A change of heart? Analysing stability and change in European party positions

Political parties face a challenging trade-off between responsiveness and ideological continuity. They are assumed to be adaptive organisations, continuously – and strategically – updating their platforms to relay the public’s evolving policy preferences and to maximise electoral support (Adams Citation2012; Laver Citation2005). Examples of such party shifts abound. Observers have noted, for example, how the populist radical right has moderated its economic position (De Lange Citation2007) to embrace the welfare chauvinism popular among its supporters (Attewell Citation2021; Magni Citation2020). By contrast, many mainstream right-wing parties have moved in the opposite direction, away from the ideological centre, on cultural issues (Meguid Citation2005) and European integration (Rohrschneider and Whitefield Citation2016) in response to the recent surge of these populist challengers. The British Conservative Party, notoriously divided on the European issue, has also grown more Euroskeptic over time. While Margaret Thatcher’s criticism of the European project once expedited her ouster as Prime Minister and party leader, it was exactly Boris Johnson’s uncompromising pledge to deliver Brexit that propelled him to power in the aftermath of the country’s referendum on EU membership.

However, notwithstanding the dynamic nature of representative democracy, there is ample reason to believe that parties cannot always move freely (Dalton Citation2016; De Sio and Weber Citation2014; Meyer Citation2013; Rovny Citation2015). Coherent, reliable, and distinctive party labels and ideologies are an integral part of representation as well (Mair Citation2008; McKelvey and Ordeshook Citation1986). They provide an important heuristic for vote choice (Downs Citation1957) and party identification (Campbell et al. Citation1960) in a context of uncertainty and limited information. Indeed, contrary to formal theory’s assumption of ‘costless spatial mobility’ (Adams et al. Citation2006), positional change is inherently risky, as it could tarnish a party’s reputation, compromise future coalition negotiations, and alienate voters and activists (Budge Citation1994; Somer-Topcu Citation2009; Tavits Citation2007).

How does a party balance these representative roles and their inherent trade-offs, and, relatedly, when is it more likely to alter its policy appeals? This study brings together insights from spatial models of elections and cleavage theory, two literatures that rarely speak to each other (Rohrschneider Citation2002), to develop a novel framework for understanding stability and change in European party positions. My starting point is the multidimensionality of the policy space in which parties compete (Bakker et al. Citation2012; Rovny and Polk Citation2019; Tavits and Potter Citation2015), which allows me to address the intimate relationship between position and salience (Meguid Citation2005; Wagner Citation2012).

Specifically, I argue that positional stability is likely observed on a dimension that is relatively salient to a party, while policy shifts are confined largely to issues outside of a party’s core agenda. Parties are rooted and invested in, even born out of, underlying social conflicts and come to represent specific issues and voter groups (Lipset and Rokkan Citation1967). As a result, it is especially on its salient, or primary, dimension that a party has a reputation to uphold and defend (Hooghe and Marks Citation2018), and where the inherent risks associated with positional shifts require a proper change of heart. Yet, by remaining ideologically true to its voters’ key concerns, a party will experience fewer constraints when strategically shifting on less salient, secondary issues (Elias et al. Citation2015).

Cross-sectional time series analysis of Chapel Hill Expert Survey (CHES) evaluations of party positions (Bakker et al. Citation2015; Polk et al. Citation2017), covering fourteen Western European countries from 1999 to 2019,Footnote¹ confirms that absolute positional change increases as an ideological dimension’s relative salience decreases – and vice versa. The effect is weaker for larger and governing parties, as their increased public scrutiny and executive responsibilities reduce the salience differential between the two dimensions. While the main empirical models employ a continuous ratio measure of economic and cultural dimensional salience based on manifesto data (Volkens et al. Citation2020), additional analyses show similar patterns in position change by party family when dichotomously identifying their respective primary and secondary dimension. Finally, multiple robustness checks, which use alternative model specifications and operationalisations of key variables, lend further support to these findings. This study focuses on Western Europe for data availability, but its results should travel to other established democracies with a history of programmatic politics.

The finding that policy positions are more flexible on a party’s secondary dimension provides an important behavioural prediction that bridges extant literature on party strategy, issue emphasis, political cleavages, and democratic representation, especially in an increasingly multidimensional environment. Above all, it contextualises our spatial assumptions of party strategy – which have been the basis of much of the party competition literature, but which are not always borne out – to reconcile the seeming tension between predicted positional change and observed ideological stability.

Party change and stability

The study of how political parties present themselves to the public, and compete for its vote, has a rich and impressive tradition in the field of comparative politics. Since Downs’ (Citation1957) classic work on the median voter theorem, political scientists have made major advances in understanding how a party locates itself in the political landscape and what factors lead it to alter its ideological position. Spatial models of elections – which view party competition through a positional lens, with parties offering alternative policy proposals to the electorate – have become increasingly sophisticated and comprehensive. Empirical studies now account for a wide range of characteristics, including a party’s office and policy-seeking behaviour, the multifacetedness of voter preferences, and the transformative nature of the policy space (e.g. Adams et al. Citation2006; Budge Citation1994; Hooghe et al. Citation2002; Kriesi et al. Citation2006; Meguid Citation2005).

The causes of party shifts remain contested, however. Which factors explain why a party moves and, equally important, with what magnitude? Extant research suggests that party positions are highly flexible and volatile, as parties respond to a plethora of endogenous and exogenous incentives, such as electoral defeat (Somer-Topcu Citation2009), governing status (Bawn and Somer-Topcu Citation2012), economic conditions (Fenzl Citation2018; Haupt Citation2010), and the policy moves of their competitors (Laver Citation2005). Although diverse, these theories share an implicit belief in the persistence of party change (Adams Citation2012).

For all the attention given to what makes parties move, relatively few studies have been devoted to what makes them stay put (but see Dalton Citation2016; Meyer Citation2013). Yet, it is often acknowledged that position change is inherently risky (Budge Citation1994; Somer-Topcu Citation2009; Tavits Citation2007). Even Downs acknowledged that parties experience serious limitations on their room to manoeuvre, as shifting too much and too often would undermine their ideological integrity and responsibility (Citation1957: 122).

To better understand the constraints on party strategy, this study builds on cleavage theory, which is based on the intuition that party systems reflect the underlying social conflicts in society, shaped – and upset – only by infrequent critical historical junctures. Although not addressing party competition directly (Hooghe and Marks Citation2018) – at least not in the way we typically think of it, as a positional dynamic of action and reaction in a competitive environment – this framework nevertheless raises two valuable insights for the study of party strategic behaviour.

First, it sheds light on the relative durability of party-voter commitments. In their pivotal work, Lipset and Rokkan (Citation1967) trace the origins of party system development to persistent tensions in society between religious, socio-economic, and political groups. These so-called cleavages create solidary communities and lead to the emergence of political parties to express their respective collective interests. The very nature of political conflict then further reinforces and perpetuates an ideological divide’s group identities (Bornschier Citation2010). From this perspective, a party thus enjoys strong linkages with its supporters based on a meaningful and lasting ideological reputation.Footnote² And we now know that such reputations can be strengthened or revived over time, e.g. through a party’s performance in government, and crucially shape electoral competition (Bélanger and Meguid Citation2008; Budge and Farlie Citation1983). In the words of Lipset and Rokkan, ‘parties do not simply present themselves de novo to the citizen at each election; they each have a history and so have the constellations of alternatives they present the electorate’ (Lipset and Rokkan Citation1967: 2).

Second, although not all of Lipset and Rokkan’s divisions proved equally resilient, or shaped national politics to the same degree, cleavage theory directs our attention to the underlying, multidimensional structure of the political space in Western European democracies (De Vries and Marks Citation2012). And while many studies of party competition today still rely on a unidimensional framework (e.g. Adams et al. Citation2006; Ezrow et al. Citation2011), scholars increasingly assume that a second dimension now structures European politics in addition to an economic left-right divide. This second dimension goes by different labels, but typically encompasses socio-cultural issues, including (though not limited to) immigration, law and order, and environmental politics (Hooghe et al. Citation2002; Kitschelt Citation1994; Kriesi et al. Citation2006).Footnote³ There is thus a growing understanding that ‘multidimensionality cannot be evaded in political practice’ (De Sio and Weber Citation2014: 870).

The above discussion suggests that, in contrast to the strategic interactions in a typical Downsian world, party-voter linkages can be durable and multidimensional. It highlights the causal factors that foster stability in and constraint on a party’s policy platform over multiple election cycles. Whether it is by virtue of deeply-rooted, cross-cutting societal divisions or less distant track records in office, ideological reputations matter and are an integral part of political contestation.

Research design

The above hypotheses are tested using pooled data of party positions from 1999 to 2019, measured by the Chapel Hill Expert Survey (Bakker et al. Citation2015; Polk et al. Citation2017), which provides information on the ideological positions of almost one hundred parties on an economic and a cultural dimension across fourteen Western European countriesFootnote⁴: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Portugal, Spain, Sweden, and the United Kingdom.Footnote⁵

The validity and reliability of party position data have been debated extensively. This study uses expert evaluations because, unfortunately, existing public opinion surveys do not gauge voter perceptions of parties’ multidimensional positions over time. But expert placements have been rigorously tested, cross-validated, and shown to perform as well as, if not better than, other measures, including those based on mass surveys and manifesto data (Marks et al. Citation2007; Steenbergen and Marks Citation2007). Moreover, expert and voter data of party positions are strongly correlated (Dalton and McAllister Citation2015), which suggests that the position shifts estimated here are both reliable and likely picked up by voters. It could be argued that experts are more prone to detect continuity in party platforms, but this is not a concern for this study, as it analyzes positional stability as well as change. By contrast, manifesto data, despite being widely viewed as a valid measure of salience (see below), have received criticism for conflating issue importance with position (Gemenis Citation2013). Nevertheless, various sensitivity analyses confidently show that the main findings of this study are not dependent on any particular measure or data source (see Online appendices 6–10).

I use the absolute difference in the average expert evaluation of a party’s position between waves, both on the economic and the cultural dimension, because my theoretical argument is about ideological flexibility, not direction.Footnote⁶ That is, the greater a party’s ideological shift from its previous position on a dimension, the higher its score on the dependent variable – regardless of the direction of change. Separate models are specified for each dimension. The histograms for the dependent variables show that, despite their substantive and historical differences, party change is relatively similar across the two dimensions (see Online appendix 3).

Analysis

This study employs a cross-sectional time series analysis with party-wave as the unit of observation. The dependent variables are first-differenced, tracking party change on each dimension from one survey year to the next, to analyse the direct effect of a dimension’s relative salience level on Y. The generalised model can be specified as follows:|Δ position(i,t)|=α+β1 salience(i,t)+β2 covariates(i,t)+β3 interactions(i,t)+ϵ(i,t)where the dependent variable is the absolute positional shift of party i on either the economic or the cultural dimension at t compared to t₋₁; the key independent variable is a party’s relative dimensional salience at time t (measured at the most recent national election); the covariates include a party’s vote share, government participation, age, leadership dominance, ideological extremism, and the effective number of parties, all of which vary over time, with the exception of party organisation; and the interaction terms model the effect of relative salience on position change by electoral strength and government status, respectively. Note that while the dependent variables are differenced, the main independent variable is not. The model includes the level of a dimension’s salience (t) to explain party change (t – t₋₁), not the difference in salience between t and t₋₁. After all, a party is expected to move more on a dimension that is less salient, even when the relative salience of that dimension does not change.Footnote⁹

Tests show that the residuals at time t₋₁ are a good predictor of the residuals at t, indicating a problem with serial correlation, specifically of the panel-specific AR(1) type. Instead of using the traditional correction that leads one to include a lagged dependent variable on the right-hand side, a Prais-Winsten solution is used, as the former approach has been shown to produce bias and wash out the effect of the variables that are of theoretical interest (Plümper et al. Citation2005).Footnote¹⁰ Unexpectedly, given the multitude of parties, nested within countries, the data also show panel-level heteroscedasticity, so corrected standard errors are used.

Table 1 confirms this study’s main hypothesis that the relative salience of an ideological dimension predicts a party’s positional shifts. Models 1 and 2 – which report the results without and with the interaction terms, respectively – show that the more salient the economic dimension, the less likely a party is to change position on that dimension. On the cultural dimension, by contrast, this finding is reversed (Models 3 and 4): as the salience ratio increases, and the cultural dimension thus becomes relatively less important than the economy, position change is more likely. The relevant coefficient estimates are statistically significant across the board, but do have smaller magnitudes on the cultural dimension, which likely reflects its inherent complexity compared to economic left-right (Rovny and Polk Citation2019). It combines a diverse set of policy issues on which parties could take inconsistent positions. Populist radical right parties and voters, for example, increasingly combine largely conservative attitudes with more modern, progressive views on morality issues (Lancaster Citation2020), which would produce a more muddied image of dimensional position change.

Moving on to the covariates, government parties are less likely to change position, which suggests that they are constrained by their executive functions and do not try to upend the status quo. As expected, the positive estimate for leadership-dominated parties is particularly strong and conforms to extant work on the importance of party organisation for party strategy (e.g. Schumacher and Giger Citation2018). The statistical significance and sign of the other variables – vote share, party age, extremism, and the effective number of parties – vary by dimension. This could be an indication that there are substantial differences between those parties for whom the economic dimension is more salient and those who primarily emphasise cultural issues (see also Tavits and Potter Citation2015). While the bivariate correlations between these variables are low (see Online appendix 5), and new parties are thus not necessarily smaller or more radical, this is nevertheless an important observation that will be explored in more detail below.

The results reported in Table 1 offer a first indication of the relationships between the included variables, but the preferred way to interpret especially the interaction terms is by means of marginal effects plots. Not only does this allow us to examine directly the conditional relationship of relative dimensional salience on party position change at different values of the moderating variables, the greater magnitude of relative salience’s estimates in Models 2 and 4 (compared to the baseline specifications) indicate that we need to account for these interactions to fully understand their intricate relationships.

To reiterate, the expectation is that electoral strength and government participation weaken the relationship between relative salience and positional flexibility. Figure 2 confirms this prediction with regard to vote share. On the economic dimension, the marginal effect of relative salience is negative as before, but its magnitude declines as a party’s electoral strength increases.Footnote¹¹ In fact, the effect is no longer statistically significant for parties with a vote share greater than 15%, when the 95% confidence interval includes the null. Similarly, the positive coefficient estimate for relative salience on the cultural dimension decreases in size, albeit slightly, before becoming statistically indistinguishable from zero for larger parties (around 13%). Thus, for electorally strong parties, which have broader issue agendas that encompass both ideological dimensions, relative salience no longer predicts party change. Nevertheless, given that about two thirds of the included party-wave observations fall below these electoral thresholds, the reported relationships remain meaningful and valid for the vast majority of parties.

Next, Figure 3 plots the marginal effect of relative salience by government status.Footnote¹² As expected, on both dimensions, the estimates converge on zero once a party enters government. On the economic dimension, the marginal effect loses half of its magnitude compared to opposition parties, whereas on the cultural dimension it is no longer statistically significant. Enjoying government status thus comes at a cost. When taking on executive responsibilities, a party is forced to address a wide variety of issues, including on its less salient dimension, and it concedes at least partial control over its own agenda. This constrains its behaviour and limits its ability to adopt tailored strategies on individual dimensions.

Taken together, these findings provide strong and consistent support for the main hypotheses of this study. And the magnitude of the presented effect sizes is substantively meaningful, individually approximating up to one third of a standard deviation of the dependent variables. To further interpret the results, consider the following illustrative examples. Between 2006 and 2010, the Dutch Partij voor de Vrijheid moved from a strongly rightist position to the near centre on the economic dimension – a shift far bigger than its cumulative position changes over multiple waves on its more salient cultural dimension (2006–2019). In the midst of Europe’s financial crisis, and presumably knowing that economic attitudes do predict its electoral support (Attewell Citation2021), this was likely a strategic decision – one also adopted by other radical right parties. For instance, similarly large economic shifts were made by Front National (France) and Freiheitliche Partei Österreichs (Austria), parties for whom the cultural dimension was considerably more salient than the rest of the party system. By contrast, the smallest positional changes between waves on the economic dimension were made either by established parties like Partido Social Democrata (Portugal) and Partido Popular (Spain), outspoken left-wing parties like Enhedslisten (Denmark) and Kommounistikó Kómma Elládas (Greece), or both, e.g. in the case of Socialdemokraterna (Sweden) and Partij van de Arbeid (the Netherlands).

Multiple robustness checks produce strongly similar results and further corroborate the above findings (see Online appendices 6–10). The negative relationship between relative dimensional salience and party position change is confirmed in additional tests that include: (1) a dependent variable which uses positional data derived from the manifesto project, specifically Lowe et al.’s (Citation2011) log transformation; (2) an alternative, expert-level evaluation of economic and cultural salience, the key independent variable, as reported by CHES; (3) a combined ratio measure of position change similar to the operationalisation of relative salience; (4) an examination of long-term party change over more than one wave, i.e. t compared to t₋₂, to rule out that the results are driven by short-term zigzagging; (5) a lagged measure of vote share to confirm that it gauges a party’s electoral strength, not the appeal of its most recent positional shift; (6) country and survey wave fixed effects and (7) an analysis of positional shifts on individual economic and cultural policy issues. The latter suggests that salience similarly affects party change at lower levels of aggregation.

Discussion

Every political party faces cross-pressures to respond to citizens’ evolving preferences while simultaneously maintaining ideological continuity. To evaluate representative democracy in Western Europe, it is important to analyse when and how parties alter and adapt their policy platforms. This study presents a novel lens to better understand how these demands shape party behaviour. It brings together insights from spatial models of elections (e.g. Adams Citation2012; Downs Citation1957) and cleavage theory (e.g. Lipset and Rokkan Citation1967; Hooghe and Marks Citation2018) to explain stability and change in parties’ multidimensional ideological positions. By embracing the multidimensionality of contemporary politics in Western Europe, this study provides a set of predictions about where the strategic actions of parties, as described by spatial modellers, are more likely to be observed.

A party faces constraints especially when shifting its established agenda (Hooghe and Marks Citation2018; Meyer Citation2013), as its ongoing commitments to its core agenda make change undesirable and risky. This leads them to actively engage with their secondary dimension (Elias et al. Citation2015). Conversely, strategic policy shifts, which have been the focus of much of the party competition literature (see Adams Citation2012), are particularly likely on issues less salient to a party. Certainly, parties can and do change position on their core dimension (see, e.g. Adams et al. Citation2006; Ezrow et al. Citation2011), but such (potentially high profile) policy shifts are relatively less common and auspicious. The reported findings convincingly confirm this relationship between dimensional salience and party change, both when treating dimensional salience continuously and dichotomously, as a binary distinction between a party’s primary and secondary dimension. Furthermore, the results are robust to multiple alternative operationalisations.

These findings have several important implications and contribute, directly or indirectly, to a variety of literatures. In addition to those already mentioned – spatial modelling and cleavage theory – its findings relate to extant research on, among others, salience (Bélanger and Meguid Citation2008), issue ownership (Budge and Farlie Citation1983; Petrocik Citation1996) and entrepreneurship (Hobolt and De Vries Citation2015) and niche parties (Meguid Citation2005).

This study also makes a methodological contribution to an ongoing debate about the measurement of parties’ programmatic appeals (Marks et al. Citation2007; Steenbergen and Marks Citation2007). By integrating both expert surveys and manifesto data, it provides one way in which the respective strengths of each methodology can be optimised. It further develops a novel measure of relative dimensional salience that can be used in future research on party strategy.

Additional research is needed, however. For instance, how does public opinion shape the salience of an ideological divide? And do voters pick up on parties’ positional shifts, large or small, on both dimensions? While this study is not the first to establish the connection between position and salience, it provides a new avenue to analyse their intricate relationship. Yes, a party might emphasise the issues popular among both its supporters and the general public to ‘yield’ greater voter support (De Sio and Weber Citation2014) – and mute those that could harm its electoral fortunes (Rovny Citation2012, Citation2013) – but the appeal of the theoretical argument presented here is that it does not do away with the positional dynamics of party competition altogether. Position change is risky, but that does not mean that a party will accept the status quo, especially on issues outside of its core agenda. Rather, altering one’s positional appeals could change its electoral fate. As such, this perspective can better explain how party strategies might vary over time, as parties respond to the changing preferences of both their voters and their competitors.

Crucially, although the relationship between positional flexibility and dimensional salience is now well established, no theoretical argument was presented about the direction of these policy shifts. Do parties use outside issues to amplify their ideological distinctiveness or do they pursue a median voter strategy on their secondary dimension in an attempt to appeal to more centrist voters? They could, for instance, employ policy-based cues to chase undecided swing voters, as opposed to the emotional appeals used to mobilise loyal supporters (Rohrschneider Citation2002). Or perhaps they converge on the position of their main rival? Indeed, we have seen that mainstream right parties, such as the conservatives and Christian democrats, have done exactly that by approaching radical right challengers on the cultural dimension. Regardless, while some have concluded that policy shifts are generally more appropriate on the economic dimension (Tavits Citation2007), this study suggests that what is deemed pragmatic or principled varies by party. We do not have all the answers yet – and better data on particularly voters’ multidimensional attitudes are much needed – but this study provides a first step in a broader research agenda on multidimensional representation.