Abstract
There has been a dramatic rise in voting for populist parties in Europe over the past thirty years. We assess the role of government labor market policy in dampening or provoking populist sentiment. Drawing from a panel of 134 elections from 1990 to 2021 and pooled cross-sectional data from eleven waves of the European Social Survey, we find evidence that populist parties fared worse where countries provided more robust income support to workers experiencing unemployment. The effect was stronger among those individuals who had experienced unemployment and among current and former trade union members. This suggests that the welfare cuts and labor-market reforms pursued since the early 2000's may have alienated vulnerable segments of the population and driven them toward populist parties.
Keywords
Introduction
The past 25 years have seen a striking increase in support for European populist parties. In 1998, populist parties drew support from less than 10% of European citizens. Only two capitals on the continent—Bratislava and Bern—had populist politicians serving in government. In 2024, populist parties held 27% of national parliamentary seats in Europe and served in eight different governments; they were part of a confidence-and-supply agreement in one other (Heinö, 2024).
There are major differences among European populist movements of the Right and the Left. All of them, however, share an antagonism to existing mainstream political parties and political institutions. And all of them, in different ways, are skeptical of or hostile to central aspects of European integration. Some, especially on the Left, oppose austerity measures and fiscal constraints they see as imposed by European institutions. Others, especially on the Right, resent European policies toward refugees, asylum-seekers, and immigration generally. Still others are more broadly concerned that the European Union and the international trading system have eroded too much of their nations’ sovereignty or accelerated processes of deindustrialization that have devastated many communities.
Many of the material sources of the upsurge in populist sentiment in Europe and elsewhere have been well established. There is ample evidence for the impact of economic distress, both due to international economic trends and to automation (Anelli et al., 2019; Colantone and Stanig, 2018; Rodrik, 2018). Certainly, there are important cultural, ethnic, and traditionalist bases of populist sentiment and rhetoric (Norris and Inglehart, 2019), and there is clear evidence of both material and nonmaterial sources of populism (Frieden, 2022). Our focus is on a different economically based source of populist sentiment: national government welfare policies. We argue that social policies, particularly spending on programs that maintain incomes for working-age adults who are fully or partially unemployed, have had an impact on the appeal of populism.
In this article, then, we follow up on findings that economic distress tends to stimulate populism. We ask two related questions. First, does the existence of an ample unemployment insurance system that softens the impact of negative economic trends reduce political discontent? More precisely, do higher levels of spending on labor markets and higher unemployment insurance replacement rates (UI RR) reduce the discontent that leads to populist voting? Second, do cuts to unemployment insurance programs stimulate this discontent? In particular, have reductions in UI generosity as a result of labor market reform or austerity increased the likelihood that affected parts of the population will support populist parties?
We argue for two distinct channels by which government social policies, and specifically labor market programs, have affected the strength and nature of populist sentiment. The first channel operates over the long term: countries that evolved a broader and deeper social safety net for working adults facing actual or potential economic disruption have experienced less of a populist backlash than those who have not. This suggests that some form of the “compensation hypothesis”—that compensating those harmed by economic changes, or most vulnerable to harm, can mitigate the socio-political impact of those changes—may be correct.
On this dimension, we find evidence that higher expenditures on labor market programs predict lower populist vote shares, controlling for other factors. From a panel analysis of 134 national legislative election results, we show that where governments provide more generous support to the unemployed, and more funding to help individuals return to work, populist parties are less successful electorally. This relationship is most robust for populist parties but also observed in some models when we separately examine radical right-wing and radical left-wing parties, respectively. Our analysis of pooled survey data from the European Social Survey also indicates that more generous unemployment protection may moderate support for populist parties. We find that higher levels of social expenditures on labor markets, and more generous UI replacement rates, predict a lower likelihood that a respondent will have supported a populist party in the previous election.
The second channel connecting government policy to populism is more recent: countries whose governments have significantly reduced spending for out-of-work individuals, whether due to labor market reform or austerity measures, have experienced a greater backlash against political and economic integration. Both the labor market reforms pursued in many northern European countries in the 2000s, and the austerity measures imposed on southern European countries in the 2010s have reduced the generosity of unemployment insurance that replaces the income of workers facing short and long-term employment disruption. These reforms may have had a particularly negative impact on precisely those segments of the population that were already experiencing significant economic insecurity related to globalization, deindustrialization, and technological change (Autor et al., 2003).
We find that cuts to unemployment benefits are associated with greater support for populist parties in general and radical left-wing parties in particular. All other things equal, a 5 percentage point increase in the unemployment insurance replacement rate from its mean reduces the predicted probability of an individual voting for a populist party from 8% to 5.8%, and for far left parties from 4% to 3.2%. This effect is more pronounced for individuals who have experienced three or more months of unemployment and among current and former labor union members.
Our theory is tested against a rival hypothesis suggesting that the effect of social protection on support for populist parties may depend on a country's level of immigration and individuals’ immigration preferences (Burgoon and Schakel, 2022; Vadlamannati, 2020). Consistent with prior research, we find that opposition to immigration increases support for populism and that general welfare state spending does not mitigate support for populism. However, our key finding, that higher labor market spending is associated with lower support for populist parties, is not influenced by immigration levels or individual attitudes toward immigration. Individuals with negative attitudes toward immigration also express less support for populist parties when unemployment insurance is more generous. This finding aligns with evidence that well-designed and targeted welfare policies can dampen support for populist parties among those facing socioeconomic risk (Vlandas and Halikiopoulou, 2022).
The correlational nature of our analysis does not allow us to make strong causal claims about the relationship between unemployment insurance programs and support for populist parties. The relevant policies in this domain are set at the national level and are therefore endogenous to a wide variety of other socio-economic and political features of the countries in question. While establishing a causal effect for such policies is admittedly challenging, this does not diminish their importance as subjects of empirical investigation. Our findings highlight a strong and consistent empirical relationship between unemployment insurance spending and the appeal of populism to national populations. The results suggest that welfare expenditures at the national level plausibly shape public support for populist parties. These conclusions are robust across a range of alternative specifications and are consistent with recent studies that use granular data and spending discontinuities to demonstrate causal relationships between spending cuts and support for populist parties (Colantone and Stanig, 2018; Dickson et al., 2024; Fetzer, 2019).
Populism can and does thrive in a variety of contexts and its development and durability is shaped by a wide array of factors (Gidron and Hall, 2020; Norris and Inglehart, 2019; Rodrik, 2018). We recognize that long-term economic, social, and cultural changes are the underlying forces behind growing support for populist parties. However, we contend that the structure and organization of the social safety net and its trajectory of change is one overlooked factor that likely mediates people's experience of these developments. By lessening the effects of economic and cultural change on livelihoods, compensation may reduce the extent of grievances and limit the appeal of populist political parties. While labor market and social policy reforms may have been justified, their uneven distributional impact has had politically important and in some cases explosive effects.
Theory: Compensation and populism, austerity and populism
The countries of Western Europe have undergone substantial socio-economic changes over the past fifty years, in particular the shrinkage of labor-intensive manufacturing. The decline of low- and middle-skilled, high-paid jobs in industry has in turn been connected to the rise in populist sentiment; this can be seen especially with the geographical concentration of support for populism in declining industrial regions (Broz et al., 2021). These economic trends are largely the result of economic integration and technological progress, which suggests important theoretical questions of both a positive and normative nature. Because economic integration and technological progress, like most economic developments that create aggregate welfare gains, produce losers as well as winners, they can lead to political conflict. Indeed, in the political arena the distributional effects may outweigh the welfare effects, especially if the concerns of real or expected losers are more intense than those of winners, and if the losers are well-organized and well represented in the political system.
The political feasibility of welfare-improving policies with substantial distributional effects may, as a result, depend on using some of the welfare gains to compensate losers. One strand of the literature that addresses the issue focuses on what might make economic integration politically feasible in democratic political systems. Scholars have noted that more open economies tend to have larger governments, and have surmised that this is due to the greater need to compensate those threatened by the vagaries of the world economy (Cameron, 1978; Rodrik, 1998). In an influential series of country studies and a summary volume, Katzenstein (1985) examined the small open economies of Western Europe. He showed that they were largely forced by the fact that their small size made economic openness a necessity to devise comprehensive social safety nets to protect their citizens from the potential harms that openness might bring.
This “compensation hypothesis” should apply more broadly to any disruptive socio-economic development. Here, we use it to attempt to explain the impact of a social safety net on the political response to both specific trade shocks and the broader process of deindustrialization and automation that has reduced the availability of high-paying manufacturing jobs and increased the insecurity and risk faced by routine workers (Kurer, 2020; Rehm, 2009).
The basic proposition is simple: policies that insure against income loss and protect workers and communities from instability can mitigate a potential political backlash against adverse trends. When economic changes are the cause of discontent, such policies can be seen as compensating the losers for their losses. Most directly, they replace income for the individuals who are experiencing economic dislocation and help them find alternative employment. More indirectly, they provide assurance to workers whose livelihoods are threatened by globalization and technological change. Labor market spending thus can reduce support for populist political parties that exploit economic (and cultural) grievances.
We do not suggest that countries with relatively generous welfare states will not still see an increase in support for populist parties generated by economic and cultural change. As noted earlier, populist politics stems from structural, long-term trends that affect nearly all advanced industrial democracies. However, we do suggest that countries with more extensive social support will see comparatively lower support for populist parties, all other things equal. Our proposition that labor market policy can affect the level of support for populist parties is in line with the broad policy feedback literature that has demonstrated how policies shape patterns of political participation (Campbell, 2011; Larsen, 2019). Despite the importance of the issue, there have been only a few attempts to evaluate whether this expectation has been borne out over the past 20 years. The careful empirical studies that have examined the question have produced small or inconsistent results (Gingrich, 2019; Halikiopoulou and Vlandas, 2016; Rickard, 2023). 1
While the logic of compensation suggests that countries with well-developed social policies to assist citizens facing economic difficulties, an alternative view argues that the compensation dynamics observed for much of the second half of the twentieth century may no longer apply to all European contexts (Burgoon and Schakel, 2022; Rathgeb and Busemeyer, 2022). In countries where immigration levels are high and social welfare is perceived as disproportionately benefiting immigrants, a large social welfare state may reinforce rather than mitigate the anti-globalization nationalism that fuels support for populist parties (Burgoon and Rooduijn, 2021; Burgoon and Schakel, 2022; Cavaille and Ferwerda, 2023; Vadlamannati, 2020). As discussed below, we have theoretical reasons to expect labor market spending will not be perceived as disproportionately benefiting immigrants. However, we still consider and test the possibility that citizen electoral responses to unemployment spending could be shaped by both immigration levels and immigration views.
Labor market spending and cross-national variation
To assess the relationship between compensation and support for populist parties, we examine spending on unemployment insurance and labor markets more broadly. We choose to focus on labor market spending rather than other kinds of social spending for several reasons. First and foremost, unemployment insurance programs provide the most direct and immediate form of assistance to those facing job and income loss generated by globalization and deindustrialization (Burgoon, 2001: 521). For this reason, labor market spending is often viewed as following a compensatory rather than a protective logic (Vlandas and Halikiopoulou, 2022). Social investments in education, childcare, and housing may be an important tool at the societal level to deal with the impact of economic change, but in many cases they are not targeted at the individuals and households who have faced job loss. This sort of spending may in fact be irrelevant for middle-aged and older workers threatened by economic uncertainty.
Second, traditional contributory income maintenance programs such as unemployment insurance and pensions are consistently prioritized by the less educated and less skilled individuals and groups most adversely affected by economic integration and technological change and most likely to support populist parties (Garritzmann et al., 2018: 844). Survey research suggests that populist voters are the most likely group to support increased spending on traditional cash transfer programs such as unemployment insurance and the least likely to support new investments in education, childcare, and workfare programs (Garritzmann et al., 2018; Häusermann, 2018; Häusermann et al., 2020).
Figure 1 presents the average unemployment insurance replacement rates for two months and 24 months of unemployment across 16 Western European countries from 1990 to 2024. Replacement rates vary substantially across countries, ranging from as high as 90% in Luxembourg to below 50% in countries such as Ireland and Greece. This indicates that EU citizens experience markedly different levels of protection in the event of job loss. As the dislocating effects of economic globalization and deindustrialization have become more pronounced, these cross-national differences in the quality of the social safety net are likely to shape the political response to economic change.
Unemployment replacement rates in European countries, 1990–2024.
Figure 1 also points to changes in the generosity of unemployment insurance programs over time. The average trend has been toward less generosity, particularly for the long-term unemployed. In 2001, a childless single adult living in the 16 countries analyzed in this study, and earning the national average wage, would have received around 47% of their previous earnings after 24 months of unemployment. By 2013, the average replacement rate was nearly a fifth lower, just 39% under the same conditions. While replacement rates have increased over the last decade in response to popular demands, they remained at an average of just 42% in 2024.
In many countries, unemployment insurance systems have also become more conditional, requiring workers to work for longer periods before being eligible, and to enroll in training or search for work as a condition for receiving benefits. These changes can be seen in overall labor market spending adjusted for the business cycle. Figure 2 details the trend in unemployment-adjusted labor market spending as a percentage of GDP across 16 Western European countries. We can see once again significant differences in the level of spending across countries. Overall, countries in northern Europe spent twice as much on average than countries in southern Europe.
Unemployment-adjusted labor market spending in European countries, 1990–2022.
We can also observe significant within-country change over time. Although these changes were not uniform, the overall trend has been toward lower levels of spending on labor markets. Between 1991 and 2015, unemployment-adjusted spending declined by roughly 40%, driven by both labor market reforms and austerity measures. Many northern European countries enacted spending cuts as part of labor market reforms designed to increase workforce participation (Bonoli, 2010; Hemerijck, 2015; Jenson, 2011). During the Eurozone crisis, some countries were also subject to structural adjustment programs that reduced social welfare spending, including unemployment insurance (Hermann, 2014). We expect such reductions to generate voter resentment and increase the appeal of populist parties (Campbell, 2011; Kurer et al., 2019; Pierson, 1996).
We thus have several theoretically grounded expectations which stem from or qualify the compensation hypothesis. Much like the original compensation hypothesis, we expect that countries that have evolved more substantial social safety nets for out-of-work adults will experience less of a populist upsurge, controlling for other factors. This is in essence about the impact of an established high safety-net political-economy equilibrium upon the rise of populist voting. However, we do not expect that all social welfare spending will dampen support for populism, given the diverse constituencies that different spending programs benefit. Rather, we expect labor market spending targeted at established, full-time workers to have the greatest impact.
Our second expectation about change over time provides a corollary to the compensation hypothesis. We expect that countries whose governments did, over the course of the past 30 years, undertake labor market reforms or austerity measures that limited cash transfers to those facing short- or long-term unemployment experienced a more significant increase in populist voting, while governments that maintained or expanded labor market programs experienced a less significant increase in populist voting. This expectation might explain why Scandinavian countries such as Sweden or Denmark that cut labor market expenditures could have experienced an increase in populist voting even as they maintained a comparatively generous welfare state.
Finally, as a check on our own assumptions, we consider the alternative hypothesis that welfare spending and populist voting may be positively correlated. The hard version of this argument is that more generous welfare states may be particularly prone to populist appeals given citizens’ concerns that their established systems may be undermined by European integration, globalization, or increased immigration (Greve, 2019; Kitschelt and McGann, 1997). The softer version of the argument is that the effect of compensation will depend on an individual's immigration views and the degree to which nonworking immigrants are perceived as using the welfare state (Burgoon and Rooduijn, 2021; Burgoon and Schakel, 2022).
Defining populism and measuring social expenditures
The word populism has been used to describe a wide range of social movements and political programs, but the term is now widely associated with a variety of political parties outside of the political mainstream (Bonikowski and Gidron, 2016; Mudde and Kaltwasser, 2017). Substantial heterogeneity notwithstanding (Vasilopoulou, 2017), all populist parties share a number of common characteristics. Nearly all populist parties emphasize an antagonism between citizens and elites, pitting “the people” against the elites (Mansbridge and Macedo, 2019). In Western Europe, populist parties of the Left and Right share two other common features. Nearly all are opposed to key aspects of European integration and employ elements of nationalism (Halikiopoulou et al., 2012; Vasilopoulou, 2018), and nearly all draw disproportionate support from the traditional working class, which has seen its relative position decline as a result of European integration and technical change (Bornschier and Kriesi, 2012; Gidron and Hall, 2020; Oesch, 2008).
We recognize, of course, that there are important differences in both the ideologies and policy programs of left- and right-wing populist parties as well as the educational levels of their social support base (Rooduijn et al., 2017; Visser et al., 2014). These differences notwithstanding, left and right populist parties both use populist discourse to articulate the interests of a common category of citizens negatively affected by economic globalization, deindustrialization and technological change (Rooduijn and Akkerman, 2017). From this vantage, votes for left-wing and right-wing populist parties can equally be viewed as expressions of discontent with processes of European economic and political integration that are widely viewed as benefiting elites at the expense of others (Kurer et al., 2019; Rodrik, 2018).
To categorize populist parties, we use the PopuList, an overview of populist parties developed by a consortium of political scientists (Rooduijn et al., 2023). This categorization substantially overlaps with separately generated lists of far right and far left parties. In this article, we focus primarily on analyzing the determinants of populist parties, while also reporting results for radical right-wing and left-wing parties. A full list of the political parties classified as populist and/or radical is provided in the Online Appendix.
To measure welfare spending, we rely on several indicators. First, we examine overall social spending as a percentage of GDP, using the OECD's Social Expenditures Database (OECD, 2025). This measure encompasses both cash payments to households as well as social services spending. This broad spending category has complex redistributive implications (or none) and may or may not be directly targeted at those facing economic shocks or dislocation (Burgoon, 2001; Busemeyer and Garritzmann, 2019). It therefore provides an important baseline for examining the effect of labor market spending.
To capture spending on labor markets we rely on a measure of overall labor market spending as a percentage of GDP. Two-thirds of this spending is for traditional unemployment insurance programs that provide those who lose their jobs with a certain percentage of their former salary. This is the spending that most directly and immediately aids workers facing economic shocks (Burgoon, 2001). Another third covers expenditures on counselling, subsidies to employers, job search assistance, and vocational training programs designed to facilitate or incentivize workforce participation (Clasen and Clegg, 2012; Jenson, 2011). Given that effective compensation systems have historically relied on both skills development (including retraining) and income maintenance to help workers adjust to changes in the global economy (Katzenstein, 1985), we examine “passive” and “active” labor market spending jointly. However, we present disaggregated analyses of these measures in the Online Appendix.
We supplement our analysis of labor market spending with an examination of unemployment replacement rates using data from the Social Insurance Entitlements Database (SIED) developed by the Swedish Institute for Social Research (Nelson et al., 2024). In the main text, we report results for the average unemployment benefit for a single worker after 26 weeks of unemployment, expressed as a percentage of the average production worker's wage. We additionally present results for families, as well as replacement rates adjusted for labor force coverage, in the Online Appendix.
Empirical analysis
We analyze two different datasets to evaluate whether compensation conditions populist support. First, we analyze election results from a panel of 16 countries between 1990 and 2021. We then examine the thesis further using eleven waves of the European Social Survey. In both analyses, we are interested primarily in two related analytical questions: (1) Do countries with comparatively high levels of compensatory spending on labor markets face lower levels of populist voting, controlling for other conditions? (2) Do changes in the generosity of labor market programs affect support for populist parties?
While it is impossible to fully disentangle the effects of cross-national differences from those of temporal change, we approximate this by estimating both multilevel models with random intercepts for country and fixed-effects models with country dummies. If the nature of the social-democratic welfare state limits the appeal of populist parties, we expect these parties to have lower levels of support in countries that spend more on compensation, all other things equal. If changes in social spending, especially to more economically precarious segments of the population, affect electoral support, we expect populist parties to be more successful in countries that have cut more from earlier levels and less successful in countries that have increased spending compared to expected baselines. Conversely, if compensation dynamics are no longer effective, due to increases in immigration or other factors, than we expect to observe a null relationship or even a positive relationship between welfare spending and populist voting.
Social welfare spending and populist vote shares, 1990–2021
As a first evaluation of the relationship between social welfare spending and populist voting, we run ordinary least squares (OLS) regressions using an unbalanced panel of 134 parliamentary elections held in 16 European countries from 1990 to 2021. Following previous literature, the focus is on Western European countries with established party systems and common social cleavages (Kriesi, 1998). 2 While we exclude postcommunist countries in Eastern Europe from the main analysis due to differences in social cleavages and welfare-state structures, we report results for an expanded sample of 24 European countries (including eight from Eastern Europe) in the Online Appendix. Many of the broad relationships between social spending and populist voting found in Western Europe can also be seen across this larger group of countries.
The dependent variable is the proportion of votes received by populist parties in national legislative elections. Our main explanatory variables of interest are the three welfare expenditure variables outlined earlier: (a) total social spending as a percentage of GDP; (b) public spending on labor markets as a percentage of GDP; (c) the unemployment insurance replacement rate for a single worker after 26 weeks of unemployment. By capturing different aspects of the generosity of the welfare state, these measures in combination allow us to make an initial assessment of whether populist voting is on average lower in countries where compensation systems are more robust.
We incorporate a range of economic controls. Because overall labor market spending reflects both short-term business cycles and structural labor market conditions, we control for each country's annual unemployment and labor force participation rates. Since macro-economic performance conditions support for populist parties (Rooduijn and Burgoon, 2018), we control for per capita income. Similarly, to account for the link between deindustrialization and populism, we add the share of industrial employment (Swank and Betz, 2003). National institutional quality is measured using Transparency International data, reflecting the tendency of populist parties to frame themselves as remedies to endemic corruption. Inflation, which has been shown to correlate with populist voting (Funke et al., 2023), is likewise accounted for, and immigration is measured using Eurostat data on immigration flows. All independent variables are standardized. Table 1 presents summary statistics for the variables used in the panel analysis.
Panel summary statistics.
To estimate the effects of changes in spending over time, some models include country fixed effects, thereby restricting the analysis to within-country variation and allowing us to assess how increases and decreases in spending relate to populist support. Given important distinctions between radical right-wing and left-wing parties, we additionally estimate each specification separately for these two groups, relying on the PopuList classification. To address potential serial autocorrelation, the models incorporate country-specific linear time trends. This approach allows each country to follow its own gradual trajectory over time, thereby absorbing unobserved, unit-specific dynamics such as gradual cultural change or persistent macroeconomic conditions that could otherwise bias coefficient estimates. We retain year fixed effects to capture common shocks across European countries and cluster standard errors by country.
We begin by examining scatterplots of labor market compensation and populist vote share. As shown in Figure 3, populist support is generally higher in countries with lower labor market spending and lower unemployment replacement rates. This pattern reflects both cross-national differences in spending and populist vote shares, as well as within-country changes. For example, as illustrated in the left panel, Germany's labor market spending declined sharply following the 2003–2005 Hartz reforms, which spurred the creation of the left-wing populist party Die Linke. Since then, German populist vote shares have surged from 4% in 2002 to 22% in 2017.
The relationship between labor market compensation and populist vote share.
In the right panel, we highlight the trajectory of Greece. Prior to the Eurozone crisis, the generosity of Greek unemployment insurance measured in earnings replacement rates was comparable to that of Scandinavian countries. Following the Troika-imposed structural adjustment programs, however, the system was transformed into one of the least generous in the EU. The sharp decline in unemployment insurance not only deepened the social costs of the crisis but also created fertile ground for political mobilization on the extremes, contributing to the rise of Syriza on the left and Golden Dawn on the right.
Table 2 presents results from the first stage of our analysis, which examines the relationship between equilibrium levels of social compensation and populist vote share. We observe a clear negative relationship between overall social spending and populism. Across two of our three welfare measures, and both of our labor market spending indicators, higher levels of compensation are associated with lower populist vote shares. Countries that devote more resources to supporting the unemployed and provide more generous replacement rates tend to have lower support for populist parties. At the same time most of our controls are not significant predictors. While higher per capita income predicts lower populist vote shares, inflation, industrial employment, institutional quality, and immigration rates are not systematically related.
Determinants of populist vote share.
Note: p-values in parentheses. *p < .05, **p < .01, ***p < .001.
Table 2 also reports estimates from models with two-way fixed effects, which account for all time-invariant country characteristics and common year shocks (Columns 2, 4, and 6). We continue to observe clear negative correlations between labor market compensation and populist vote share. Substantively, the estimates imply that a 5-percentage-point increase in a country's unemployment insurance replacement rate from its mean corresponds to a 20% reduction in predicted populist vote share, falling from 14.2% to 11.3%. Given that most countries have reduced benefit generosity over this period, particularly for long-term unemployment, these retrenchments may help explain the sustained rise of populist parties.
In Table 3, we present models with far-right and far-left vote shares as the dependent variables. Several macroeconomic covariates emerge as significant predictors. In some models, higher unemployment rates are linked to lower right-wing support but higher left-wing support, while stronger institutional quality corresponds to reduced right-wing vote shares.
Determinants of far right and far left vote share.
Note: p-values in parentheses. *p < .05, **p < .01, ***p < .001.
The compensation variables are less consistent predictors of far-right and far-left voting than they were of overall populist support. Overall social spending is not strongly correlated with either right- or left-wing vote shares. Labor market spending, by contrast, emerges as a more consistent predictor: higher levels of such spending are negatively correlated with radical left-wing vote shares, while higher unemployment insurance replacement rates are negatively correlated with radical right-wing vote shares. Taken together, these results suggest that more generous labor market policies may also dampen support for parties at both extremes of the political spectrum.
Welfare spending and populist voters: A multi-level analysis
While the panel analysis provides evidence that lower levels of compensation and decreases in spending over time are associated with higher populist vote shares, it does not allow us to control for individual-level characteristics that might affect support for populist parties. By constructing multi-level models that combine country-level statistics with individual-level survey data, we can more precisely identify which parts of the population are voting for populist parties and determine whether and how these groups’ political preferences are affected by the level and type of spending on labor markets and other welfare measures. This approach, common in political economy studies (Abou-Chadi and Wagner, 2019), makes possible an analysis of both between and within unit variation while still taking into account the nested structure of the data (Bell and Jones, 2015).
To analyze the determinants of populist voting, we estimate multilevel logistic regression models with random intercepts for countries and clustered standard errors, as well as OLS linear probability models with two-way fixed effects. We use eleven waves of the European Social Survey, a semi-annual survey of public attitudes conducted by the European Research Infrastructure Consortium, to assess the determinants of populist voting in 84 of the 95 national parliamentary elections that occurred in our 16 countries over the period 1999–2021. Our dependent variable is a binary indicator of whether a respondent reported voting for a populist party in the previous election. 3 Since the question asks respondents who they supported in the previous election, responses are coded for the relevant election year.
Previous research has found that unemployment and experiences of economic hardship increase the likelihood of supporting left-wing populist parties (Kurer, 2020) while individuals in routine occupations more vulnerable to globalization or technology-induced disruption are more likely to support right-wing populist parties (Colantone and Stanig, 2018; Oesch, 2008; Oesch and Rennwald, 2018). The ESS survey includes several questions that make it possible to assess whether individual economic circumstances shape support for populism. The first is a question that asks whether an individual has ever experienced three months or more of unemployment. We include a dummy variable in the regressions that indicates whether a respondent has this prior unemployment experience. The second is a question that asks an individual whether they are “living comfortably,” “coping,” “finding it difficult” or “finding it very difficult” on their current household income. We create dummy variables for each of these responses, excluding “living comfortably” from the regression analysis.
Previous research has found that immigration preferences condition support for redistribution (Burgoon, 2014) and shape how the public responds to changes in social welfare expenditures (Burgoon and Rooduijn, 2021). We therefore include in our model a measure capturing individuals’ views on whether immigration makes their country a better or worse place to live. We additionally include a dummy variable for trade union membership because trade unions have historically played a central role in the construction, defence, and, in some cases, administration of unemployment insurance systems in Europe (Clegg and Heins, 2024). Given their longstanding political mobilization around labor-market protection, union members constitute a constituency for whom changes in unemployment insurance may be especially salient.
Because the analysis now operates at the individual level, we control for key demographic characteristics such as gender and age. To allow for a nonlinear relationship between age and populist support, age is modeled using dummy indicators for six cohorts, with respondents under 30 serving as the reference category. The models also incorporate a standard set of individual-level covariates commonly used in studies of populism, capturing place of residence (urban, suburban, small town, village, or rural), educational attainment, and occupational characteristics.
To categorize educational attainment, we rely on an ESS question on schooling that has been harmonized into the International Standard of Classification of Education (ISCED) developed by the United Nations Educational, Scientific, and Cultural Organization (UNESCO). The ISCED classification divides educational attainment into five tiers, ranging from “less than lower secondary” to “higher tertiary education.” The largest category of education, lower secondary attainment (ISCED II), serves as the reference group. To categorize occupation, we use a question from the ESS that asks respondents to state their current or former occupation, which is subsequently classified into the 10-tiered International Standard Classification of Occupation (ISCO) developed by the International Labor Organization. In all models, the mid-skill category of clerk serves as the reference group. Additional details on these indicators are provided in the Online Appendix.
The multi-level models incorporate the five macroeconomic indicators used in earlier analyses: unemployment, labor market participation, inflation, GDP per capita, and the share of industrial employment. We also examine whether perceived institutional quality and measures of immigrant inflows condition the results. Standard errors are clustered at the country level, with year dummies included to capture common shocks. All specifications are weighted using poststratification weights to correct for sampling and nonresponse bias within countries, as well as population size weights to ensure that each country's contribution reflects its share of the total European population.
Table 4 reports the regression results. For each of the three dependent variables, we estimate both multi-level logit models with random country intercepts (RE) and OLS linear probability models with two-way fixed effects (FE). As expected, many of the control variables are statistically significant. At the country-year level, higher levels of industrial employment are associated with lower populist vote shares, consistent with previous findings that deindustrialization has fuelled populist support. At the individual level, men and current or former union members are more likely to vote for populist parties, as are individuals who hold negative views about immigrants. By contrast, many education, occupation, and domicile indicators are not statistically significant, perhaps reflecting differences in which social groups are drawn to right- versus left-wing populism.
Determinants of populist voting.
Note: p-values in parentheses. *p < .05, **p < .01, ***p < .001.
The results also provide support for our compensation hypothesis. While total social spending does not show a systematic relationship with populist voting, both measures of labor market programs do. Individuals are less likely to support populist parties in countries that allocate a larger share of resources to labor market programs and where unemployment insurance replaces a greater share of previous earnings. This pattern is consistent across both the random-effects logit models and the fixed-effects linear probability models.
Finally, those who have either experienced unemployment or view their household finances as precarious are consistently more likely to support populist parties. Across all specifications, individuals who have been unemployed for at least three months show a higher probability of voting populist. These findings align with existing literature that links job disruption and adverse economic circumstances to higher levels of populist support.
Table 5 reports results when we separately examine radical right-wing and radical left-wing voting. Disaggregating the dependent variable highlights important differences between the constituencies of right- and left-wing parties that have also been found in previous research (Rooduijn et al., 2017). Supporters of radical right-wing parties are more likely to be male, to live outside cities, and to lack tertiary education, whereas radical left-wing voters are more likely to reside in urban areas and to have completed university studies. Immigration preferences also diverge: right-wing voters hold negative views about immigrants while left-wing voters positive views. Left-wing voters are also more likely to have experienced extended unemployment and to report facing adverse economic circumstances. Trade union members, by contrast, are more likely than their respective baselines to support both radical right and radical left parties.
Determinants of right-wing and left-wing voting.
Note: p-values in parentheses. *p < .05, **p < .01, ***p < .001.
Source: Author's calculations using the European Social Survey.
We also gain new insight into the relationship between labor market compensation and populism. As shown in Table 5, higher spending on unemployment programs is negatively associated with support for radical left-wing parties. This relationship appears robust: increases in labor market spending and higher unemployment replacement rates both predict lower likelihoods of voting for the far left. By contrast, we do not find a statistically significant association between labor market spending and voting for the far right.
To probe whether the observed relationships between compensation and populist voting may be causal, we estimate several interaction models. We focus on three groups where populist voting is especially likely and where labor market spending may be most impactful: current or former trade union members, and individuals with prior unemployment experience. If unemployment protection influences populist support, then these groups should become less likely to back populist parties when unemployment benefits are more generous. To test the rival welfare chauvinist hypothesis, we also estimate the interaction model for immigration preferences and the UI replacement rate.
Figure 4 presents coefficient plots from the interaction models for populist, right-wing, and left-wing voting. The results indicate that, under more generous unemployment insurance systems, groups otherwise most prone to populist voting—union members and individuals with unemployment experience—are predicted to exhibit lower support for populist parties and left-wing parties than they would under less generous systems. Similarly, individuals with more pro-immigrant attitudes are predicted to be less supportive of populist and right-wing parties at higher levels of unemployment replacement, although this effect does not reach conventional levels of statistical significance (p < .05).
Impact of UI changes on populist support.
These counterfactual simulations imply that the negative association between labor-market support and populist voting extends beyond a simple cross-sectional correlation. Higher levels of social protection are associated with a lower marginal propensity to vote populist among core constituencies, such as current and former labor union members, relative to contexts characterized by weaker protection. The analysis further indicates that this association may be stronger among individuals with pro-immigration views, but it remains present when controlling for these views and when restricting the sample to individuals who hold negative views of immigrants. 4 Taken together, the results are consistent with the possibility that welfare institutions are linked to the political preferences of constituencies otherwise most predisposed toward populism.
Figure 5 illustrates the substantive significance of the relationship between UI generosity and voting for a populist (left panel) or far left party (right panel), using the interaction model reported in Figure 4. A 5 percentage point increase in the unemployment insurance replacement rate from its mean reduces the predicted probability of an average individual voting populist from 7.96% to 5.80%, when all other variables are held at their means. For a current or former trade union member, the effect is even greater, dropping from a 8.90% to 6.59% likelihood of voting populist. This represents a sharp decline, underscoring how more generous social protection may substantially dampen the electoral appeal of populist parties.
An individual's predicted likelihood of supporting a populist party or far-left party at different unemployment replacement rates.
Discussion
Our analysis provides insights into the relationship between labor market spending and support for populist parties. First, we find that higher levels of compensatory spending for unemployed workers are associated with smaller populist vote shares in national elections held in 16 European countries. We further demonstrate that both labor market spending and unemployment insurance generosity are negatively correlated with populist vote shares across multiple models, controlling for a wide array of macro-economic indicators. Notably, this relationship is not observed when we examine overall social welfare spending, social cash transfers or social services spending. 5
On the second dimension, we find evidence that reductions in spending on income maintenance since the 1990s, as well as austerity measures enacted following the eurozone crisis, contributed to the rising electoral fortunes of populist parties on the right and left. In the panel analysis, we demonstrate that within-country decreases in labor market spending are associated with higher populist vote shares. In the multi-level analyses, we show similarly that reduced labor market support is associated with a higher likelihood that voters will support a populist party or a far-left party. These effects are particularly pronounced among current and former trade union members and among those individuals who have previously experienced unemployment.
These relationships are observed across multiple model specifications in two independently generated datasets that cover different lengths of time. Our results are robust when we exclude nonvoters from the analysis; remove controls for occupation and education; restrict the sample to individuals opposed to immigration; and exclude certain European regions and countries. They also hold when using alternative measures of unemployment generosity and different measures of immigration, such as the size of the foreign-born population. Some key findings persist when we expand the sample to include eight Eastern European countries, which have historically structured labor market support quite differently. 6
Our findings suggest that long-term commitments to social programs aimed at insulating workers from the risk of economic disruption limit populist voting, while cuts to these programs, whether as a result of labor market reform or austerity, stimulate populist voting. Even after holding immigration attitudes constant, higher levels of labor-market compensation continue to reduce the predicted probability of supporting populist parties, suggesting that the dampening effect of social protection operates independently of immigration preferences. All in all, our evidence suggests that the compensation hypothesis is still relevant for explaining the rise of populism (Vlandas and Halikiopoulou, 2022). And, it suggests that in core areas of the welfare state such as unemployment insurance, the logic of compensation still holds its own against the welfare chauvinist logic that may predominate in other welfare state arenas (Burgoon and Schakel, 2022).
Since these results are observational, caution is warranted in drawing causal inferences. The consistent negative correlations between labor market expenditures and populist vote share could relate to factors omitted from this analysis, while the association between labor market reform and increased austerity on the one hand, and the rising fortunes of populist parties on the other hand, may reflect parallel historical trends which are not causally related. However, there are reasons to believe these relationships are not merely coincidental.
First, while there are some common movements, there is significant variation in both welfare spending and populist voting in the period examined. The model specifications we developed isolate this variation, controlling to the extent possible common historical developments through year fixed effects.
Second, we explored some of the micro-foundations of a potential causal link, demonstrating that many of the individuals most prone to support populist parties, including trade union members, and those who have experienced sustained periods of unemployment, become less likely to support populist parties when national unemployment insurance programs become more generous. We have also shown that the relationship is strongest in the welfare spending area that provides the most direct and immediate relief to those left facing economic distress, and which is most strongly supported by populist voters.
Finally, we have demonstrated that the observed relationship is robust to a variety of specifications and controls. Across multiple models and measures of spending, we have shown that populist parties are weaker in countries that spend more on compensation, and that cuts to welfare spending, as a result of labor market reform and austerity, are strongly associated with rising support for populist parties. Furthermore, we have identified plausible mechanisms by which compensation may affect political preferences by dampening the effects of globalization and technological change on livelihoods and thereby reducing the grievances among the groups most affected.
Conclusion
Europe's political systems are under challenge from populist movements and parties that reject core aspects of the post-World War II regional and international order. This challenge brings to mind long-standing arguments that the insecurity generated by economic change could be politically explosive if the concerns of those harmed were not addressed with adequate “compensatory” social policies (Kapstein, 1994; Rodrik, 1998; Ruggie, 1994).
Indeed, we find that higher levels of labor market spending help moderate support for populism among those who have seen their relative economic and social status decline. We also find that as European governments have cut unemployment and other programs that maintain incomes for established, full-time workers, these cutbacks have fueled support for populist parties opposed to core principles of European integration. Reductions in spending, especially on unemployment insurance and other labor market policies, have stimulated support for populism. These effects are most pronounced among the occupational and educational groups whose jobs are most vulnerable to deindustrialization and technological change, and among individuals who have experienced unemployment.
Our analysis suggests that appropriate social policies can limit the populist backlash, while labor market reforms and austerity measures can stimulate such a backlash. The relevant social and labor-market policies may be essential to long-term political stability. While a good case can be made for spending more on education and childcare, as well as programs that support refugees, the disabled, and other vulnerable individuals, these investments need not come at the expense of traditional compensatory programs such as unemployment insurance. The policy implications are clear, even if the political path to implementing appropriate policies is not.
Supplemental Material
sj-docx-1-eup-10.1177_14651165261446395 - Supplemental material for Compensation, austerity and populism: Labor market spending and voting in 16 Western European countries
Supplemental material, sj-docx-1-eup-10.1177_14651165261446395 for Compensation, austerity and populism: Labor market spending and voting in 16 Western European countries by Chase Foster and Jeffry Frieden in European Union Politics
Supplemental Material
sj-zip-2-eup-10.1177_14651165261446395 - Supplemental material for Compensation, austerity and populism: Labor market spending and voting in 16 Western European countries
Supplemental material, sj-zip-2-eup-10.1177_14651165261446395 for Compensation, austerity and populism: Labor market spending and voting in 16 Western European countries by Chase Foster and Jeffry Frieden in European Union Politics
Supplemental Material
sj-zip-3-eup-10.1177_14651165261446395 - Supplemental material for Compensation, austerity and populism: Labor market spending and voting in 16 Western European countries
Supplemental material, sj-zip-3-eup-10.1177_14651165261446395 for Compensation, austerity and populism: Labor market spending and voting in 16 Western European countries by Chase Foster and Jeffry Frieden in European Union Politics
Footnotes
Acknowledgments
We thank Lorenza Antonucci, Tim Büthe, Amy Catalinac, Stephen Chaudoin, Felix Hagemeister, Peter Hall, Sophie Hill, Iain Osgood, Stephanie Rickard, Tobias Rommel, Peter Rosendorff, David Steinberg, and two anonymous reviewers for valuable comments on earlier versions of this article. We are also grateful to participants in the Global Research in International Political Economy (GRIPE) online webinar, TUM's Behavioral & Empirical Work In Progress (BEWIP) Seminar, Harvard's Seminar on the State and Capitalism since 1800, and workshops at Johns Hopkins University and New York University. Finally, we thank the editors of European Union Politics for their guidance throughout the editorial process.
Author contributions
The authors contributed equally to the article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
The replication materials for this study are publicly available at the DOI of this article.
Supplemental material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
