Social class, economic and political grievances and radical left voting: The role of macroeconomic performance

After the collapse of communism, it was expected that radical left parties (RLPs) across Europe would struggle to maintain their position in the political arena. Yet, RLPs clearly managed to survive. Now more than 15 years after March and Mudde (2005) asked ‘What's left of the radical left?’, one of the answers is that RLPs gained a permanent position in countries such as Germany (Die Linke), Greece (SYRIZA) and the Netherlands (Socialist Party). The electoral success of the radical left is often suggested to be sparked by the economic crisis at the end of the first decade of this century, which affected many European countries. A worse economic climate is argued to be conducive to radical left voting because the message of RLPs is one of economic transformation and income redistribution, which would particularly appeal to people during a recession. Empirical studies indeed found that RLPs attract more voters in countries with higher unemployment rates (March and Rommerskirchen 2015; Gomez and Ramiro 2019).

Economic downturns are not the only adversities that RLPs have been shown to profit from. Experiencing individual economic hardship makes the radical left more appealing as well. Low income, unemployment, insecure employment, poor economic well-being and relative deprivation have all been linked to radical left voting (Rooduijn et al. 2017; Rovny and Rovny 2017; Rooduijn and Burgoon 2018; Burgoon et al.2019). In addition to these economic difficulties, being in a low socioeconomic position in society, and with that, having poorer prospects and facing higher economic risks, promotes voting for RLPs. Notably, previous research showed that a large part of the radical left electorate consists of people in lower social classes, such as manual workers and lower routine non-manual workers (Sperber 2010; Bowyer and Vail 2011; Gomez et al. 2016; Ramiro, 2016; Rooduijn et al. 2017).

The preference of those in lower classes to vote for (radical) left parties is traditionally referred to as class voting (Lipset and Rokkan 1967; Nieuwbeerta 1996; Evans 2000; Jansen et al. 2013). Class voting literature argues that members of the working class use elections to express dissatisfaction with their position in society (Lipset and Rokkan 1967). Although this dissatisfaction can stem from both cultural and economic issues (Langsæther, 2019), classic class voting studies focus on the economic dimension of class (Evans 2017; Rennwald 2020). Indeed, RLPs campaign first and foremost on the topic of the economy and redistribution of income (March and Mudde 2005). Moreover, despite the rise of new left issues, RLPs still distinguish themselves from other parties by strongly emphasizing traditional left issues (Fagerholm 2017). As RLPs defend the economic interests of lower socio-economic strata, especially the dissatisfied working class would vote for them.

Looking at what kind of grievances may explain the preference among lower classes to vote for RLPs, it has been shown that dissatisfaction with the economy, democracy and income inequality are all positively associated with electoral support for RLPs (Rooduijn et al.2017; Hansen and Olsen 2021). Furthermore, Oesch and Rennwald (2018) showed that production workers vote for the left out of economic dissatisfaction (measured by attitudes towards income inequality). However, this study did not distinguish radical left voting from green or social democratic voting and did not examine political grievances. Langsæther (2019) demonstrated that economic left-right values account for a sizeable share of class differences in support for RLPs, but again, political dissatisfaction was not part of the study. With the current study's focus on the role of both economic and political grievances to understand the class cleavage in radical left voting, we make a first contribution to the literature.

Another aspect of class voting for RLPs that has received limited attention is that class voting may depend on macroeconomic conditions. Examining political ideology instead of social class, Gomez and Ramiro (2019) found that higher unemployment increases support for RLPs regardless of ideological radicalism, rather than among those with a radical left ideology in particular. Furthermore, Rooduijn and Burgoon (2018) argue that poor macroeconomic conditions can either deepen or dampen radical left voting among people who experience economic difficulties. Even though they did not find support for any such moderation, their theoretical approach is relevant to contextualize the role of structural position in radical left voting in Europe. RLPs may appeal even more to lower classes when their structurally disadvantaged socioeconomic position is exposed under poor economic conditions, for example because they are more likely to become unemployed and are thus more threatened by economic downturns than higher classes. It is an open empirical question whether poor macroeconomic performance does matter as a moderator of social class, as opposed to a subjective assessment of economic well-being (cf. Rooduijn and Burgoon 2018). The second contribution of this study is, therefore, to examine if differences between people in lower and higher social classes in RLP support are conditional on macroeconomic circumstances. In doing so, we focus on Rooduijn and Burgoon's deepening hypothesis rather than their dampening arguments (i.e. relative deprivation and risk aversion), because we expect people in lower social classes to benefit from good macroeconomic performance in terms of economic well-being.

In sum, this study addresses the following research questions:

1. To what extent do social classes differ in their likelihood to vote for RLPs?

2. To what extent is the gap between social classes in radical left voting mediated by feelings of dissatisfaction with (a) the economy, (b) democracy and (c) income inequality?

3. To what extent is the gap between social classes in radical left voting moderated by macroeconomic performance, that is, (a) wealth, (b) unemployment and (c) income inequality?

To answer these research questions, we apply multilevel analysis to cross-national data from the European Social Survey (ESS), collected biannually between 2002 and 2018. This enables us to study class voting for RLPs in different economic climates across a large(r) number of European countries and years. Moreover, it provides us the opportunity to take into account not just cross-sectional differences between countries, but also changes over time within countries (Fairbrother 2014).

RLPs are defined by their both radical and left-wing position. They are regarded radical in the sense that they reject the current economic system of capitalism and advocate an alternative economic and power structure, but with democratic means. Furthermore, they are left-wing as they conceive of economic inequality as the largest issue in contemporary society that they wish to solve through major income redistribution (March and Mudde 2005; Dunphy and Bale 2011). In line with studies on the classification of these parties (e.g. March 2011; Gomez et al.2016), RLPs position themselves left of social democratic and green parties on the political spectrum. Finally, many European RLPs are considered populist (Rooduijn and Akkerman 2017), but we do not consider this a defining feature of the radical left party family.

Class inequality and the accompanying conflict between classes have been the topic of political and sociological studies since its early days. The Marxist definition of social class, rooted in historical materialism, differentiates between those who own the means of production (capitalists) and those who use the means of production to produce without owning them (the working class). This has been adopted in the class voting literature as the distinction between manual and non-manual workers, even though the latter group is too broadly defined to be the electorate that owns the means of production (Nieuwbeerta 1996). Recent class voting literature (Jansen et al.2013; Evans 2017; Rooduijn et al.2017) situated in this tradition has used the more elaborate EGP schema to differentiate between classes (Erikson et al.1979).

A widely used alternative way to operationalize social class is the Oesch schema, which – besides the vertical or economic divide – includes a horizontal or cultural dimension that puts emphasis on differences in marketable skills and work logics (Oesch 2006). This class schema displays similar explanatory power as the EGP schema in the study of voting (Langsæther 2019), but performs better when predicting radical right voting (Kitschelt 2013; Langsæther 2019). Because of the strong emphasis of the EGP schema on the economic divide between classes, we consider it still relevant in predicting partisan choice (Evans and De Graaf 2013; Langsæther 2019), as did recent class voting studies (Jansen et al.2013; Evans 2017; Rooduijn et al.2017).

Social class is a source of differing political interests and diverging voting behaviour. The class cleavage strongly determines political values and party preferences (Lipset and Rokkan 1967; Rennwald 2020; Emanuele 2021). The tendency of people from the same class to more strongly prefer a party family compared to people from other classes is referred to as class voting (Evans 2017). Class voting can be the result of differences in cultural values (Van der Waal et al.2007; Langsæther 2019; Rennwald 2020), but is traditionally thought of as the result of differing economic values, such as redistribution preferences (Evans 2017; Langsæther 2019). According to the latter view, classes have different economic values, which leads to class voting because people are inclined to vote for parties that align with their economic values (Evans 2017; Langsæther 2019).

It can be expected that people in lower social classes are overrepresented among the radical left electorate because RLPs claim to defend the economic interests of the working class and share their preferences regarding state influence on the economy (Langsæther 2019). Supporting this class voting assumption, individuals in lower classes have indeed been found to be more likely to vote for RLPs (Bowyer and Vail 2011; Gomez et al.2016; Ramiro 2016; Rooduijn et al.2017). The current study retests this relation, but does so for a large(r) number of European countries and a large(r) period of time. Both manual workers and lower routine non-manual employees are in this study classified as lower social classes because they share a disadvantaged position in the labour market that is addressed by the manifestos of RLPs (Sperber 2010). The main ideological message of RLPs thus resonates with these lower classes.

Members of the working class may not only vote for RLPs for economic reasons, but also out of resentment. People who are more dissatisfied with the economy, democracy and income inequality have a higher likelihood to vote for a RLP (Rooduijn et al.2017; Hansen and Olsen 2021) and it has been suggested that lower classes are more likely to support RLPs compared to higher classes because of these feelings of dissatisfaction (Oesch and Rennwald 2010). Put differently, lower social classes vote for RLPs precisely because their economic position causes them to be dissatisfied. Of course, this argument on economic and political grievances neatly aligns with the economically disadvantaged position lower classes are in. RLPs intend to improve the situation of the working class by radically changing the economic structure of society through major income redistribution (Fagerholm 2017).

Additionally, lower classes may feel that mainstream parties do not pay sufficient attention to their problems and become discontent with conventional politics (Sperber 2010; Ramiro 2016). Many RLPs address these concerns by expressing a populist message, blaming neo-liberal elites for the backlash against austerity and poor circumstances of the working class (Rooduijn and Akkerman 2017). Lower classes share their discontent with RLPs and may view these parties, as opposed to mainstream parties, as their best hope to improve their position (Rooduijn and Burgoon 2018).

This study stresses the importance of feelings of dissatisfaction with the economy, democracy and income inequality as explanations of class voting for the radical left. We formulate the following hypotheses:

Changing macroeconomic conditions are often assumed to explain changes in class voting over time. With increasing prosperity, the association between class and voting behaviour would become weaker (Clark et al.1993; Nieuwbeerta and Ultee 1999). Inglehart (1997) argues that modernization led to improved economic conditions and a higher standard of living, creating also more security among lower classes as they can make a decent living and are less likely to lose their livelihood. Because material concerns diminished in salience, the role of material interests in voting would have lost its significance too (Nieuwbeerta and Ultee 1999; Jansen et al.2013). Instead, partisan choice of the working class would be more strongly based on cultural issues (Van der Waal et al.2007). In economically secure times, lower classes are thus expected to be less inclined to vote for RLPs. After all, there is less need to vote for parties that defend their economic interests and aim to change the economic system.

In contrast, during economic downturns, members of the working class might experience financial difficulties or they may fear that they will lose their job, making RLPs more attractive. Poor macroeconomic performance is also known to fuel support for radical leftist views (Blekesaune 2007). Macroeconomic adversity exposes the already disadvantaged position of lower classes even more, which makes the message of RLPs more appealing. This line of reasoning is similar to the general argument underlying the deepening hypothesis as proposed by Rooduijn and Burgoon (2018), that the less well-off would experience even more economic insecurity if macroeconomic conditions are unfavourable. We formulate hypotheses about the moderating role of two indicators of macroeconomic performance, namely wealth (Visser et al.2014) and unemployment (March and Rommerskirchen 2015; Gomez and Ramiro 2019):

Goldthorpe and colleagues (1968) introduced another explanation for changes over time in the salience of class voting. They claimed that when macroeconomic conditions become better, the personal situation of in particular members of lower social classes improves relative to other classes, which means that the income and living standard of manual workers and lower routine non-manual employees start to overlap with that of higher, non-manual classes. As a result of this decreasing income inequality, lower classes have less need to use their political vote to defend their economic interests (Jansen et al.2013). Conversely, higher levels of income inequality may provide an incentive to lower classes to improve their relatively disadvantaged position (Nieuwbeerta and Ultee 1999). Lower classes may, therefore, be more motivated to vote for RLPs under these circumstances given that RLPs propagate major income redistribution to combat excessive income inequality. This leads us to formulate the following hypothesis: (H5) The more income inequality in a country, the larger the gap between individuals in lower and higher social classes in their likelihood to vote for a RLP.

We use data from nine rounds (2002–2018) of the European Social Survey (ESS) to test our hypotheses.¹ Data were collected every two years in a changing set of European countries. Face-to-face interviews were conducted with individuals of 15 years and older. Respondents were sampled differently across countries, ranging from fully randomly drawn samples to stratified four-stage samples. Yet, samples were designed in such a way that they are representative of the target population.²

Only those county-year combinations where respondents could vote for a RLP and where at least 25 respondents voted for the radical left were selected. The latter selection was necessary to obtain reliable estimates of the random coefficients for social class in the multilevel models. As respondents were asked which party they voted for in the last elections, people who did not vote were excluded, which includes people who were not eligible to vote due to their age or citizenship status. Respondents who voted for an ‘other party’ or ‘none of the above’ were also excluded from the sample because it could not be determined whether or not this ‘other party’ was radical left. As individuals become eligible to vote when they are 18 years in most European countries, we excluded people younger than 18. Finally, respondents with missing values on the predictor variables were excluded (3.8%). The final dataset consists of 134,665 individuals, 124 country-year combinations and 19 countries.

Respondents were asked whether they voted and if so: ‘Which party did you vote for in that [the last national] election?’. We dichotomized this variable, with 1 meaning that a respondent voted for a RLP and 0 that a person voted for another party. We defined RLPs based on our theoretical definition and classifications in previous studies (March 2011; Gomez et al.2016; Rooduijn et al.2017). First, the party needed to belong to the communist or socialist party family, which follows the party family approach (Mair and Mudde 1998). Additionally, the party needed to be located between 0 and 3 on the 0 to 10 left-right scale, as determined by party expert surveys, which follows March and Rommerskirchen (2015). Information on both indicators was derived from the Parliaments and Governments (ParlGov) database (Döring and Manow 2021). A few parties were not included in the ParlGov database and required additional research (e.g. the Communist Workers’ Party in Finland and New Anticapitalist Party in France). Some of them are a merge of two or more parties that were previously defined as RLP (e.g. the United Left Alliance in Ireland and Red Party in Norway). A few other parties were categorized as radical left because they belonged to The Left in the European Parliament (GUE/NGL), which is identified as a radical left group in the European Parliament. Some parties are a member of this group, but were not coded as RLP because these parties either belonged to another party family or were not located between 0 and 3 on the left-right scale. Table S1 in the online supplemental material gives an overview of all included countries and RLPs.

The ESS uses the International Standard Classification of Occupations (ISCO) to record respondent's current or previous occupation. Respondents were assigned ISCO88 scores in the first five rounds of the ESS; as of round six they were given ISCO08 scores. All ISCO scores were recoded to the EGP class schema (Erikson et al.1979).³ As discussed, we prefer this measure of social class as we are mainly interested in the economic divide between lower and higher social classes and because this class schema is widely used in the class voting literature.

The original measurement has eleven categories. We collapsed these into two categories for several reasons. First, we are theoretically interested in the differences between lower and higher social classes. Second, models with more than two random coefficients for social class categories did not converge due to the fact that in some country-year combinations there are null cells, meaning that no one in a certain class voted for the radical left. We defined lower routine non-manual employees, manual supervisors, (un)skilled manual workers and farm labourers as lower classes. Other classes were defined as higher classes as they are in a more advantaged position.⁴

We excluded people who did not have an ISCO score. We could have added a separate category for unemployed and/or non-employed individuals, but again, for both theoretical (our hypotheses are on the gap between lower and higher classes) and empirical reasons (model convergence issues prevented us from including a random coefficient for that category) we decided not to. Instead of using unemployment and/or non-employment status, which is only available at the moment the survey was administered, we chose to use as much information as possible with regard to social class, which is available for previous occupation in case of un- or non-employment.

Dissatisfaction with the economy and democracy were measured by asking respondents how satisfied they were with ‘the present state of the economy’ and ‘the way democracy works’. They could express their opinion on a scale from 0 (satisfied) to 10 (dissatisfied). Dissatisfaction with income inequality was measured with the following statement: ‘the government should take measures to reduce differences in income levels’. Answer categories ranged from 0 (disagree strongly) to 4 (agree strongly). Although the item does not measure dissatisfaction with income inequality directly, we assume that respondents are more dissatisfied with income inequality if they want the government to reduce it.

Data on wealth and unemployment for all country-year combinations were derived from the website of the World Bank. Wealth was measured by Gross Domestic Product (GDP) per capita in current US dollars. We divided this measurement by 1,000 to get a meaningful interpretation of the regression coefficients. Unemployment was measured by the percentage of the labour force that is unemployed. Finally, income inequality was measured by the Gini coefficient in equivalized household disposable (post-tax, post-transfer) income. These figures were obtained from the Standardized World Income Inequality Database (SWIID). The higher the coefficient, the more income inequality.

We control for several variables that have been established in previous research as predictors of radical left voting (e.g. Gomez et al.2016; Ramiro 2016; Rooduijn et al.2017). Educational attainment refers to the highest level of completed education, based on the International Standard Classification of Education (ISCED) and condensed into primary, secondary and tertiary educated. Religious belonging was measured with the question ‘Do you consider yourself as belonging to any particular religion or denomination?’ (0 = no and 1 = yes). We also control for sex, age (divided by 10 for ease of interpretation) and whether people were born in the country of residence. Finally, we included a dummy variable indicating whether the country of residence is part of Western (0) or Eastern Europe (1). Table S2 in the online supplemental material presents descriptive statistics for all variables.

To test our hypotheses, we performed three-level logistic regression analysis with the ‘lme4’ package in R (glmer function and maximum likelihood Laplace approximation as estimation procedure), with individuals nested within country-year combinations nested within countries. We included between- and within-country variables for all indicators of macroeconomic performance to distinguish effects of the level of and changes in these indicators (see, for example, Fairbrother 2014).⁵ Interval variables were centred at their mean. The intraclass correlation is 0.299/(3.290 + 0.299) = 0.083 for the country level and 0.092/(3.290 + 0.092) = 0.027 for the country-year level. This implies that 8.3% and 2.7% of the total variance in radical left voting can be attributed to differences between countries and country-year combinations, respectively. Both are statistically significant based on likelihood ratio tests.

Model 1 in Table 1 includes all but the mediating variables. It shows that lower social classes are more likely to vote for RLPs versus voting for another party than higher social classes, controlled for the individual and contextual characteristics. Their odds to support a RLP are 1.4 (e^0.324) times greater and their likelihood to vote for a RLP is 2.9 percentage points higher. Given that 10.4% of the voters in our sample support the radical left, we can speak of a strong effect size. The results are in line with our first hypothesis.

We also observe that the likelihood to vote for the radical left is higher when unemployment increases and when the level of income inequality is higher. Furthermore, all else equal, men, the lower educated and religious people are less likely to cast their vote for a RLP than for another party.

To test hypotheses 2a to 2c, Model 2 includes all variables from Model 1 and introduces economic and political grievances as mediating variables. The results first show that economic and political grievances are positively related to radical left voting. Individuals who are more dissatisfied with the economy, democracy or income inequality are more likely to vote for a RLP. The results also show that the gap between people in lower and higher classes in their likelihood to support RLPs has decreased considerably as a result of including the dissatisfaction measures. Although the gap is still significant, the coefficient of lower classes is much smaller compared to Model 1 (41.4% of the AME is mediated). This indicates that the class cleavage in radical left voting is partly attributable to economic and political grievances.

Formal mediation tests are not straightforward and still in development for multilevel logistic regression models. We, therefore, used the KHB method in an individual-level analysis that included dummy variables for the country-year combinations (Breen et al.2021). We deem this acceptable as the mediation takes place solely at the individual level and because we do not have expectations on cross-national variation in the mediation. The results of this analysis showed that all dissatisfaction variables significantly mediate the effect of social class on radical left voting, with income inequality being the most important mediating variable (29.8% explained), followed by dissatisfaction with democracy (7.1%) and the economy (5.2%).⁶ These findings support hypotheses 2a, 2b and 2c.

Interestingly, and again comparing models 1 and 2, it seems that economic and political grievances are able to fully explain the effects of changes in unemployment and the level of income inequality. If unemployment increases or the level of income inequality is higher, people feel more dissatisfied with the economy, democracy and income inequality, which in turn makes them more likely to support the radical left. Note also that adjusting for economic and political grievances leads to a significant effect of the level of unemployment in a country (i.e. a suppression effect), although it was already almost significant in model 1 (p = 0.054). Surprisingly, the higher the share of unemployed people, the lower the support for RLPs.

In the next step, we estimated a model in which we introduced a random coefficient for social class at both the country and country-year level, but in which we left out economic and political grievances because we are interested in the total effect of social class. There was significant variation in the class gap in radical left voting across countries and country-years (results not shown here), which we aimed to explain by wealth, unemployment and income inequality. However, none of the cross-level interaction terms between social class and the indicators of macroeconomic performance reached statistical significance, as shown in models 3a to 3c in Table 2. This means that hypotheses 3 to 5 are not supported.

We decided to continue with testing whether our explanations of class voting for the radical left are conditional on macroeconomic performance. Even though we did not formulate hypotheses on this, it is theoretically plausible that, instead of the relation between social class and radical left voting, the effects of economic and political grievances are moderated by macroeconomic performance. We argue that in contexts of poor macroeconomic performance, economic issues are likely more salient and, in these circumstances, economic grievances could be the deciding factor in voting for the radical left. Put differently, the effect of dissatisfaction with the economy is reinforced. In contrast, economic issues and thus also the role of economic grievances in voting decisions might be of less importance when the economy is booming. Under these favourable conditions, political grievances could be more consequential for radical left voting behaviour, for instance because income inequality is then seen as even more unjustified. As the results in Table 1 have shown that lower classes are more likely to vote for RLPs out of economic and political dissatisfaction, these opposing forces could even explain why we did not find moderation of class voting for the radical left.

Figure 3 shows that increases in wealth (GDP per capita) strengthen the positive effects of dissatisfaction with the economy and income inequality on the likelihood to vote for a RLP. We also observe that a higher level of wealth strengthens the positive effect of dissatisfaction with income inequality. This holds by and large across the full range of GDP per capita, with some exceptions at the lowest values in the upper left panel. Next, neither the level of unemployment nor changes therein have a moderating influence. Finally, the results show that a higher level of income inequality weakens the individual-level effect of dissatisfaction with income inequality, whereas increasing income inequality strengthens this effect as well as the effect of dissatisfaction with the economy. Figure 3 again demonstrates that the AMEs are significant across almost the full range of the Gini coefficients.

We performed several sensitivity analyses to check the robustness of the results with regard to the hypotheses and the additional cross-level interaction models. First, we ran the models without Cyprus (see Tables S6–S10) because it has the highest share of radical left voters and is, therefore, a potential influential case, and without Russia (Tables S11–S15) because it is a special case in terms of communist tradition and the political situation. The results on the hypotheses did not change. Overall, the results of the additional analysis also did not change. There are, however, two extra findings when excluding Cyprus: the higher the level of wealth in a country, the weaker the effects of dissatisfaction with the economy and democracy (see Models 4a and 4b in Table S8). All in all, this robustness check strengthened our confidence in the main results.

Second, voting behaviour refers to the last general elections, which means that all predictor variables are in principle measured ahead in time of the outcome variable. Although we examine associations, we analysed an alternative dependent variable, namely to which political party people feel closer to than all the other parties, if any. The correlation between radical left voting and feeling closer to a RLP is 0.842 (p < 0.001), indicating high overlap. The estimates of this sensitivity analysis can be found in Tables S16–S20. The results show high resemblance to the main findings, but there is a noteworthy difference as well. Hypothesis 5 is supported in the robustness analysis, yet only for changes in income inequality within countries over time. If income inequality increases, the gap between lower and higher social classes in their likelihood to feel closer to a RLP is larger. It is difficult to determine whether this divergence is substantively meaningful, a result of analysing a different sample or having a different causal order. All other results regarding the hypotheses are equal to the main results. Our interpretation of this is that the associations we have found are by and large robust and that analysing voting for a RLP, even though it is about the last general elections, does not lead to biased conclusions.

RLPs have consolidated their position in many European countries in the past decades. As previous research showed, their electorate predominantly consists of individuals in lower social classes whose economic interests RLPs claim to defend (Bowyer and Vail 2011; Gomez et al.2016; Ramiro 2016). It has been suggested that lower classes also vote for RLPs because they are dissatisfied with the economy and politics (Oesch and Rennwald 2010; Sperber 2010). The first aim of this study was to test whether economic and political grievances are able to explain class voting for the radical left. Furthermore, during and after the economic crisis of 2008, RLPs became electorally more successful (March and Rommerskirchen 2015; Gomez and Ramiro 2019), indicating that macroeconomic conditions may influence the decision to vote for a RLP. However, little is known about the class cleavage in radical left voting in relation to macroeconomic performance. The second aim of this study was, therefore, to examine to what extent the class gap in RLP support depends on the level of and changes over time in wealth, unemployment and income inequality. To fulfil these aims, we applied three-level regression analysis to high-quality data from the ESS, covering a large number of countries and spanning over 16 years.

Our first conclusion is that class voting remains relevant. Replicating the findings of previous research (Bowyer and Vail 2011; Gomez et al.2016; Ramiro 2016), we showed that people in lower social classes are more likely to vote for RLPs than their counterparts in higher social classes. Moreover, and an empirical contribution to understanding class voting for the radical left, they do so because they are dissatisfied with the economy, democracy and income inequality. Dissatisfaction with income inequality turned out being the most important reason why lower-class individuals support RLPs. An ideological match is thus made, as RLPs and their voter bases both want income differences to be reduced (Langsæther 2019). Economic and political grievances were, however, not able to fully explain why lower classes are more likely to vote for a RLP compared to higher classes. Part of the appeal of RLPs for lower classes must, therefore, be because of other factors than the economic and political aspects studied here. For example, cultural views, relating to international solidarity but at the same time a sceptical attitude towards international organizations such as the EU and the NATO, have been linked to radical left voting (Ramiro 2016; Rooduijn et al.2017; Wagner 2021) and could also play a role in the decision of lower classes to vote for a RLP. Including these factors was beyond the scope of this study, but future research may want to test such explanations of class voting for the radical left. It also seems promising to combine economic and cultural explanations of radical voting behaviour by examining feelings of social marginalization (Gidron and Hall 2020).

Our second conclusion is that economic and political grievances do not only partly explain class voting for RLPs, but also fully explain the contextual-level effects of unemployment and income inequality. Although we did not propose any hypotheses on the direct effects of the macroeconomic indicators, we found that people are more likely to support RLPs if unemployment increases or if there is more income inequality, which is in line with prior studies (March and Rommerskirchen 2015; Gomez and Ramiro 2019). We provided a novel empirical finding by showing that a rise in unemployment and a higher level of income inequality fuel economic and political grievances and, in turn, support for the radical left.

Our third conclusion is that poor macroeconomic performance is not conducive to class voting for the radical left. None of the indicators of macroeconomic performance, that is, (changes in) wealth, unemployment and income inequality, moderated the class gap in radical left voting. Hence, poor macroeconomic performance does not appear to stress the (material) concerns of people in lower social classes even more, at least not in such a way that it increases their already higher likelihood to support a RLP. Our findings are in agreement with those of Rooduijn and Burgoon (2018). They found that people who find it difficult to cope on their present income are more likely to cast their vote for a RLP. Reaching a similar conclusion as we do, this effect of economic well-being was not influenced by wealth, unemployment and income inequality. Yet, individuals who experience economic difficulties were found to be more likely to vote for a progressive RLP (compared to a mainstream party) if the level of immigration is low (ibid.). This suggests that class voting for the radical left may also depend on cultural country characteristics rather than economic ones. Future studies could – next to cultural factors at the micro level as mediators – focus on cultural factors at the macro level as moderators.

Our fourth conclusion, arriving at new insights while performing the analyses, is that macroeconomic performance does moderate the associations between economic and political grievances and voting for the radical left, yet mostly not as one would initially expect. Unexpectedly at first sight, voting for RLPs as a result of dissatisfaction with the economy and income inequality is more pronounced when a country's wealth increases over time. Similarly, RLPs receive more electoral support because of dissatisfaction with income inequality in more affluent countries and in countries where there is less income inequality. It could be that a relative deprivation mechanism is at work here (see Rooduijn and Burgoon 2018). If the economy is performing well, people might be inclined to vote for a RLP because they feel that they do not benefit from these favourable circumstances. This fuels their economic and political grievances and, subsequently, makes RLPs more attractive. Connecting the results from the mediation analysis to this additional moderation analysis indirectly informs us about class voting. We found that lower social classes are more dissatisfied with the economy, democracy and income inequality and, therefore, more likely to vote for radical left parties than higher social classes. If, then, economic and political grievances are moderated by macroeconomic performance, this should happen more often among lower social classes.

However, we also found that increasing income inequality over time, as opposed to the absolute level of income inequality, strengthens the effects of dissatisfaction with the economy and income inequality on the likelihood to support a RLP. The question is to what extent people are actually aware of widening income inequality in the short term, but our results suggest that such increasing income differences do matter. We would have not picked up on these diverging moderating effects if we did not distinguish within- and between-country effects (Fairbrother 2014). Whenever feasible, it would be worthwhile if future studies do so as well.

A final suggestion for future research to study the role of context in class voting, is to use a more elaborate measurement of social class, consisting of more than two categories. Additional analyses showed that, as expected, the findings at the individual level using the seven-class version of the EGP schema are similar to the main findings, but a limitation of our study is that for reasons of model convergence we were not able to examine the moderating role of macroeconomic performance for this wider range of social classes. Although we theoretically do not expect major differences, it is an open empirical question whether this is the case. Future research should, therefore, aim to answer this question.

All in all, this study has contributed to the literature on radical left voting behaviour by providing new insights into the reasons why lower classes vote for RLPs and under which circumstances. It has shown that social class remains relevant in explaining RLP support. RLPs throughout Europe are in part electorally successful because they manage to attract dissatisfied working-class people, indicating that the economic interests of lower classes still align with the ideological backbone of the radical left. What is more, RLPs predominantly benefit from economic and political grievances of the electorate in times of economic prosperity, pointing towards a prominent role of feelings of relative deprivation in mobilizing voters to support the radical left.

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

Mark Visser is an Assistant Professor in the Department of Sociology at Radboud University, Nijmegen, the Netherlands. His major research interests include older workers and retirement, social inequality, voting behaviour and the welfare state. He studies these topics in longitudinal and comparative perspective and has published on these subjects widely in international journals.

Marcel Lubbers is Professor of Interdisciplinary Social Science: Relations between groups and cultures at Utrecht University/ICS and leading ERCOMER (European Research Centre on Migration and Ethnic Relations).

Twan Karremans finished his Research Master 'Social and Cultural Science'. His main interests include quantitative research methods and topics as social and economic integration of migrants. Currently, he is a Data analyst at Market Research Agency MarketResponse in Utrecht, The Netherlands.

Marlou Ramaekers is a PhD candidate at the Department of Sociology at Radboud University Nijmegen. Her research focuses on direct person-to-person helping and the impact of family life and neighbors on the provision of help. Other research interests of hers include voting behavior and social identity.