Introduction
Europe has experienced several marked, cross-national trends in electoral results during the 21st century, including growing vote shares for radical right, radical left, and, latterly, green parties, and declining vote shares for social democrat, and, less emphatically, conservative, and liberal parties. This article asks: what has caused these changes? We contribute to existing explanations for these cross-country electoral trends, which have utilised two theoretical approaches: cleavage theory, which emphasises the role of long-term socio-structural transformations affecting policy attitudes and voter links with parties; and party entrepreneurship, which emphasises political actors strategically setting the public issue agenda to their own advantage. We propose that electoral outcomes have resulted from marked changes in what Europeans perceive as the most important issues affecting their countries, but that these are determined primarily by societal trends, in addition to party agenda setting. Our explanatory approach is therefore partially a synthesis of these two approaches, while incorporating other issue competition approaches regarding ‘issue convergence’ and beyond.
We use national-level and individual-level panel data to explain voting for six party families—conservative, social democrat, liberal, radical right, radical left, and green—across 28 European countries at both the national and individual level. At the national-level we augment Eurobarometer data with data from the Manifesto Project Database, Eurostat, and the Global Historical Climatology Network, at the individual level we use European Election Study augmented with Chapel Hill Expert Survey data, and British Election Study panel data, and we support our claims with evidence from the European Social Survey and European Values Study. Using panel data, we show that the public salience of immigration increases the vote share of the radical right and reduces that of conservative, social democrat, and radical left parties. Similarly, the public salience of unemployment increases the vote share of the radical left and social democrats and reduces that of radical right and green parties; and the public salience of the environment increases the vote share of the greens and reduces that of the social democrats and radical left. In turn, we show that the public salience of each of those issues is driven by societal trends—immigration and unemployment rates and temperature anomalies—as well as agenda-setting by parties. We find that the party system issue agenda, which has mixed effects on party family vote share, is partially set by public issue salience but not by our societal trends. Together, these results have significant theoretical implications which we discuss.
In line with party issue competition and ‘wave-riding’ theoretical approaches, we argue and show that the agency of parties to set the terms of the political debate to their own advantage is highly constrained—at least on the issues analysed in this article—and that they are partially forced to chase the issues that voters decide are important. As such, democratic politics is driven by forces partially exogenous to it, with longer-term socio-structural changes episodically manifested in shorter-term, newsworthy, and vote-shifting events. If societal trends and events continue to become more transnational, domestic party competition across countries will become subject to increasingly synchronous trends, as has happened in Europe, further limiting domestically focused political entrepreneurs and other “agenda setters”. Finally, whereas party systems and public attitudes are formed by long-term socio-structural forces and resultant social groups, shorter-term volatility is likely to result from considerations over the most important issues of the day, so that one’s shorter term cognitive and behavioural reconciling of the political and social worlds is primarily dictated by prioritisation. More substantively, our approach also sees recent electoral transformations as potentially fleeting, with the public’s issue agenda having the potential to change in a way that no longer benefits respective non-mainstream parties, notwithstanding longer-term structural changes. Finally, in our discussion we consider the challenges of this approach, particularly the complexities of measuring short-term, societal trends and ‘issues’ in a theoretically grounded way that recognises the role of the media as mediators and supply-side considerations, not least regarding how parties fight to claim advantageous reputations over issues of public concern.
Existing explanations for Europe’s changing electoral landscape
The starkness and novelty of recent electoral trends in Europe have prompted significant academic enquiry, which we argue can be imperfectly divided into two theoretical camps. The first strand of explanations sees electoral change as the outcome of long-term socio-structural transformations. These approaches are generally based on the assumptions of cleavage theory (Lipset and Rokkan, Reference Lipset, Rokkan, Lipset and Rokkan1967), according to which party systems reflect conflicting interests amongst social groups and remain rigid until disrupted by rare ‘critical junctures’ (for review see von Schoultz, Reference Von Schoultz, Arzheimer, Evans and Lewis-Beck2016). In this theoretical vein, parties have minimal agency over the forces determining long-term trends in electoral results and are, rather, the political manifestation and representation of those forces. Numerous scholars have utilised this theoretical approach to explain recent changes in Europe. Hutter and Kriesi (Reference Hutter and Kriesi2019; see also, e.g., Kriesi et al., Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2006, Reference Kriesi, Grande, Dolezal, Helbling, Höglinger, Hutter and Wüest2012; Dalton, Reference Dalton2018; Hooghe and Marks, Reference Hooghe and Marks2018) have argued that long-term social transformations resulting from globalisation are reshaping European party systems by ‘opposing those who benefit from this process against those who tend to lose’ (Kriesi et al., Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2006: 921), with the ‘losers’ turning to the radical right and ‘winners’ turning to the greens, who had already been conceptualised as beneficiaries of earlier social transformations including the ‘silent revolution’ towards post-material values and the related expansion of higher education (Inglehart, Reference Inglehart1977; Kitschelt, Reference Kitschelt, Dalton and Kuechler1990). Hooghe and Marks (Reference Hooghe and Marks2018: 17) focus less on long-term socio-structural forces and more on the short-term effects of the Eurozone crises and the ‘migration crisis’ on raising the salience of the ‘transnational cleavage’, still determined by the ‘winners’ and ‘losers’ of globalisation. Notably, in each of these accounts, salience is conceptualised and measured as the emphasis that parties, rather than voters, place on issues.
Several party-specific explanations for electoral trends have also taken a similarly socio-structural approach but without the assumptions of cleavage theory. The decline of social democratic parties has been explained by the shrinking of the traditional working class, de-industrialisation and ‘embourgeoisement’ (Benedetto et al., Reference Benedetto, Hix and Mastrorocco2020; Gingrich and Häusermann, Reference Gingrich and Häusermann2015) and attitudinal divides between middle-class and working-class centre-left voters (Abou-Chadi and Wagner, Reference Abou-Chadi and Wagner2019, Reference Abou-Chadi and Wagner2020; Keating and McCrone, Reference Keating and McCrone2013). Accounts for the rise of the radical right have similarly focused on either economic competition or cultural threat deriving from increasing immigration and globalisation (e.g., Inglehart and Norris, Reference Inglehart and Norris2019; for a review, see Arzheimer, Reference Arzheimer and Rydgren2018). Whereas the electoral growth of green parties has been argued to result from value change and educational expansion, increased radical left success has also been interpreted in socio-structural terms attributing it to the support by the ‘losers of globalisation’, the north-south divide within the Eurozone, unemployment, and anti-globalisation attitudes (March and Rommerskirchen, Reference March and Rommerskirchen2015; Visser et al., Reference Visser, Lubbers, Kraaykamp and Jaspers2013).
The second strand of explanations sees recent electoral change in Europe as the result of shorter-term strategic party behaviour and ‘entrepreneurship’ (see Budge, Reference Budge2015, for earlier findings on party attempts to strategically set the issue agenda). Notably, de Vries and Hobolt (Reference De Vries and Hobolt2020: 6) argue that oligopolies of ‘dominant’ parties, subject to decreasing voter loyalty, have been under ‘attack’ by challenger parties, who challenge by ‘introducing issues that can drive a wedge between coalitions’ and using anti-establishment rhetoric. Similarly, Green-Pedersen (Reference Green-Pedersen2019) explains party system change via new issues that are determined by the strategic considerations of political parties. The key assumptions of these approaches are that parties have a high degree of agency over what issues are considered important to the public, in direct contrast to the passive conceptualisation of cleavage theoretical approaches. According to this approach, disruptive, challenger parties have introduced political issues that are new or ignored by mainstream parties—such as immigration, the environment and, in southern Europe, unemployment—leading to greater interest in those issues among the voters, who change their voting behaviour accordingly. Additionally, there exist other party-centric or ‘supply-side’ accounts for recent changes, such as those focussing on social democratic parties ideologically moving to the centre (Rennwald and Evans, Reference Rennwald and Evans2014; Escalona and Vieira, Reference Escalona and Vieira2014); and explanations focusing on centre-left parties being ‘squeezed’ by radical left, green and radical right parties (Cuperus, Reference Cuperus, Keating and McCrone2018).
Our explanatory approach
We contribute to the above explanations for recent electoral trends in Europe by integrating structural changes and the issue competition literature, and by incorporating distinct extant theoretical approaches from the broader issue competition, issue salience, and issue ownership literatures.
The broad issue competition literature is comprised of numerous strands with distinct theoretical assumptions. Although argued to ‘lack … conceptual clarity’ (Guinaudeau and Persico, Reference Guinaudeau and Persico2014: 312), these approaches share a focus on variation in the emphasis that parties place on issues, shown to be of increasing importance to voting compared to positional considerations (Green-Pedersen, Reference Green-Pedersen2006; Reference Green-Pedersen2007). The more prominent strand, which emphasises strategic agenda-setting and ‘entrepreneurship’ by parties, has been applied to explaining Europe’s recent electoral trends, as identified above. It argues that parties emphasise the issues that they ‘own’ (Petrocik, Reference Petrocik1996; Petrocik et al., Reference Petrocik, Benoit and Hansen2003, see below for discussion on issue ownership) and that doing so is advantageous electorally (Brazeal and Benoit, Reference Brazeal and Benoit2008; Budge and Farlie, Reference Budge and Farlie1983). The second theory—the ‘wave-riding’ theory and related findings of ‘issue convergence’—argues, instead, that parties either emphasise those issues that are salient to the media (typically used as a proxy for the public; Ansolabehere and Iyengar, Reference Ansolabehere and Iyengar1994; Dalmus et al., Reference Dalmus, Hänggli and Bernhard2017) or public (Wagner and Meyer, Reference Wagner and Meyer2014; Damore, Reference Damore2005; Sigelman and Buell, Reference Sigelman and Buell2004; Green and Hobolt, Reference Green and Hobolt2008; Sides, Reference Sides2006) and that doing so is electorally advantageous, particularly if the party also owns the issue (Abbe et al., Reference Abbe, Goodliffe, Herrnson and Patterson2003).
When parties ‘own’ issues, their electoral outcomes become linked with the public issue salience of those issues, regardless of their strategic approach (Bélanger and Meguid, Reference Bélanger and Meguid2008). This ‘ownership’ has been shown to result from two concepts: perceived competence and broader association. Perceived competence in dealing with an issue is more short-term and primarily determined by governing party performance evaluations (e.g., Green and Jennings, Reference Green and Jennings2017, Reference Green and Jennings2012; Craig and Cossette, Reference Craig and Cossette2020; Lanz and Sciarini, Reference Lanz and Sciarini2016; Petrocik, Reference Petrocik1996; Petrocik et al., Reference Petrocik, Benoit and Hansen2003) and perceived intra-party unity (Greene and Haber, Reference Greene and Haber2017) rather than by opposition or challenger party activity. However, long-term prioritisation of an issue by a party also leads to association with that issue (Walgrave et al., Reference Walgrave, Lefevere and Tresch2012, Reference Walgrave, Tresch and Lefevere2015; Grossmann and Guinaudeau, Reference Grossmann and Guinaudeau2021), from which ‘challenger’ party families like the radical right, radical left, and greens are particularly likely to benefit (Seeberg, Reference Seeberg2017, on the long-term and cross-national stability in issue ownership links). Indeed, recent electoral changes in Europe have often been preceded by the perceived loss of competence of multiple mainstream parties on a publicly salient issue (like immigration, unemployment, and the environment) followed by the rise of previously marginal challenger parties that have the benefit of long-term prioritisation of that issue and internal unity on it (Davidsson and Marx, Reference Davidsson and Marx2013; Hobolt and de Vries, Reference Hobolt and de Vries2015). Moreover, the ability of parties to change what they own is limited, primarily to issues that already had weak party ownership reputations (Tresch et al., Reference Tresch, Lefevere and Walgrave2015; Walgrave et al., Reference Walgrave, Lefevere and Nuytemans2009). Overall, ‘party competition plays out more by what politicians actually do and what impression they give about their deeds and competence than by exercising influence on the issue agenda’ (Meyer and Müller, Reference Meyer and Müller2013: 484).
Recent findings have shown that parties are highly constrained in their ability to set the public issue agenda and, to varying extents, are forced to follow the voters—in short, ‘Parties respond to voters, but voters do not necessarily respond to parties’ (Klüver and Spoon, Reference Klüver and Spoon2016: 663; also, Spoon and Klüver, Reference Spoon and Klüver2014; Burdina, Reference Burdina2014; Stimson, Reference Stimson2004; Klüver and Sagarzazu, Reference Klüver and Sagarzazu2016; Grossmann and Guinaudeau, Reference Grossmann and Guinaudeau2021). Similarly, Barberá et al. (Reference Barberá, Casas, Nagler, Egan, Bonneau, Jost and Tucker2019: 898) show that the public’s issue agenda leads that of elected politicians more so than vice versa. Larger, more well-resourced parties and more senior party members have been shown to particularly respond to public concerns due to office-seeking motivations (Wagner and Meyer, Reference Wagner and Meyer2014; Ennser-Jedenastik et al., Reference Ennser-Jedenastik, Haselmayer, Maria Huber and Fen2022) and to the collective party system issue agenda (Green-Pedersen and Mortensen, Reference Green-Pedersen and Mortensen2015).
Instead, recent findings have also shown that public issue salience is partially determined by exogenous events and trends independently of political and media cues (Singer, Reference Singer2011; Klüver and Sagarzazu, Reference Klüver and Sagarzazu2016; McAllister and bin Oslan, Reference McAllister and bin Oslan2021; Dennison, Reference Dennison2019, Reference Dennison2020; Gilardi et al., Reference Gilardi, Gessler, Kubli and Müller2021; Seeberg, 2017; Reference Seeberg2022). Media agenda-setting—outside of mediating the effects of and framing exogenous social trends and events—is constrained in its selection of what issues to cover (Amsalem and Zoizner, Reference Amsalem and Zoizner2022; Pritchard and Berkowitz, Reference Pritchard and Berkowitz1993; Gentzkow and Shapiro, Reference Gentzkow and Shapiro2010) and follows more than influences both public trends and political actors (Zoizner et al., Reference Zoizner, Sheafer and Walgrave2017; Kleinnijenhuis and Rietberg, Reference Kleinnijenhuis and Rietberg1995; Brandenburg, Reference Brandenburg2002; Soroka, Reference Soroka2003). Indeed, Barberá et al. (Reference Barberá, Casas, Nagler, Egan, Bonneau, Jost and Tucker2019: 883) show that the ability of the public to affect the party system issue agenda is maintained even after controlling for the agenda-setting effects of the media, while Amsalem and Zoizner (Reference Amsalem and Zoizner2022: 234) show that citizen competence is sufficiently high to mean that media influence is largely diminished in realistic scenarios. Overall, agenda-setting has been argued to be largely exogenous to both parties and media (Gilardi et al., Reference Gilardi, Gessler, Kubli and Müller2021: 190). Indeed, the recent cross-national nature of trends in public issue salience in Europe also suggests that the determinants of the issue agenda are ultimately exogenous to divergent domestic party or media agenda-setting capabilities (Ford and Jennings, Reference Ford and Jennings2020).
The above findings suggest that: (1) the issues that the public find important affect the issues that politicians emphasise (‘wave-riding’ and ‘issue convergence’) and, possibly to a lesser extent, vice versa; (2) the electoral fortunes of respective parties are linked to certain issues via ‘ownership’, dictated both by shorter-term perceptions of party competence, especially for mainstream and governing parties, and longer-term issue associations, particularly for challenger and niche parties; and (3) the issues that the public finds important vary according to societal trends and ‘real-world’ events. Overall, these relationships are summarised in Figure 1.
Figure 1. Theoretical framework.
Source: Author’s own elaboration
How do these theoretical propositions relate to and complement the two major strands of explanations for recent changes in Europe’s electoral outcomes? Both the above findings and cleavage theoretical approaches emphasise societal trends. However, whereas cleavage theoretical approaches focus on medium- and long-term structural trends that fundamentally affect society and its political cleavages but may less acutely form part of the political conversation, such as changing climatic or trading patterns, public issue salience is more likely rooted in short-term social trends and events that become of public interest due to their relative and perceived uncertainty, risk, novelty, and need to be immediately resolved (Dennison, 2021; Reference Dennison2019 for review on public issue salience), such as natural disasters and economic or migratory crises (often episodic expressions of the longer-term structural trends). Similarly, compared to the longer-term explanatory power of extant cleavage theoretical approaches, the rapidity of trends visible in Figure 2, such as over 80 per cent of Germans listing immigration as the most important issue affecting their country in 2016, up from less than 10 per cent two years earlier, highlight the limits of relying on long-term cleavage theoretical approaches only. Finally, the volatility of issue salience reflects its ordinal measurement (‘most important’) that makes each issue’s salience relative to that of other issues, in line with psychological findings on the tendency of humans to prioritise (and their behaviour to be dictated by) only a few issues at a time (Boninger et al., Reference Boninger, Krosnick, Berent, Fabrigar, Petty and Krosnick1995; Dennison, Reference Dennison2019).
Figure 2. The demand-side issue salience of the environment, immigration, and unemployment across the EU28.
Notes: Measure (y-axis) is percentage of the population responding to ‘what do you think are the two most important issues affecting your country?’ Source: Eurobarometer, 2005–2019
Related to the issue competition literature, short-term variation in issue salience may have long-term consequences for party preferences via the breaking and creating of voting habits (Shachar, Reference Shachar2003; Achen and Bartels, Reference Achen and Bartels2016). Moreover, although parties have agency over public issue salience via agenda-setting (e.g., Green-Pedersen and Mortensen, Reference Green-Pedersen and Mortensen2010; Soroka, Reference Soroka2003; Kingdon and Stano, Reference Kingdon and Stano1984; Baumgartner and Jones, Reference Baumgartner and Jones1993), because this relationship also works in reverse (and more strongly so, according to findings outlined above), such agency is ultimately constrained by external, societal developments and the need to be seen to respond to them competently (Grossmann and Guinaudeau, Reference Grossmann and Guinaudeau2021).
Recent findings, and the above considerations, are therefore highly supportive of the original assumptions of both issue voting and cleavage theory. Regarding issue voting, as Budge (Reference Budge2015: 766; see also Bélanger and Meguid, Reference Bélanger and Meguid2008) stated, ‘election issues appeared as the result of external developments rather than party manipulation. Parties might try to emphasise favourable issues to suit prevailing circumstances. But the latter dictated the agenda’. Regarding cleavage theory, Lipset and Rokkan (Reference Lipset, Rokkan, Lipset and Rokkan1967: 94) observed that short-term variation in electoral outcomes results from a ‘hierarchy of cleavage bases’ according to their relative ‘political weight’ as determined by societal trends and events exogenous to the party system, rather than media or political cues. In sum, both the public’s and the politicians’ agendas are ultimately at the mercy of what British Prime Minister, Harold Macmillan, in 1957 described as the major challenges to his government: ‘events, dear boy, events’.
Hypotheses
Based on the above discussion, we produce four hypotheses. We choose three issues with which to test them for both theoretical and empirical reasons: immigration, unemployment, and the environment. Empirically, in terms of public issue salience, these issues stand out throughout the 21st century in Europe both for the scale of their change and the cross-national trends in their variation, as we describe below. The three issues are regularly respectively associated with the radical right, radical left, and greens parties (Budge Reference Budge2015: 769, Table 2). The issue ownership of environmental issues by the Greens is particularly pronounced (Abou-Chadi Reference Abou-Chadi2016: 421). For attendant societal trends, we use immigration rates, temperature anomalies and unemployment rates. Although other societal trends, such as irregular immigration rates or flooding, may make alternative or even more powerful predictors of the salience of these issues, we choose the theoretically broadest possible metrics (see also McAllister and bin Oslan, Reference McAllister and bin Oslan2021; Hoffmann et al., Reference Hoffmann, Muttarak, Peisker and Stanig2022). Thus, we produce the following hypotheses:
H1. An increase in the public issue salience of immigration, unemployment, and the environment, each negatively affect the vote shares of conservative, social democrat and liberal parties, and, respectively, positively affect the vote shares of radical right, radical left, and green parties.
H2. An increase in, respectively, the immigration rate, unemployment rate, and temperature anomaly is associated with: (H2a) an increase in the respective public issue salience of immigration, unemployment, and the environment; and (H2b) increase in the proportion of the party system issue agenda devoted to those three issues respectively.
H3. An increase in the proportion of the party system issue agenda devoted to, respectively, immigration, unemployment, and the environment, covaries with an increase in the respective public issue salience of those three issues and, additionally, we expect the effects of public issue salience on the party system issue agenda to be stronger than vice versa.
Methods
To test Hypothesis H1, we use fixed effects panel data models at the aggregate level. Such models test only the effects of within-country variation in the explanatory predictors on the within-country variation in the dependent variable. As such, all unobserved time-invariant between-country effects are controlled for. At the national level this includes, for example, the historic issue agenda or party system, and fixed economic, demographic, political (e.g., the electoral system), and social factors. We test the respective effects on the national vote share of six respective party families across Europe and controlling for that country’s time invariant fixed effects, resulting in eighteen models—predicting the vote share of each of the six party families using the three respective issues.
For Hypothesis H2a, we again perform fixed effects panel data models at the aggregate level to test the effects of the three societal trends, respectively, on the public salience of each issue. We first test the effect of just the societal trend on public issue salience, then do the same controlling for waves (i.e., public issue salience trends common across Europe and their common determinants), and, third, additionally controlling for the proportion of the party system agenda devoted to that issue. In the final model, we use controls for an interaction effect of days since the last election and the party system issue agenda, since (as described below) our data for the party system issue agenda is taken from election manifestos and, so, this effect is likely to wane throughout the electoral cycle (though we know that parties are internally constrained to use their manifesto to inform their strategy not just on election day but thereafter, e.g., Meyer, Reference Meyer2013). For Hypothesis H2b, we again perform fixed effects panel data models of the effects of both the respective public issue salience and societal trends on the proportion of the party system agenda devoted to each issue.
Hypothesis 3 is tested in the models testing H2a and H2b since the effects of the party system issue agenda on public issue salience and vice versa are tested in those models. However, to further test H3 we also run more robust but less efficient panel vector autoregression models with two dependent variables to test the direction of effects, with granger causality Wald tests, reported and discussed further in the online Appendix 2.
We further test Hypothesis 1 at the individual-level using logistic models that predict switching to one of the six respective party families amongst individuals who did not vote for the party family in the previous election or did not vote at all. We do so amongst only those who did not vote for the party at the last election to capture the predictors of change—since our driving question is explaining change—rather than predicting static vote choice. Crucially, these models include not only issue salience measures but also control for policy preference measures for the three issues, to control for the effects of attitudinal change as a potential cofounder because, despite aggregate-level attitudinal stability, there may still be significant and effective individual-level attitudinal change (Druckman and Leeper, Reference Druckman and Leeper2012). Standard errors are clustered by country. We also run individual-level fixed effects panel data models as a robustness check (reported in Appendix 2) that again control for all time invariant predictors. At the individual-level these include (typically) gender and other socio-demographics, pre-adulthood socialisation, and psychological predispositions.
Data
We use three sources of data to test Hypothesis 1 at the national-level. We construct a dataset built around our public issue salience data, for which we use data from the Eurobarometer surveyFootnote 1. Since 2005 it has asked a representative sample of every respective country in the European Union ‘what do you think are the two most important issues affecting your country?’. Surveys are typically biannual (though occasionally less) so that each country has 29 observations during the period 2005–2019 (the UK and Germany had an extra survey one year so have 30 observations). Respondents are offered around 14 responses, which have changed occasionally over time. We use the percentage responding ‘immigration’, ‘unemployment’, and ‘climate change and the environment’, respectively, in each country. We match this with data measuring the party system issue agenda (Green-Pedersen and Mortensen, Reference Green-Pedersen and Mortensen2010) at each country’s most recent election from the Manifesto Project Dataset (Volkens et al., Reference Volkens, Burst, Krause, Lehmann, Matthieß, Regel, Weßels and Zehnter2021) which includes the emphasis—measured as mentions—that each party gave to various political issues in the most recent election manifesto in most EU and OECD (and beyond) countries, going back to the mid-twentieth century. We use this source of data because it includes all countries under analysis. We create an indicator corresponding to the weighted sum of the manifesto scores for every party at the most recent election preceding the respective Eurobarometer surveys, where each party is weighted by its vote share, resulting in three respective party system issue agenda indicatorsFootnote 2 per country. The sources of vote shares are from both electoral results and opinion polling—for each EU member state from 2005 to 2019 and these are outlined in the online Appendix 1. We take opinion polls on vote intentions only from those months in which the Eurobarometer was surveyed, producing an average of all opinion polls in that country in the respective month unless there is a parliamentary election that month, in which case we use the results of the election (in France we use the first round of the presidential election). The party family classification follows the ParlGov classification of parties into party families (Döring and Manow Reference Döring and Manow2019). Overall, each observation in the dataset includes the public issue salience as reported by the Eurobarometer that month, the (constructed) party system issue agenda as reported by the Manifesto Project Dataset at the most recent election at the time of the Eurobarometer survey, and vote shares as measured by either opinion polling averages or, when possible, election results that month.
For Hypothesis 2, we again use three sources of data: the same data as above for public issue salience and the party system issue agenda (also used to test Hypothesis 3) and data for each of our respective three societal trends. Immigration rates are measured using Eurostat data from 2007 to 2018 on new immigrants per year, which is then divided by population size. The mean is 0.0098 and the standard deviation is 0.0089, so that, on average, for every 100 citizens in an EU member state, almost one new immigrant arrived per year during the period. Monthly unemployment rates are also obtained from Eurostat starting in 2005, with a mean of 8.57 and a standard deviation of 4.47. We produce values of temperature anomalies using Global Historical Climatology Network (GHCN, 2019) data and calculated as that month’s deviation from the mean surface temperature in each country’s capital city in the respective month over the period 1960 to 2004. The largest differential is 7.0° in November 2011 in Poland, which had a temperature that month of 5.4° and a mean temperature during the reference period of the Novembers between 1960 and 2004 of −1.7°. The mean is 1.49 and the standard deviation is 1.29, so that, on average, each EU member state was 1.49 degrees hotter during the time series than their respective average during the period 1960 to 2004.
To test Hypothesis 1 at the individual-level, we use the 2014 European Election Study (EES) voters survey (Schmitt et al., Reference Schmitt, Hobolt, Popa and Teperoglou2016), which included representative samples in all the then 28 EU member states. It asked respondents about: their previous vote choice in national elections and how they would vote if there were an upcoming election; what they consider to be the most and the second important issue affecting their country; several policy attitudes on a 0–10 scale (immigration policy; prioritising environmental protection over growth; and state intervention in the economy). For the sake of party classification, we use the 2014 Chapel Hill Expert Survey (Polk et al., Reference Polk, Rovny, Bakker, Edwards, Hooghe, Jolly, Koedam, Kostelka, Marks, Schumacher, Steenbergen, Vachudova and Zilovic2017) because it can be linked to the EES. Those countries that do not have one of the respective parties are removed from the respective model. Our panel data robustness check uses British Election Study data and is described in Appendix 2.
Analyses
Changing levels of issue salience, distributions of policy attitudes, and party family electoral support
In Figure 2, we plot the changing public issue salience of the three issues for each EU member state from 2005 to 2019. We group the member states largely along regional lines—north-western Europe, southern Europe, and central and eastern Europe. In all three regions, the sequence of the ups and downs of the salience of the three issues is the same with unemployment becoming most salient first, followed by the increasing salience of immigration, which in turn was replaced by the environment. However, there are region-specific variations to this overall pattern with immigration and, later, the environment reaching a higher peak in north-western Europe.
By contrast, when we consider policy preference change across the same three issues in Figures A1, A2, and A3 in Appendix 2, we see remarkable stability and very little polarisation from the early 2000s to the late 2010s, in line with previous findings and making the latter a less plausible predictor of electoral change over time at the national level.
The effects of public issue salience and party system issue agenda on electoral outcomes
Hypothesis 1 is tested at the national-level using the 18 models in Table 1. There are three models—one for each party—across each of the six party families. For the purposes of comparison, both public issue salience and the party system issue agenda are standardised to allow for comparison of effect sizes.
Table 1. Fixed effects panel models of the effect of public salience and the party system agenda salience on vote shares of six party families
Notes: Standard errors in parentheses. *** P < 0.01, ** P < 0.05, * P < 0.01.
We see strong, largely consistent evidence to support Hypothesis 1. Notably, in terms of the party-specific hypotheses whereby the three mainstream parties are hurt by the rise of public salience of the three issues and the three challenger parties respectively benefit from them, we see five statistically significant effects in the expected direction and one in the contrary direction (social democrats benefitting from the public salience of unemployment together with the radical left). For immigration, public salience is shown to positively affect the vote share of the radical right and negatively affect the vote share of conservative, social democrat and, to a lesser extent, radical left parties (with no statistically significant effect on the liberals or greens). The public salience of unemployment increases the vote share of the radical left and, to a lesser extent, the social democrats, while negatively affecting the radical right and greens (with no statistically significant effect on the conservatives or liberals). Similarly, the public salience of the environment strongly increases the vote share of the greens and reduces the vote share of the radical left and, to a lesser extent, the social democrats (with no statistically significant effect on the conservatives, liberals, or radical right).
Of note, the effects of the party system issue agenda are more mixed. In terms of comparison, the standardised effect sizes show that, for unemployment, the party system issue agenda and public issue salience have similar sized effects for the three ‘challenger’ parties but the effects of party system issue agenda are far larger for the mainstream parties. Indeed these are two of the largest effects in all of the 18 models (a one standard deviation increase in party discussion over unemployment, takes 2–2.5 percentage points off the centre right and centre left vote shares), perhaps a reflection of the incumbency status these parties more often hold and thus the blame they receive for unemployment. The other very strong effect is the public salience of immigration on radical right vote share (it also has fairly strong negative effects on conservative and social democrat vote share) compared to the almost universal lack of statistical significant effect on electoral results of the proportion of the party issue agenda devoted to immigration—in short, the disruptive effects of immigration as an issue come from political demand not supply. Whereas the effects of the unemployment issue come both from the political supply side and demand side, and the effects of the immigration issue come primarily from the demand side, the effects of the environment issue are more mixed in their roots, even if the directions of the effects are largely intuitive.
We ran several models to further investigate these relationships. At the regional level we found strong negative effects of unemployment on social democrats in southern Europe and strong positive effects in the two other regions. We also ran dynamic Arellano-Bond estimator panel-date models with similar results, albeit lower efficiency.
The effects of real-world trends and party system agenda on public issue salience
Our tests of Hypothesis 2a are reported in Table 2, which are largely supportive. Each of the societal trends has a positive, statistically significant effect on the salience of its respective issue when tested alone, in models 1, 4, and 7, respectively. A one standard deviation increase in new immigrants per capita increases the percentage of individuals listing immigration as the most important issue by 8.4 percentage points. A one standard deviation change in the unemployment rate increases the salience of unemployment by 17.3 percentage points. Finally, a one standard deviation change in the temperature anomaly increases the salience of the environment by 0.7 percentage points. When wave controls are added the effect disappears only in the case of temperature anomalies (model 8), possibly the result of greater transnational consistency in climate change or greater public consciousness over time. When the party system issue agenda effects—all of which are statistically significant—are added (models 3, 6, and 9), the effects of the two societal trends remain statistically significant. We also include an interaction term for ‘days since the previous election’ in case the power of agenda-setting wanes over time (see Dekeyser and Freedman Reference Dekeyser and Freedman2021), which is statistically significant in the case of unemployment, giving party system issue agenda an overall positive effect after two years. As such, the first test of our Hypothesis 3 is also supportive: public issue salience and the party issue agenda covary. Because both the independent variables are standardised, we can compare their relative effects, seeing that societal trends had a far larger effect than party issue salience in the case of immigration and unemployment, while the reverse is true in the case of the environment.
Table 2. Fixed effects panel data models of the effect of societal trends and party system issue agenda on public issue salience
Notes: Standard errors in parentheses. *** P < 0.01, ** P < 0.05, * P < 0.1; dichotomous wave controls not shown
The effects of societal trends and public issue salience on the party system issue agenda
Our tests of Hypothesis 2b is displayed in Table 3, which are unsupportive of the Hypothesis that the party system issue agenda is affected by societal trends. In none of the cases are the societal trends statistically significant, meaning that any effect that they have is via public issue salience (given the support for Hypothesis H2a in Table 2 above). The effect of public issue salience is positive and statistically significant for immigration and the environment, but not for unemployment. This suggests that, in the cases of immigration and the environment, parties respond to public issue salience and only indirectly in response to societal trends.
Table 3. Fixed effects panel data models of the effect of societal trends and public issue salience on the party system issue agenda
Notes: Standard errors in parentheses. *** P < 0.01, ** P < 0.05, * P < 0.1; Year controls not shown
Together, Tables 2 and 3 also test our Hypothesis 3 of covariance between the public issue salience and the proportion of the party system issue agenda devoted to our three issues. However, we only find evidence to support this in one direction: public issue salience affects the party system issue agenda but not vice versa. As such, our Hypothesis 3 of covariance is rejected and our Hypothesis 3a is partially supported. This finding is supported by our more robust but less efficient panel vector autoregression models with the measures as two dependent variables, in Appendix 2, along with panel vector autoregression granger causality WALD tests (Abrigo and Love, Reference Abrigo and Love2016; Sims, Reference Sims1980; Bose et al., Reference Bose, Hravnak and Sereika2017), which show that the public issue salience of immigration granger causes the party issue agenda devoted to immigration, to an even greater extent than the latter’s own lag. For the other two issues, however, there is no additional confirmation of Hypothesis 3a.
The individual-level effects of salience on vote choice
We now validate the relationship between salience and vote switching at the individual level, allowing us to see if this relationship holds when attitudinal change is controlled for. As shown in Table 4, using EES data we find individual-level evidence that the public issue salience of, respectively, immigration, unemployment, and the environment, increases the chance of switching to radical right, radical left, or green parties, respectively. This strongly supports Hypothesis 1. By contrast, conservative and liberal parties repel voters who perceive the environment as highly salient, and the liberals also repel voters who perceive unemployment as such. In these models, only the social democrats are not affected by the voters’ salience considerations. Regarding these mainstream parties, then, we only see partial support for Hypothesis 4. We also see that policy attitudes induce switching, too. Independent of salience, pro-immigration attitudes induce switching to the social democrats, radical left, and the greens, and preclude switching to the radical right. Opposition to state intervention in the economy encourages switching to conservative parties and precludes switching to social democrat parties. Finally, anti-environmentalism encourages switching to conservative parties and precludes switching to green parties. These results confirm the complementary relevance of spatial and salience voting for electoral change, despite aggregate-level policy preference attitudinal stability (see Druckman and Leeper, Reference Druckman and Leeper2012).
Table 4. Individual-level predictors of intending to switch one’s vote to each of the six party families
Notes: Standard errors in parentheses. *** P < 0.01, ** P < 0.05, * P < 0.1; Standard errors clustered by country; Socio-demographic controls not shown. European Election Study, 2014
In Appendix 2, we see the results of five fixed effects panel data models using British Election Study data acting as a robustness check for Hypothesis 4. The results are strong in their support of Hypothesis 1: changing to see immigration as salient has a statistically significant negative effect on one’s chance of voting conservative and social democrat, and positive one on voting for the radical right; changing to see the environment as salient has a statistically significant effect and also has a statistically significant negative effect on one’s chance of voting conservative and social democrat as well as a positive one on voting for the greens. In both cases, the effects of the salience of two issues are of a greater magnitude than the effects of the analogous policy preferences in every Hypothesised relationship (in these models there is no radical left party).
Overall, our analyses support Hypotheses 1, 2a, and 3, but not 2b. In other words, Europe’s changing electoral outcomes can be explained by changing societal trends affecting public issue salience, which in turn affects the party system issue agenda. Changing public issue salience, but also party system issue agendas, then affect voting for all six of Europe’s party families, even when controlling for corresponding attitudes.
Discussion
In this article we asked what caused the marked, cross-national, unprecedented trends in European electoral results in the 21st century. We built on previous findings and theoretical approaches to offer a novel, synthetic explanation for these specific changes that builds on extant theoretical approaches. We take a comprehensive empirical approach, first, assessing changes in the policy attitudes and public issue salience of three issues—unemployment, immigration, and the environment—across the EU and then using national-level panel data to show that the public salience of these issues is partially rooted in actual unemployment rates, immigration rates and environmental change, to a greater extent than in party agenda setting. By contrast, the party system issue agenda seems to be driven by the concerns of the public, but not by societal trends. We show that public issue salience of immigration increases the vote share of the radical right and reduces that of conservative, social democrat, and radical left parties; the salience of unemployment increases the vote share of the radical left and social democrats and reduces that of radical right and green parties; while the salience of the environment increases the vote share of the greens and reduces that of the social democrats and radical left. These findings support and expand the well-known claims of issue voting theory and findings regarding the centrality of salience to the effects of issue ownership (Bélanger and Meguid, Reference Bélanger and Meguid2008). The party system issue agenda’s emphasis on immigration and the environment has weaker and only intermittent effects, while its emphasis on unemployment has large and consistent effects for respective party families as we would theoretically expect. The implications of our findings for the party system issue agenda and vote choice are that the issues parties choose to emphasise—collectively the issue agenda—matter for vote choice but to an extent that varies by issues, and that (for immigration and the environment at least), parties are partially following the public’s priorities anyway. The power of parties to affect vote choice by agenda setting—at least in the aggregate—proves to be limited, as increasing evidence elsewhere has shown. We largely validate the effects of public issue salience at the individual-level, incorporating complementary spatial motivations, with cross-sectional pan-European data and panel data from the UK.
We distinguish between the profound, slow shifts in structural social cleavages and aggregate-level political attitudes, on the one hand, and the short-term volatility of societal events and trends, public issue salience and electoral outcomes on the other. The basic parameters are changing slowly under the surface whereas what we see in fast-moving party competition is the result of shorter-term considerations that may disappear as quickly as they arise, even though they each are to some extent rooted in real-world changes. Recent electoral changes in Europe are therefore in line with what Lipset and Rokkan (Reference Lipset, Rokkan, Lipset and Rokkan1967: 94) argued, as well as with the arguments of the key proponents of issue voting and recent findings on agenda-setting. Electoral outcomes partially reflect societal trends and events over which parties have only limited control and to which they must react while campaigning, as publics are quick to reorient their political priorities according to the perceived urgent matters of the day while maintaining their deeper, more robust convictions. The same can be said of governance: ‘events, dear boy, events’.
Based on our findings, we offer two predictions: first, if exogenous trends and events continue to become more transnational, domestic party competition across countries will continue to become subject to trends that are increasingly synchronous, as has happened in Europe; and second, recent changes are potentially fleeting. If the public issue salience and performance evaluations change away from what has benefitted respective non-mainstream parties recently, these parties may lose out again in the issue competition.
Our article has several limitations, which future research should attempt to overcome. First, we only test three issues of many that are likely to have significantly affected recent changes in Europe’s electoral landscape and may or may not operate according to the theoretical framework that we offer (see also, Guinaudeau and Persico, Reference Guinaudeau and Persico2014, on the difficulties of conceptualising policy issues). Future research should consider testing further issues and attendant societal trends, in the case of our issues including irregular immigration or natural disasters (see McAllister and bin Oslan, Reference McAllister and bin Oslan2021). Indeed, the validity of our theoretical approach will certainly vary by issue and by context (Soroka, Reference Soroka2003; Kingdon and Stano, Reference Kingdon and Stano1984; Baumgartner and Jones, Reference Baumgartner and Jones1993), meaning that other issues with less obvious real-world-party politics linkages, such as European integration or abortion rights, should also be tested. Similarly, the specific context of the time-period in question may limit the generalisability of our findings: it may be the case that parties do not have a great degree of leeway to set the agenda during times of immediate crisis but during ‘normal’ times of politics they are less constrained and thus more able to affect the electorate’s priorities. Second, and related, we do not test the link between issues and parties but instead assume that it is via perceived competence in the short-term (at least for incumbent parties) and association over the longer term, via repeated emphasis (particularly for challenger parties). Indeed, though we mentioned opposition and government and challenger and established party dynamics, we do not systematically disentangle them empirically despite there being good reason to expect distinct dynamics as has been shown in the issue agenda literature. Issues are likely to vary in the extent to which parties—especially governing ones—can control their public salience—for example, domestic economic policy, unemployment and immigration rates are likely to be more controllable than terrorist attacks or international affairs. On the supply-side, future research should perform similar analyses using the credibility, ownership, and cues of parties on these and other issues (Zaller, Reference Zaller1992; Druckman et al., Reference Druckman, Fein and Leeper2012). We focus on issues associated with crises during the time-period under examination. Third, we do not incorporate cueing or framing into our analysis—i.e., the extent to which parties can affect public spatial consideration—but, aside from the agenda, parties may have agency through framing issues to their advantage (potentially combined with wave-riding). Future research may want to consider these dynamics. Fourth, though we may infer causality from our results given the relationships shown and temporal precedence, our parsimonious models means that there may still be omitted variable bias and issues of reverse causality—future studies should further interrogate these relationships.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1755773923000085.
Acknowledgements
This study was made possible by generous funding from the Leverhulme Trust (grant number: ECF-2021-342), the Swedish Research Council (grant number: 2019-00504) and the European Research Council (ERC_SYG_2018 grant no. 810356).
Open access
