The rightward shift and electoral decline of social democratic parties under increasing inequality

Abstract

Recent electoral results reveal a pronounced decline in the fortunes of Social Democratic parties. Much of the decline debate has revolved around their rightward policy shifts, which have turned Social Democrats away from their founding principle of equality in an age of increasing inequality. Thus, this article examines the interconnections of these major changes in the Western political economy. In doing so, it contributes to the identification of income inequality as a key mechanism moderating Social Democratic policy offerings and their support. It does so through aggregate-level election results and individual-level survey responses on a sample of 22 advanced democracies, over 336 elections, from 1965–2019. Results reveal that rightward economic movements of Social Democrats significantly reduce their vote share under higher levels of income inequality or when they are combined with rightward socio-cultural movements. The findings provide an important explanation for the pronounced electoral decline of Social Democratic parties.

KEYWORDS :

• Income inequality
• voting
• social democracy
• mainstream left
• party competition

Social Democratic parties across the West have undergone substantial changes over the past generation (Mudge 2018). Prime among them is a pronounced economic shift in the policy domain. Much evidence exists that this rightward shift towards the centre has caused substantial damage to Social Democratic parties as enduring institutions in the long run (Arndt 2013; Horn 2020; Karreth et al. 2013; Loxbo et al. 2019; Schwander and Manow 2017). The shift has occurred despite equality being a founding principle of social democracy and whereby the promulgation of redistribution was once a leading remit of Social Democratic parties (Bartolini 2000; Mudge 2018). Concurrently, income inequality has risen substantially over this period and has come to be viewed as one of the greatest challenges facing the advanced West (Piketty 2014; Stiglitz 2013). However, extant literature has not examined the impact of income inequality on Social Democratic party decline. According to ‘conflict theory’, Social Democrats should benefit electorally from rising inequality, but this article shows that they only do so if they offer clear left-leaning economic positions, as party supply needs to meet citizen demand.

The welfare state has traditionally been a strong means of mobilisation for Social Democratic parties, as they have been perceived by electorates as being dedicated to first expanding and later preserving the welfare state (Schumacher et al. 2013; Schwander 2019). Consequently, Social Democratic parties tend to benefit if welfare state issues are salient during electoral campaigns (Bélanger and Meguid 2008). Despite this advantage, analyses of party manifestos have confirmed in what Lipset (2001) describes as The Americanisation of the European Left, the gradual movement of Social Democratic parties towards the ideological centre, through the abandonment of many distinctly traditional leftist positions (Arndt 2014: 780). This was achieved through greater embrace of the market, via increasing financialization, privatisation and deregulation, as well as reductions in tax and the welfare state, causing the parties to become less egalitarian during this period of rising inequality.

Prime strategic explanations for this party movement have been to foster an image of being a strong steward of the economy to increase their chances of gaining office, or to be more palatable as coalition partners (Keman 2011; Kraft 2017). In Western Europe, the motivations also lay with the goal of mobilising new constituencies to become ‘catch-all’ parties that would be offsetting the decline of their traditional base through globalisation and declining industrialisation (Evans and Tilley 2012). While in Eastern Europe, it was viewed by leaders as a way of distancing themselves from their Communist legacies and to increase the prospects of attaining European Union membership (Benedetto et al. 2020).

Initially, this ‘Third Way’ party strategy seemed to be achieving strong election results in the late 1990s and early 2000s, however, there is much evidence that it has achieved the opposite effect, with substantially declining vote shares across Europe (Arndt 2013; Horn 2020; Karreth et al. 2013; Loxbo et al. 2019). Many of these parties typically achieved over 40 percent vote share but have recently collapsed into the low 20 s and some even into single digits (see Figure 2). The evidence is ‘mixed regarding mobilization of new target constituencies’ (Arndt 2014: 780), and it has led traditional supporters to increasingly embrace new challenger parties. For example, in one of the most pronounced cases of welfare reform – the German Social Democratic (SPD) Hartz laws of 2003 to 2005 – Schwander and Manow (2017) find the reforms as contributing to the party’s recent decline by inducing many former supporters to abstain from voting, or to vote for a new socialist party (Die Linke). This rival party on the left that has been able to establish itself firmly in the German system, owing largely to these SPD losses (Bowyer and Vail 2011). Lost support has disproportionately occurred among the lower classes, despite Social Democratic parties typically expected to represent the interests of citizens in the bottom half of the income distribution, since they provide the primary conduit for the constituency to exercise demands for redistribution to combat rising inequality.

Figure 1. Social democratic vote share and Gini Index, 1965–2019.

Accordingly, this study examines the policy changes and decline of Western Social Democratic parties over the past half century. It does so through both aggregate- and individual-level tests, as to whether income inequality and Social Democratic rightwards policy movements, negatively impacts their support. Thus, it fills important gaps in the literature. As there exists no comparative work linking income inequality to both electoral behaviour and Social Democratic party positions; nor has the second dimension been investigated simultaneously alongside the state-market economic dimension in determining Social Democratic electoral decline. Furthermore, previous research has been primarily focussed on Western Europe, so the incorporation of five non-European countries,¹ expands our knowledge beyond the usual regional scope.

This study is situated at the intersection of the comparative political economy literature on the political consequences of inequality, as well as the electoral behaviour literature on party programmatic shifts and vote choice. A review of this literature, along with the key hypotheses of the study, is discussed in the next section. The research design is then outlined, followed by a test of the expectations on a sample of 22 advanced democracies, over 336 elections, from 1965 to 2019. The study concludes with a discussion of the key implications and avenues for future enquiry.

Social democratic positions, income inequality and decline

Social Democratic parties have witnessed electoral decline in recent years, in what is one of the most consequential global political trends. Combined with the rise of populism, this has led many to question whether social democracy is in crisis. Central to the debate is the pronounced reduction in class voting compared to the heyday of Social Democrats in the 1960s and 70 s. Much evidence now exists that the far-right has made substantial inroads among the working class in recent years (Afonso and Rennwald 2018; Mosimann et al. 2019; Rydgren 2013) and Social Democratic parties have suffered losses amongst its traditional working-class base (Arndt 2013; Gingrich and Häusermann 2015; Rennwald 2020). Some scholars believe this is occurring because voters are increasingly being mobilised via identity rather than distributive politics (Häusermann et al. 2020; Häusermann and Kriesi 2015). Most notably, populists on the far right have been able to capitalise and make identity issues more salient, which has precipitated a debate surrounding the benefits of Social Democrats pursuing a so-called ‘accommodation strategy’. However, changing cleavages have been accompanied by rising income inequality throughout the West and according to conflict theory, inequality’s rise should then be making distributive concerns more salient, but Social Democrats need to cater to these demands to benefit electorally.

Social democratic positions

Traditionally, Social Democrats are the primary actors actively promoting policies that favour labour and the lower classes through state development and redistribution. In contrast, business and centre-right parties tend to promote marketisation and income concentration at the top, while opposing redistribution (Huber et al. 2019). Consequently, both the working class and labour unions have traditionally been a strong base of support for Social Democrats, as their vote shares were higher in countries with high levels of union and party membership in the post-war era (Hopkin 2020). However, the neoliberalization of Social Democratic parties through their gradual embrace of privatisation and job market flexibility since the 1980s, has increasingly distanced the parties from their base in many Western countries (Mudge 2018). For example, Thau (2018) provides evidence that from 1961–2004, the Danish Social Democratic Party increasingly associated with business and attempted to appeal to the middle class, while substantively diminishing its appeals to the working class in later years.

Despite Social Democratic economic moderation, recent evidence shows that voters do indeed listen to parties and understand their policy messages, especially on the issue of redistribution (Somer-Topcu et al. 2020). As Rueda and Stegmueller (2019: 187) demonstrate using European Social Survey data, that the poor are ‘uniformly in favour of redistribution and therefore more likely to vote for redistributive parties’. Rueda (2018) also finds that individuals with high redistribution preferences are 70 percent more likely to vote for leftist parties and this article finds that Social Democratic voters are overwhelmingly in favour of redistribution (see Online appendix A11). Consequently, as Social Democrats were the principal advocates of an expansive welfare state where they were able to establish issue ownership, and as their traditional base largely supports left-leaning economic policy, it will be tested whether their vote share increases, if they adopt these policies:

H1: Social Democratic parties gain vote share if they adopt stronger redistributive policy positions.

Conflict theory and inequality

If Hypothesis 1 is supported, then the electoral decline of Social Democratic parties is likely owing to factors other than their economic policy offerings. However, the second hypothesis, which is based on conflict theory, introduces income inequality as an important conditioning mechanism into the relationship between Social Democratic policy offerings and their support.

Conflict theory builds on Meltzer and Richard’s (1981) median voter model and posits that increasing inequality leads to demands for a more generous redistributive policy because the median voter has more to gain from redistribution under rising inequality. Consequently, rising inequality should increase support for Social Democratic parties if they offer more redistribution. However, the model is challenged in practice because countries with the highest market inequality tend to redistribute the least and countries with the least market inequality redistribute the most, resulting in the so-called ‘Robin Hood paradox’ (Moene and Wallerstein 2001).

This could be occurring due to multiple reasons. People largely underestimate the true extent of income inequality, often by substantial amounts (Hauser and Norton 2017). Second-dimension issues can also trump economic preferences, especially when people are diverted away from the pursuit of their material self-interest. For example, parties on the right tend to emphasise values and non-economic issues to distract voters’ attention away from their economic interests when inequality is high (Tavits and Potter 2015). Correspondingly, leftist parties should be able to counter this strategy by emphasising redistribution during periods of high inequality, to capitalise on the increased potential constituency that inequality generates. However, Barth et al. (2015), demonstrate that increased inequality causes parties on the left, across the Organisation for Economic Co-operation and Development (OECD 2020), to shift their manifestos to the right and towards less welfare generosity. Unless turnout is high – as Pontusson and Rueda (2010) find that higher turnout causes parties on the left to respond to greater inequality with more redistribution. But unless parties are effectively offering redistribution under high inequality, then turnout is unlikely to be elevated for lower classes (Polacko 2021).

The rightward shift of Social Democrat parties has coincided with a decline in living standards for many across the West. Roughly 10 percent of national income in most countries has moved from the share of labour to capital (Hopkin and Blyth 2019), the middle class has been thoroughly squeezed (Nolan and Weisstanner 2021), and the bottom half of the American population has witnessed zero real income growth since 1980 (Piketty 2020: 835). Therefore, the failure of Western governments to act on increased demands for redistribution stemming from increasing inequality, points to a supply issue, which could in part be owing to Social Democratic rightward economic policy movements. Thus, it will be measured whether Social Democratic party support increases during periods of higher inequality, especially among the lower classes, only if those parties adopt stronger redistributive policy positions in their election manifestos:

H2: Social Democratic parties gain vote share if they adopt stronger redistributive policy positions, only during periods of higher income inequality (H2a). The effect will be most pronounced among lower income earners (H2b).

Nativism and accommodation

Over the past few decades Western class coalitions have altered due to globalisation and the changing structure of economics and society. The decline in the industrial workforce has been associated with the decline of Social Democratic support (Benedetto et al. 2020) and the parties now rely on middle-class voters to a much greater degree than in the past. For example, manual workers comprised half of all employed voters of mainstream Social Democrats in the early 1980s, which has now plummeted to around 20 percent, while middle-class support has more than doubled from roughly 25 to more than 60 percent (Engler and Zohlnhöfer 2019: 1621). Less educated and lower income private sector workers were also more likely to vote for left-wing parties between 1950–1970, which is the opposite for the period 1990–2020 (Piketty 2020: 863–867). Middle-class voters now tend to be highly educated and socio-culturally liberal, which has put them at odds with a changing working class that has gradually become more concerned over increased immigration and cultural liberalism since the 1980s (Gingrich 2017; Häusermann et al. 2013; Piketty 2020: 40). These concerns tend to stem from increased migrant labour market competition in an era of stagnating working-class wages; perceptions of a greater fiscal burden generated from immigrants in an era of austerity; and a perceived loss of cultural identity (Rueda and Stegmueller 2019: 145–146). This has contributed to lower class individuals moving rightward on the second issue dimension (Gingrich and Häusermann 2015).

Far right parties have capitalised on these changes and made substantial inroads with the working class, through a strategy that effectively blurs their economic positions (Rovny 2013), and/or by offering different shades of welfare chauvinism (Afonso and Rennwald 2018), which combine ‘defence of the nation with defence of the welfare state’ (Morgan 2018: 121). Thus, Social Democrats face a dilemma in maintaining a comprehensive welfare state in increasingly multicultural societies, without losing public support. This development has recently prompted a heated debate as to whether Social Democrats would now reap electoral gains if they offered more restrictive immigration. Spoon and Klüver (2020) offer compelling evidence in 15 elections for six Western countries from 1998–2013, that such an accommodation strategy does significantly benefit mainstream left parties. Denmark provides an instructive example. Hjorth and Larsen (2020) find in a Danish survey experiment that accommodation does cost votes among pro-immigration voters, but these voters tend to defect to other parties on the left, who typically support Social Democrats in coalitions. As was the case in the 2019 Danish election, where in a rare victory for Social Democrats, the Danish Social Democratic Party won power at the head of a left-wing coalition by pursuing accommodation (Kosiara-Pedersen 2020). However, increasing support for a left bloc is context dependent on a party system with multiple parties on the left that can form coalition government together, which is not always the case.

Others have challenged the merits of accommodation by introducing different scoping conditions into the analysis. Abou-Chadi and Wagner (2019) find that Social Democrats can gain vote share by taking up greater investment-oriented positions on the investment–consumption growth strategy spectrum (as opposed to the common state–market dimension) if they also take up liberal socio-cultural positions, and if unions are limited in their capacity to mobilise against such shifts. Social investment policies such as education, skills training, and childcare funding are popular with the public, but moving left on the second dimension may alienate the working class unless combined with greater redistribution. Hence, Loxbo et al. (2019) examine the extent that welfare state generosity conditions the electoral impact of shifts to the right on the second dimension. They find that Social Democratic parties only lose votes from rightward turns on the second dimension, when combined with low levels of welfare generosity.

The empirical scholarship on the electoral effects of accommodation remains inconclusive. Therefore, I test whether the adoption of more socio-cultural right-wing positions, will increase Social Democratic party support. However, combining this movement with right-wing positions on the prime economic dimension should prove detrimental to Social Democratic vote share. This stems from their historical reputational advantage as the leading protectors of the welfare state (Arndt 2013; Schwander and Manow 2017), which has gradually become eroded due their embrace of welfare retrenchment and austerity (Horn 2020). This brand dilution then becomes magnified when the party family simultaneously abandons socio-cultural liberalism.

H3: Social Democratic parties gain vote share if they adopt more right-wing positions on the socio-cultural dimension (H3a), unless combined with economically right-wing positions (H3b).

Data and methodology

Methodology

These hypotheses are tested from a uniquely created dataset based on the five waves of the Comparative Study of Electoral Systems (CSES 2019, 2020), party level data from the Comparative Manifesto Project (MARPOR) (Volkens et al. 2020), and other country-level data for 22 advanced democracies. The data based on the CSES surveys comprises 158,822 individuals in 85 elections, from 1996–2018, while the aggregate-level data contains 336 elections from 1965–2019.² As the hypotheses specifically apply to established democracies where party policy offerings are perceived to matter to voters, case selection is based on a country’s level of democracy and economic development. Therefore, OECD membership is required, as well as a longstanding history of a dominant Social Democratic party on the centre-left of a country’s party system.³

The data for the individual-level analysis contains individuals nested within countries over time, therefore, multilevel models are applied to repeated cross-sectional data. As the dependent variable is dichotomous, logistic mixed-effects models are estimated, which include both fixed and random effects. Since the number of elections per country or year are too small in the CSES to identify election-level variance within a country or year, it is unsuitable to include random effects for both levels (Bryan and Jenkins 2016; Park 2019). Hence, since the hypotheses primarily rely on changes over time, observations are clustered at the year-level to isolate the potential effects of time-specific factors on voting, with country fixed effects.

For the aggregate-level analysis, I rely on time-series cross-sectional (TSCS) data. I estimate ordinary least squares (OLS) and country-level fixed effects with robust standard errors to ensure that unobserved differences between countries do not bias the findings (Green et al. 2001) and because the hypotheses focus primarily on intra-country over-time variation in the dependent variable, rather than cross-sectional variation. Moreover, by deriving estimates from variation within the same countries, a wide range of unobservables that vary across countries but do not change much (such as institutions), is controlled for. To ensure consistency of results temporally, decade fixed effects are also estimated.

Aggregate-level variables

The dependent variable is SD vote, operationalised as the percentage of votes cast by the registered electorate for a mainstream Social Democratic party.

Income inequality is the first key country-level explanatory variable. The most widely used measure is the Gini, operationalised as the Gini Index (range 0 to 100), where 0 represents complete equality and 100 complete inequality. The adjusted after-tax Gini is employed rather than the market income Gini because the main mechanisms leading inequality to affect voting are most likely to operate via disposable income after taxes and transfers. To account for retrospective voting – as voters are typically backward looking with a memory of roughly one year when evaluating changes and impacts of the economy – the Gini is given a one-year lag (Lewis-Beck and Stegmaier 2013). Gini t-1 rates are obtained from the Standardised World Income Inequality Database (SWIID), which maximises both accuracy and coverage (Solt 2020).⁴

The policy variables of interest measure Social Democratic party positions on redistribution (SD economic position) for each election and their socio-cultural left–right position (SD culture position). The ideological scores are tabulated from MARPOR, which allows for the post-war comparability of party manifesto positions within and across countries (Ezrow and Xezonakis 2011). It is the most popular data for the study of political parties and with few exceptions offers reliable estimates correlating highly with expert and mass surveys (Benoit and Laver 2006). MARPOR relies on party manifesto statements classified into 56 policy categories over seven domains. To measure a party’s position on redistribution, I follow Lowe et al. (2011). This method takes better account of the proportional changes on the left–right scale than the traditional Laver/Budge methodology. The left–right score of the parties is calculated by summing up the logged percentages of all the sentences in the left category and subtracting their total from the sum of the logged percentages of the sentences in the right category.⁵ An SD economic position variable is then constructed based on this party score involving 15 relevant categories (left–right from −100 to 100) of the historically largest by vote share party on the centre-left, for each election. Similarly, an SD culture position is included, involving 17 relevant socio-cultural categories, which includes positions on the environment, equality, internationalism, law and order, minorities, multiculturalism, nationalism, and traditional morality. See Online appendix A4 for a detailed breakdown of both variables.

The party/bloc chosen for each election is readily discernible,⁶ as they remain the same for each country included in the dataset (see Online appendix A2). The primary Social Democratic party/bloc position is chosen as opposed to the entire spectrum of parties on the left in a party system (Abou-Chadi and Wagner 2019; Pontusson and Rueda 2010) because small parties located on the fringes are unlikely to be considered by most voters, and the largest Social Democratic party is likely to represent the most attractive option for lower income voters. Thus, this measure more accurately captures the ideological positioning and strength of parties (most notably left party strength) within the party system (Wilford 2019).

A wide range of party system controls are included. SD vote e-1 is Social Democratic party vote share in the previous election, which should highly influence current vote share. Including a lagged dependent variable creates dynamic estimations, which account for serial autocorrelation and are superior to other models or estimators (Keele and Kelly 2006: 203). The strength of both left competitors e-1 and radical right competitors e-1 in the party system is controlled for, as left parties have been shown to negatively impact vote shares of mainstream left parties (Bale et al. 2010; Iversen and Soskice 2006), and populist right parties have made inroads with the working class at the expense of Social Democrats (Afonso and Rennwald 2018; Mosimann et al. 2019; Rydgren 2013). Left party support has also been shown to be negatively affected by incumbency (Abou-Chadi and Wagner 2019). Therefore, a dummy incumbent variable is added and coded as 1 when the mainstream Social Democratic party controls government. A measure of electoral disproportionality is also introduced in the form of the Gallagher index, which is the average number of seats allocated to each electoral district.⁷ As cross-national evidence finds that the representation of low-income individuals is crucially dependent on the proportionality of electoral systems (Bernauer et al. 2015; Jusko 2017). Lastly, turnout is added, as higher turnout has been found to increase the vote share of leftist parties (Bartolini 2000; Pacek and Radcliff 1995). Each of the party system variables derive from MARPOR and the Comparative Political Data Set (CPDS) (Armingeon et al. 2019).

A variety of socio-economic controls are included. Such as union density, due to the substantial influence that unions have on mobilising working-class voters and in generating support for Social Democratic parties (Kerrissey and Schofer 2018).⁸ Government spending through transfers is an important instrument for reducing inequality. Thus, lagged government spending as a percentage of Gross Domestic Product (GDP) is added from the CPDS and supplemented by International Monetary Fund (Mauro et al. 2015) data. As poor economic conditions have been linked with voting for non-mainstream parties (Funke et al. 2016; Lewis-Beck and Stegmaier 2000), key measures of economic performance such as GDP growth and unemployment, are also included and lagged. Both derive from the CPDS.

Individual-level variables

The individual-level variables are all drawn from the CSES. The dependent variable is a dummy indicating a vote for the country’s leading Social Democratic party (SD voted).

Household income divided into five quintiles (lowest to highest), supplies the key explanatory variable at the individual-level. Quintiles were chosen because they provide the best means of comparison between income groups and across time. I include age and education as both can have an important effect on political attitudes and voting behaviour. Education is measured as a categorical variable ranging from 0 to 4 (low to high). Union members and females have been shown to be significantly associated with voting for mainstream left parties (Iversen and Rosenbluth 2006) and conservative voting has been linked with rural living (Jennings and Stoker 2016). Therefore, female, union, and rural dummy variables are also included. Lastly, a person’s political ideology is measured on a 0–10 left–right scale, as it is amongst the strongest and most consistent predictors of political preferences (Jost 2006).

Results

Descriptive analysis

Firstly, the trends in income inequality and Social Democratic vote share are investigated. Figure 1 plots the average of each variable since 1965 in the sample of 22 countries, via a local polynomial smoother. We can see that the diverging trends for each started in the same period. Inequality declined slightly until the mid-1980s when it then began its upward climb. It rose substantially throughout the 1990s, then levelled off and has been rising substantially again in recent years. The Gini Index has risen roughly 2 points over the entire period with most countries exhibiting a similar trend (see Online appendix A12). Contrastingly, Social Democratic vote share held steady until the 1980s at roughly 33–34 percent, when it then began to decline steadily, until falling dramatically from 31 to 25 percent since 2000.

Figure 2. Social democratic vote share by country, 1965–2019.

Figure 2 disaggregates Social Democratic vote share by country over time. We can see that Social Democratic parties in Europe have experienced the largest declines and have been practically eviscerated in France, Greece, and the Netherlands in recent years. The New Democratic Party in Canada, which has never held office at the federal level, is the only party to increase its vote share over time, albeit very marginally, and starting from a low base.

Next, we investigate Social Democratic policy changes. In both the economic and socio-cultural realms, they have moved rightwards over time. Figure 3 displays the rightward economic trend by plotting Social Democratic position by country from 1965–2019. We can see that the largest rightward turn occurred in the 1990s for most, before moving leftwards in recent years. Generally, the liberal market economies of the Anglosphere have moved the most rightwards along with Greece, which all happen to be higher inequality countries. The overall trend for SD culture position is similar, and each policy variable increases roughly 1 point on the left–right scale (see Online appendix A13).

Figure 3. Social democratic economic position by country, 1965–2019.

Aggregate-level estimation results

In order to test the hypotheses for SD vote at the aggregate-level, models are estimated using ordinary least squares (OLS) and country-level fixed effects with robust standard errors. Table 1 presents the results from three different models. Model 1 includes each of the aggregate-level variables. Previous Social Democratic vote is significantly correlated to SD vote at (p < 0.001). When unemployment t-1 and disproportionality is higher, people are significantly less likely to vote for Social Democrats. Counter to expectations, but in line with the recent results of Benedetto et al. 2020, I find a negative relationship between union density and Social Democratic voting. Income inequality exhibits little effect. When Social Democrats are economically right-wing, they gain vote share, but lose vote share when they are right-wing on the second dimension. Neither is statistically significant, and we do not find evidence in support of H1 – that Social Democrats benefit from leftist economic policy, or H3a – that an accommodation strategy works.

Table 1. Aggregate-level regression results predicting social democratic vote.

	Model 1	Model 2	Model 3
SD Economic Position	0.080	2.314**	−0.257
	(0.187)	(0.709)	(0.283)
SD Culture Position	−0.029	−0.006	−0.126
	(0.141)	(0.137)	(0.153)
Gini t-1	0.094	0.011	0.125
	(0.243)	(0.232)	(0.258)
SD Economic Position # Gini t-1		−0.077**
		(0.025)
SD Economic Position # SD Culture Position			−0.074*
			(0.032)
Government Spending t-1	0.135	0.120	0.131
	(0.141)	(0.130)	(0.141)
GDP Growth t-1	−0.047	−0.070	−0.082
	(0.149)	(0.136)	(0.136)
Unemployment t-1	−0.612*	−0.572*	−0.591*
	(0.231)	(0.206)	(0.213)
Union Density	−0.096	−0.125	−0.108
	(0.089)	(0.087)	(0.088)
Turnout	0.084	0.109	0.124
	(0.096)	(0.093)	(0.093)
SD Vote e-1	0.382***	0.394***	0.374***
	(0.067)	(0.062)	(0.063)
Incumbent	−0.625	−0.707	−0.558
	(0.790)	(0.804)	(0.745)
Disproportionality	−0.293**	−0.246*	−0.278*
	(0.101)	(0.101)	(0.110)
Left Competitors e-1	−1.080	−1.066	−1.056
	(0.783)	(0.779)	(0.797)
Radical Right Competitors e-1	0.253	0.391	0.177
	(0.596)	(0.550)	(0.597)
Constant	11.410	12.708	7.871
	(12.733)	(12.085)	(13.539)
R^2 within	0.49	0.50	0.50
R^2 adjusted	0.80	0.80	0.80
N	280	280	280

Note: Beta coefficients from a OLS regression with standard errors in parentheses.

* p < 0.05, ** p < 0.01, *** p < 0.001.

Model 2 tests for Hypothesis 2a – that as Social Democratic parties adopt stronger redistributive policy positions, their vote share increases during periods of high inequality – via an interaction between gini t-1 and SD economic position. The interaction is negative and significant. Figure 4 displays the average marginal effects of inequality by SD economic position on their vote share. It shows that the effect of inequality is positive when Social Democratic parties are left-wing on redistribution and that their vote share is substantially dampened the more economically right-wing the parties become. To aid in interpretation of the substantive magnitude of the interaction, I standardise gini t-1 so that it has a mean of 0 and a standard deviation of 1. We see that at a left-wing economic position, 1 standard deviation below the mean (-1.3), a 1 standard deviation increase in inequality is associated with roughly a 1.5 percentage point increase in SD vote. Whereas, at a right-wing economic position, 1 standard deviation above the mean (2), a 1 standard deviation increase in inequality is associated with a 0.5 percentage point decrease in SD vote. To put these effect sizes into context, a 2 standard deviation increase in inequality is equivalent to the increase in inequality in the United States over this entire period, and a 2 standard deviation rightwards economic party movement is associated with roughly a 2-percentage point decline in SD vote, Thus, we find support for Hypothesis 2a.

Figure 4. Average marginal effects of inequality by social democratic economic position on their vote with 95% C.I. (Model 2).

Model 3 then tests for Hypothesis 3 b – that the vote share of Social Democrat parties will decrease if they adopt right-wing positions on both dimensions – via an interaction between SD economic position and SD culture position. Recall that we found little support for accommodation in Model 1, since accommodation does not increase Social Democratic vote share. Here we find a similar result, as the interaction is negative and significant. Figure 5 displays the average marginal effects of SD culture position by SD economic position on SD vote. It shows that the effect is roughly zero when Social Democrats are at their economic mean. However, we see a strong positive effect when the parties are left-wing on both dimensions and that their vote share is substantially dampened the more right-wing the parties become on both dimensions. The substantive effect is that at 1 standard deviation below the economic mean, a 1 standard deviation rightward socio-cultural dimension movement is associated with a roughly 1 percentage point increase in SD vote. Whereas, at a right-wing economic position, 1 standard deviation above the mean, a 1 standard deviation rightward socio-cultural dimension movement is associated with a roughly 1 percentage point decrease in SD vote. We see an equivalent substantive effect to the inequality interaction of a 1 percentage point swing per standard deviation, except that a left-wing economic position provides roughly a 0.5 percentage point greater increase in SD vote under higher levels of inequality, than when combined with a right-wing socio-cultural position. The finding provides support for Hypothesis 3 b and is in line with the recent results of Loxbo et al. (2019).

Figure 5. Average marginal effects of social democratic socio-cultural position by economic position on their vote with 95% C.I. (Model 3).

Individual-level estimation results

To test for likelihood to vote for Social Democrat parties at the individual-level utilising the CSES, I specify a mixed-effects logistic regression. Table 2 presents the results from four different models. Model 1 includes each of the individual-level and 10 of the most relevant aggregate-level variables, due to the minimum degrees of freedom needed for the higher level in multilevel models (Bryan and Jenkins 2016). The variables largely perform as expected and most are significant. Less educated, urban, union members, and older people are significantly (p < 0.001) more likely to vote for Social Democratic parties. Individual political ideology is also significant at (p < 0.001), as left-wing people are much more likely to vote for Social Democrats (b= −0.322). As expected, lower income earners are also significantly more likely to support the party family.

Table 2. Individual-level regression results predicting social democratic vote.

	Model 1	Model 2	Model 3	Model 4
Age	0.006***	0.006***	0.006***	0.006***
	(0.001)	(0.001)	(0.001)	(0.001)
Female	−0.032	−0.032	−0.033	−0.032
	(0.018)	(0.018)	(0.018)	(0.018)
Education	−0.141***	−0.141***	−0.140***	−0.141***
	(0.009)	(0.009)	(0.009)	(0.009)
Income	−0.023***	−0.024***	−0.022**	−0.027***
	(0.007)	(0.007)	(0.007)	(0.007)
Rural	−0.180***	−0.181***	−0.176***	−0.180***
	(0.022)	(0.022)	(0.022)	(0.022)
Union	0.369***	0.370***	0.370***	0.369***
	(0.021)	(0.021)	(0.021)	(0.021)
Political Ideology	−0.322***	−0.322***	−0.322***	−0.322***
	(0.004)	(0.004)	(0.004)	(0.004)
SD Economic Position	0.024**	0.241***	−0.040***	0.032**
	(0.008)	(0.057)	(0.012)	(0.010)
SD Economic Position # Income				−0.003
				(0.002)
SD Culture Position	0.038***	0.042***	0.016*	0.038***
	(0.008)	(0.008)	(0.008)	(0.008)
SD Economic Position # SD Culture Position			−0.019***
			(0.003)
Gini t-1	−0.035	−0.016	0.008	−0.036
	(0.024)	(0.025)	(0.025)	(0.024)
SD Economic Position # Gini t-1		−0.007***
		(0.002)
GDP Growth t-1	0.039***	0.054***	0.052***	0.039***
	(0.009)	(0.010)	(0.009)	(0.009)
Unemployment t-1	−0.020*	−0.017*	−0.044***	−0.020*
	(0.008)	(0.008)	(0.009)	(0.008)
Union Density	−0.018**	−0.030***	−0.025***	−0.018**
	(0.007)	(0.007)	(0.007)	(0.007)
Turnout	0.029***	0.029***	0.031***	0.029***
	(0.004)	(0.004)	(0.004)	(0.004)
Disproportionality	0.040***	0.060***	0.069***	0.040***
	(0.010)	(0.012)	(0.011)	(0.010)
Left Competitors e-1	−0.180***	−0.272***	−0.287***	−0.180***
	(0.037)	(0.044)	(0.040)	(0.037)
Radical Right Competitors e-1	−0.109***	−0.046	−0.002	−0.109***
	(0.029)	(0.034)	(0.033)	(0.029)
Constant	2.154*	1.652	0.840	2.205*
	(0.921)	(0.938)	(0.954)	(0.922)
Variance	0.148**	0.227**	0.299**	0.148**
	(0.053)	(0.081)	(0.103)	(0.053)
Log Likelihood	−38852.15	−38844.6	−38826.02	−38851.34
AIC	77780.3	77767.2	77730.05	77780.68
BIC	78129.98	78126.08	78088.93	78139.56
Country fixed effects	YES	YES	YES	YES
N	73,281	73,281	73,281	73,281

Note: beta coefficients from a mixed-effects logistic regression with standard errors in parentheses.

* p < 0.05, ** p < 0.01, *** p < 0.001.

As expected, people living in countries with more disproportionality and competitors on both the left and radical right are significantly less likely to support Social Democrats. Both turnout and GDP growth t-1 are significant and positively related to SD vote. Income inequality exhibits a negative effect, but it is not statistically significant. We can see that when Social Democratic parties move rightward on either dimension, they are significantly more popular. Thus, we reject H1, but find evidence in support of an accommodation strategy (H3a) at the individual level.

Model 2 tests for Hypothesis 2a via an interaction between gini t-1 and SD economic position. The interaction is negative and significant at (p < 0.001). Figure 6 displays the average marginal effects of inequality by SD economic position on SD vote. Similar to Figure 4, it shows that inequality exhibits a positive effect when Social Democrats assume leftward economic positions, but that their vote share significantly decreases the less redistribution they offer. Although the magnitude is not as substantial as Figure 3 at the aggregate level, support is again found for Hypothesis 2a.

Figure 6. Average marginal effects of inequality by social democratic economic position on their vote with 95% C.I. (Model 2).

Model 3 then tests for Hypothesis 3 b – that the vote share of Social Democratic parties will decrease if they move rightwards on both dimensions – via an interaction between SD economic position and SD culture position. The interaction is negative and significant at (p < 0.001) despite both dimensions displaying a significant and positive effect in Model 1. Figure 7 displays the average marginal effects of SD culture position by SD economic position on SD vote. Here we see a positive effect of the second dimension of roughly 0.1 percentage points when Social Democrats are at their economic mean and roughly a 0.4 percentage point increase at a left-wing economic position and a 0.1 percent decrease at a right-wing economic position, when each are 1 standard deviation from the mean. Although the substantive effects are again not as strong as at the aggregate level, the finding provides further support for Hypothesis 3 b.

Figure 7. Average marginal effects of social democratic socio-cultural position by economic position on their vote with 95% C.I. (Model 3).

Model 4 tests for income effects in Hypothesis 2 by introducing an interaction between income and SD economic position, which is not significant. The direction is negative for every quintile and it is the fourth quintile that exhibits the greatest effect. Therefore, we do not find that lower income individuals experience a comparatively significant reduction in likelihood to vote for Social Democrats under higher inequality, and no support is found for Hypothesis 2 b.

Robustness tests

The findings are robust to additional controls and alternative model specifications. Additional controls are added to the aggregate-level analyses, including the economic position of the mainstream right party and the level of globalisation. Neither is statistically significant, and the main results all hold (see Appendices A5–A6).

Three alternative model specifications are undertaken at the aggregate-level. Including a lagged dependent variable (SD vote e-1) with fixed effects can potentially introduce bias and inconsistent estimations (Nickell 1981). Therefore, the aggregate-level estimations are re-run excluding: 1) decade fixed effects; 2) all fixed effects; 3) the lagged dependent variable. The main results hold for all three specifications except the policy interaction does not retain statistical significance without the inclusion of any fixed effects. However, it still displays a similar negative effect (see Online appendix A7).

An alternative measure of inequality is also employed. The Palma Ratio is calculated from the national income share of the top 10 percent divided by the bottom 40 percent. It correlates highly with Gini (r = 0.72). None of the variables substantively perform differently and the main results all hold (see Online appendix A8).

To ensure sure that the main results are not driven by the inclusion of any one country, I also undertake a jack-knife analysis, which indicates that the estimated interactions at both the individual- and aggregate-level are relatively stable and remain statistically significant whenever a country is excluded (see Online appendix A9).

Lastly, the individual-level analysis is re-run with country fixed effects clustered by year, and again we see the same results (see Online appendix A10).

Conclusion

This article investigates the recent decline of Social Democratic parties, in what is one of the most consequential political trends in recent decades. It does so through an examination of the responses of Social Democrats to increases in inequality through their policy offerings, and how they in turn, shape people’s preferences and participation in the political process. Social Democrats are the primary conduit for their traditional constituency to exercise demands on matters of redistribution, which can combat inequality. Based on data from 22 countries between 1965 and 2019, I find that rightward economic movements of Social Democrats, significantly reduces their support under higher levels of income inequality, or when they are combined with rightward socio-cultural movements.

The implications of these findings are extensive and contribute to the literatures on comparative political economy, voting behaviour, and party competition, Specifically, in terms of the political economy literature and its recent focus on the political consequences of income inequality, the findings point to a key interplay between inequality and the Social Democratic party family. As equality was a founding principle of social democracy and protection of the welfare state has historically been a strong means of mobilisation for Social Democrats, the party family’s turn away from these traditions, while inequality rises across the West, has been detrimental to their fortunes. Social Democracy is the party family most able to combat the alarming rise in income inequality, but the family is no longer a counterweight to a market-driven society fuelling inequality, and instead has largely become a contributor to it, through its focus on creating a globalised meritocratic class society based on educational inequality (Cuperus 2018). In fact, many Social Democrats have accepted inequality as a spur for individual motivation and innovation.

In terms of voting behaviour, the findings seek to address the pronounced supply–demand imbalance that exists in the literature. As scholars have so far primarily concentrated on the ‘bottom up’ or demand side of the equation and neglected the ‘top down’ party supply side. Therefore, this article builds on these findings by focussing on the policy movements of Social Democratic parties and introducing them into the framework. The rightward movements of the party family since the 1980s, are one of the most pronounced policy changes in the Western party system. At first, many Social Democrats were able to attain office, but in recent years many have witnessed substantial vote share declines. As Ferland and Dassonneville (2021) show, changes in the ideological positions of parties do cause voters to switch parties from one election to another. Although the effects are small, they are especially pronounced for the party closest to a voter ideologically. Thus, this article’s main findings can offer some explanation for recent voter shifts away from Social Democrats to challenger parties. As the findings indicate that political parties are not weak actors with regards to voting behaviour, they can impact their fate through strategic positioning on salient issues, such as income inequality.

Relatedly, the findings from this study also provide notable ramifications for party competition and strategy, as well as contributing to the debate on the electoral effectiveness of party policy shifts. The Social Democratic embrace of the economic centre has ‘watered down the left’s distinctive historical profile’ and left the parties unable to fully take advantage of increased discontent generated from the financial crisis, austerity, and rising inequality (Berman and Snegovaya 2019: 6). In fact, most Social Democratic parties embraced austerity in the aftermath of the financial crisis (Bremer 2018). Consequently, as voters are unable to spot much differentiation in the mainstream parties economically, then political competition moves to socio-cultural issues, where parties on the right have an advantage (Tavits and Potter 2015). In line with the recent results of Spoon and Klüver (2020) and Hjorth and Larsen (2020), I find some evidence that an accommodation strategy on the socio-cultural dimension can work, but it is detrimental if combined with a rightward economic strategy.

Finally, the findings of this study also shed light on why so much of the working class has abandoned Social Democratic parties. When Social Democrat’s shift rightward on redistribution, they appear to be alienating both their traditional base and much of the middle class, so they significantly lose vote share. As labour market precarity and unemployment have spread into the higher skilled middle class and flexibilization stemming from the deregulation of non-standard employment has created downward wage pressure among low- and middle-income insiders, while benefitting top earners (Schwander 2019; Weisstanner 2020). The brand dilution suffered from engaging in reforms that conflict with the party family’s traditional brand as welfare protectors, do not appear to compensate via gains from progressive movement on the second dimension, or from laying claim to acquiring economic and fiscal responsibility.

This study provides a novel avenue of enquiry into the Social Democratic party story, party competition, and the effects of inequality on political behaviour. It sheds greater light onto the issues of political inequality that persist throughout the West and draws on evidence in support of greater representation. The findings show that the policy choices presented to the electorate substantially matter for parties and for political behaviour, especially so in this age of increasing inequality. Future work might adapt this framework to better investigate the causal mechanisms behind both reduced support for mainstream parties, as well as examining if this can account for any voter movements towards the radical left and right.

Acknowledgements

I would like to thank Oliver Heath, Michael Lewis-Beck, and Kaat Smets for very valuable advice and comments, as well as two anonymous reviewers who helped substantially improve the article. Earlier versions of this article have also benefited from the advice and comments of Chris Hanretty, Ivica Petrovica, and Aidan Regan. Versions of the paper were presented at the 2019 German Association for Political Science (DVPW) Annual Meeting of the Section Political Economy at the University of Konstanz, the 2020 PhD Winter School in Political Economy at University College Dublin, and the 4th Annual University of Manchester Conference.

Disclosure statement

No potential conflict of interest was provided by the author(s).

Funding

The author received no financial support for the research, authorship, and publication of this article.

Additional information

Notes on contributors

Matthew Polacko

Matthew Polacko is a PhD candidate at Royal Holloway, University of London. His research focuses on the effects of income inequality on political behaviour. He has published in Political Studies, Party Politics, and American Behavioural Scientist, amongst others. [[email protected]]

Notes

1 Australia, Canada, Israel, New Zealand, United States.

2 Countries included: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Israel, Italy, Luxembourg (not in CSES), Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States.

3 Eastern Europe is excluded due to the instability of the region’s party system – particularly on the left.

4 Version 9.0 of the SWIID is used. It includes 100 separate imputations of inequality data, which allows for any uncertainty in estimations. For reasons of parsimony, the average estimate of these 100 imputed variables is taken from the gini_disp variable, which is an estimate of the Gini index in equivalized household market income.

5 MARPOR position computations assume that the marginal effect of an additional sentence is constant. However, a shift from zero to one would matter more for a policy position than a shift from 9 to 10 due to the diminishing impact of repeated emphasis. Hence, Lowe’s (2011) logged method addresses this by applying a ratio approach to the raw number of sentences, so that the relative balance and proportion of change on the left-right scale are accounted for, rather than just the quantity of sentences.

6 Party family classifications derive from MARPOR and are cross validated against the CSES.

7 Gallagher index is calculated by taking the square root of half the sum of the squares of difference between the vote percentage and seat percentage for each political party, in the two most recent elections.

8 Union density derives from ICTWSS version 6.1 and is taken from the OECD or interpolated in the roughly 10 percent of missing cases.