that such gains and losses might actually be realized.

Figure 6.5 provides the empirical foundation for the discussion of overlap magnitudes. It displays the percentage of voters of one party who include the relative other party in their consideration set. Inward and outward overlaps are indicated by means of directed arrows and the weight of the lines is proportional to the size of the overlap. I have pooled numbers by region and decade, but the estimates still rely on relatively limited numbers of observations, especially when it comes to overlaps with the Radical Left and the Radical Right. Hence, our focus is on tendencies and regularities, rather than on precise point estimates. The average voting propensities among the overlaps are reported in Figure 6.A3 in the appendix of this chapter.

Figure 6.5 Inward and outward overlaps as a share of the overall electorate, comparison of magnitudes

To gauge the implications of a New Left programmatic strategy, we want to know whether inward overlaps from green/left-libertarian parties outweigh outward overlaps to the Radical Right. Figure 6.5 shows the potential gains from green and left-libertarian parties compared to the potential losses to radical right parties as a share of the overall electorate and differentiated by region and decade. We see that on average, potential gains seem to outweigh potential losses in Nordic and Continental European countries (7% vs. 2–3%) but not in Southern European countries. For Nordic and Continental European countries, inward overlaps amount to twice or three times the outward overlaps. The average voting propensity of green overlap voters for the Social Democrats is about the same as the voting propensity of social democratic overlap voters for the Radical Right (i.e., about 6.5–6.7 on the scale 1–10). The question, of course, is, whether such a strategy would also lose voters to the Moderate Right, to which outward overlaps are larger. Given the relatively conservative attitudes of centrist voters especially in the field of immigration (cf. the chapter by Abou-Chadi et al. in this volume), one might expect at least some effect in this direction, which might, however, be partly compensated by additional gains from the Radical Left. Hence, overall, a New Left strategy appears likely to yield positive payoffs for social democratic parties in Continental and Nordic Europe.

The same does not hold for Southern Europe. While estimated outward overlaps to the Radical Right are very low, so are inward overlaps from within the left field. Given that the new alternative parties in the left field have mostly mobilized against the mainstream (establishment) parties, it makes sense that their voters are reluctant to consider the mainstream left as an electoral option. This somewhat lower inward potential within the left field explains why the New Left strategy seems less promising in Southern Europe than in the Continental and Nordic countries.

To evaluate likely payoffs of an Old Left strategy, we compare inward overlaps from the Radical Left to outward overlaps to the Moderate Right.Footnote ⁴ From Figure 6.5, it appears that an Old Left strategy would on balance seem highly risky, as inward overlaps at the left end of the spectrum amount to only about half of the outward overlaps at the centrist end of the spectrum in both Northwestern countries (3–4% vs. 5–8%) and in Southern European countries (2% vs. 4–5%). A comparison of average voting propensities reinforces this conclusion for Northwestern Europe: Radical left voters who do include the Social Democrats in their consideration set on average rate their propensity to vote for them only at 6.17, that is, it seems that they would be hard to “win over,” whereas the propensity of outward overlapping voters for the Moderate Right is slightly higher (6.36). On this basis, the (small) potential gains from the Radical Left seem indeed hard to realize, in comparison with the more likely losses to the Moderate Right they may entail. In Southern Europe, by contrast, outward overlapping voters on average have a rather low voting propensity for the Moderate Right.

A Centrist strategy is supposed to appeal to moderate right voters by means of more moderate policy positions mainly on the socioeconomic dimension but entails the risk of shying away voters to radical left parties in particular, and – given the very progressive policy preferences of green and left-libertarian voters on both economic and social issues – also to some extent to green and left-libertarian voters. Figure 6.5 indicates that potential gains from the mainstream right parties are indeed on average somewhat lower than potential losses to the Radical Left parties but certainly much lower than potential losses within the left field overall (i.e., green/left-libertarian and radical left parties). Inward overlaps from moderate right parties amount on average to 6–9% in Northwestern Europe and 4% in Southern Europe, which are relatively low numbers given the size of the moderate right party family. In addition, the average voting propensity of these overlapping moderate right voters for the Social Democrats is very low (6.1–6.2). On the other hand, the combined outward overlaps to green/left-libertarian and radical left parties amount to 16–22% in Northwestern Europe and 17–18% in Southern Europe, with much higher average propensities (6.3–6.85). Hence, on balance, there is reason for skepticism whether a Centrist strategy would attract large vote shares to social democratic parties and/or the Left field.

Finally, a Left National strategy would seek to appeal to inward overlapping voters of the Radical Right, at the risk of losing outward overlapping voters within the left field to green/left-libertarian and radical left parties. The assessment of the likely payoffs of such a strategy is quite straightforward: Potential losses are likely to outweigh potential gains by far. While inward overlaps from the Radical Right amount to 0–2% of the electorate in both regions, potential ensuing losses to the green/left-libertarian parties alone go in the numbers of 10–15% relative to the entire electorate. Outward voting propensities for the green/left-libertarian parties are also higher than among inward overlapping voters, at least in Northwestern Europe where outward overlapping voters to the Greens have on average very high propensities. Even if we include orthogonal additional potential gains and losses (from the Moderate Right and toward the Radical Left) into the assessment, the likely balance of a Left National strategy still seems to remain negative. This approximative assessment of payoffs is consistent with analyses of vote switching (see the chapters by Abou-Chadi and Wagner, as well as by Kitschelt and Rehm in this volume), as well as experimental survey evidence on the resonance of programmatic appeals within the potential electorate (see the chapter by Abou-Chadi et al. in this volume).