Left-Authoritarians and Policy Representation in Western Europe: Electoral Choice across Ideological Dimensions

Abstract

Citizens can face a difficult electoral decision when no party even broadly represents their views. In Western Europe, this applies to those citizens with left-wing preferences on economic issues and traditional/authoritarian preferences on socio-cultural issues. There are many voters with such ‘left-authoritarian’ views, but few parties. Hence, the former often have to choose between parties that only match their views on one of these two ideological dimensions. This study shows that whether these citizens privilege economic or socio-cultural congruence in their electoral preferences depends on the issues they are concerned about. In general, it is found that left-authoritarians privilege economic concerns and therefore prefer parties that are left-liberal. These findings have implications for our general understanding of electoral choice and of changing patterns of political competition in Western Europe.

Normative models of democratic representation assume that voters want to elect those representatives who best reflect their views, attitudes and preferences (e.g. APSA 1950; Mansbridge 2009; Schmitt and Thomassen 1999). This is obviously easier to achieve if there are representatives running for office who closely match citizens’ views. For voters to be able to pick congruent representatives, the menu of viable parties should therefore encompass, at a minimum, the most frequent combinations of views among citizens. Yet this is not always the case: sometimes there is simply no party that articulates voters’ views on key policy issues (e.g. Lefkofridi and Horvath 2012; van der Brug and van Spanje 2009).

This article focuses on a significant segment of the West European population who often face this problem: left-authoritarians. These are citizens who blend left-wing economic with traditional/authoritarian socio-cultural views. They rarely have a party that is congruent with them on both ideological dimensions (Thomassen 2012). Instead, parties tend to either combine economically left-wing with socio-culturally liberal views, or they take an economically right-wing and a socio-culturally authoritarian stance (van der Brug and van Spanje 2009). Left-authoritarians are thus faced with a situation where no option in the party system reflects their political opinions even if we reduce the policy space to just two broad dimensions. Surprisingly, this segment of European electorates has so far received relatively little scholarly attention.¹

Our analysis of cross-national individual-level data collected prior to the peak of the current financial crisis by the European Election Study/EES (van Egmond et al. 2011) confirms that the group of citizens holding this particular ‘opinion package’ has a consistent presence and considerable size across Western Europe.² In contrast, the 2006 Chapel Hill Expert Survey of party positions (Hooghe et al. 2010) shows that there are few (if any) parties in Western Europe that defend a similar package of opinions (corroborating the findings of van der Brug and van Spanje 2009). Compared to other simple packages of views, left-authoritarian attitudes are consistently and strikingly unrepresented by any party.

This article extends previous research on left-authoritarian citizens by conducting an analysis of how such voters take electoral decisions. The absence of a ‘good match’ for left-authoritarians’ ideological preferences on both economic and socio-cultural dimensions has important consequences for their decision-making at the ballot box. Unlike voters who combine views in ways that fit the policy combinations offered by political parties, left-authoritarians regularly face a choice between parties that represent either their economic views or their socio-cultural views, but not both. In other words, these voters are attitudinally cross-pressured (Brader et al. 2013).³

We argue and demonstrate empirically that the party preferences of left-authoritarian voters are shaped by their level of concern about the economy and immigration. The focus of this article is therefore on how individual-level issue concerns shape the way in which policy distance influences decision-making at the ballot box. So, left-authoritarians prefer parties that defend similar economic views if they are concerned about the economic situation, but they prefer parties that are close to them on the socio-cultural dimension if they are worried about immigration. Thus, we present evidence that voter-level concern for each dimension is central to determining how left-authoritarians choose a party to support. We show that left-authoritarians generally privilege economic over socio-cultural congruence.

This study of left-authoritarians has important implications for research on electoral behaviour and on citizens’ representation via elections and political parties. First, our findings are theoretically relevant because they uncover how voters choose between parties in situations where there is no fully congruent option. Building on the memory model of information processing, we argue that the use of policy distance to assess the attractiveness of parties depends on the accessibility of these dimensions to voters (Iyengar and Kinder 1987; Miller and Krosnick 1996; Scheufele 2000; Zaller 1992; Zaller and Feldman 1992). The standards by which we judge parties can change depending on which issues and problems currently concern us most. Our findings are thus related to, and add to, the recent literature highlighting the importance of salience to issue voting (e.g. Belanger and Meguid 2008; de Vries 2007; Green and Hobolt 2008; Singer 2011). While we concentrate on a particular segment of the European population, our theory is therefore broadly applicable as well.

Second, our findings are empirically important because they examine a section of European voters that is consistently unrepresented by any one party but crucial to contemporary West European politics. This may be particularly true at the current time, as popular dissatisfaction with economic conditions in Europe is coupled with an increase in anti-immigration sentiment. Arguably, left-authoritarian attitudes capture the current mood well and are likely to remain electorally important for some time to come. Our findings also highlight when these voters might choose to support radical-right parties; thus, our findings generalise the argument that support for such parties is greater among voters who take electoral decisions based on their socio-cultural concerns (Lubbers et al. 2002; Norris 2005; Oesch 2008; Rydgren 2008).

In the remainder of the article we proceed as follows: we begin by discussing our assumptions about the structure of the political space and our expectations about the electoral choices of individuals holding left-authoritarian views. Next, we detail our methodological approach and summarise our descriptive findings on the general absence of West European parties and the presence of many West European voters with a clearly left-authoritarian stance. We then illustrate the role of issue salience in conditioning the degree of attraction between left-authoritarians and left-liberal or right-authoritarian parties. We conclude by summarising our results and discussing their implications for representation in contemporary Europe and pose questions for further research.

Conceptualising Political Conflict in Western Europe

The fundamental assumption behind our study is that the political ideologies of West European voters are well-summarised by an economic and a socio-cultural dimension (Kriesi et al. 2008; van der Brug and van Spanje 2009). Similar divisions have been suggested by Inglehart (1977), Finer (1987), Flanagan (1987), Kitschelt (1994) and Hooghe et al. (2010). In all these depictions of political conflict, a mainly economic group of topics is seen as separate from a mainly social/socio-cultural group of topics. Kriesi et al. (2008) as well as van der Brug and van Spanje (2009) maintain that among voters these two dimensions are orthogonal, creating a two-dimensional attitude space. Based on this argument, we identify four basic combinations of attitudes or packages of views: left-authoritarian, right-authoritarian, left-liberal and right-liberal (see Figure 1). The first word in these hyphenated terms refers to the economic dimension (‘left/right’) and the second to the socio-cultural dimension (‘liberal/authoritarian’, slightly amending Kitschelt [1994]). In comparison to the conventional uni-dimensional depiction of the left–right continuum, left/right positions on these two dimensions can be freely combined (Finer 1987). In other words, a voter on the economic right need not be on the socio-cultural right. Compared to the general left–right scheme, this two-dimensional conceptualisation of the political space allows for a more nuanced (yet still parsimonious) description of voters’ policy preferences.

Figure 1 Ideological dimensions and the four basic opinion packages

As shown by van der Brug and van Spanje (2009) and Thomassen (2012) and confirmed by our analysis below, there are many voters in all four sections of the two-dimensional space. Nonetheless, most parties either combine socio-culturally authoritarian with economically right-wing attitudes or socio-cultural libertarian with economically left-wing attitudes. This leads to a gap between many voters’ views and the policy packages on offer by parties: although a considerable number of voters hold left-authoritarian views, there are generally few left-authoritarian parties.

Theorising the Electoral Behaviour of Left-Authoritarian Voters

We argue that this mismatch between voter opinions and party policy packages has consequences for the electoral preferences of left-authoritarian voters. On the one hand, there are left-wing parties with which left-authoritarian voters are ideologically congruent on the economic dimension; on the other hand, there are authoritarian parties that advocate their preferences on the socio-cultural dimension. As these voters thus generally have to choose between congruence on the economic and congruence on the socio-cultural dimension, we can say that they are, in essence, cross-pressured between two ideological dimensions. Hence, we ask: how do these voters take electoral decisions? Which dimension do they privilege in their pursuit of policy representation, the economic or the socio-cultural?

At the outset, we assume that voters will prefer parties that are as congruent with their views as possible overall. This is the basic Downsian (1957) approach that argues that policy distance matters to voters. In a uni-dimensional space, this proposition in its very basic form means that a voter will calculate the smallest difference between herself and each of the parties competing. It is less straightforward to calculate policy distances in a two-dimensional space (Humphreys and Laver 2010). It is nevertheless reasonable to assume that, if presented with two parties with the same socio-cultural position, voters should prefer the party that is closer to them on the economic dimension, and vice versa.

However, voters who are attitudinally cross-pressured will face a more difficult choice, namely between parties that represent them well on only one ideological dimension. For example, some parties will be congruent on the economic dimension and incongruent on the socio-cultural dimension, while others will be congruent on the socio-cultural dimension and incongruent on the economic dimension. We argue that voters solve this dilemma by prioritising congruence on one of these dimensions. There is already some empirical evidence for this; for example, the appeal of the radical right depends on voters’ willingness to grant pre-eminence to socio-cultural over economic issues (Ivarsflaten 2005). Thus, voters do not necessarily weight distances on the two dimensions equally.

One appealing way of understanding the weight voters will place on each dimension is by reference to the theories on accessibility and salience, which are based on the memory model of information processing (Iyengar and Kinder 1987; Scheufele 2000; Scheufele and Tewksbury 2006; Zaller 1992; Zaller and Feldman 1992). In this model, individuals store in their memory a variety of considerations relevant to their attitudes and beliefs. However, these considerations are not equally accessible and easily retrievable (Hastie and Park 1986; Krosnick 1990; Scheufele 2000). This is important because the more accessible issues are more likely to shape political behaviour (Belanger and Meguid 2008; Bonninger et al. 1995; Green and Hobolt 2008; Krosnick 1988; RePass 1971). So, a voter’s current attitude will be determined by the considerations that are most accessible (Zaller 1992; Zaller and Feldman 1992). The aspects on which we judge certain objects – including parties and politicians – thus depend in part on the accessibility of these key considerations by the individual (Iyengar and Kinder 1987). Therefore, how voters differ in the accessibility of key issues has consequences for how they form attitudes and take political decisions.

In this article, we argue that voters will prioritise congruence on that dimension which is more accessible to them (Zaller and Feldman 1992). Specifically, we argue that the accessibility of a dimension will depend on the extent to which a voter is concerned about the issues that underlie it. Individuals who believe that the country is heading in the wrong direction in a given area will see this issue dimension as more important, and proximity on this dimension will in turn influence electoral preferences more.

Applying this theory to our two-dimensional space, we expect individuals to place more weight on congruence on the dimension they are more concerned about. Thus, the more individuals are concerned about the economy, the more important the economic positions of the parties should be to them. In turn, the parties’ socio-cultural positions should become less important. Other individuals who are more concerned with socio-cultural matters should more easily retrieve and thus base their preferences on the socio-cultural positions of the parties, with their economic positions assuming less importance.

We therefore argue that an individual’s concern about an issue dimension determines that dimension’s accessibility and thus its weight in preference formation. The question of which factors affect whether a voter becomes concerned about a particular issue is beyond the scope of this article. It is plausible that the influence of the media and party campaigns through priming is significant (Iyengar and Kinder 1987; Krosnick and Kinder 1990). Negativity bias may also mean that it is easier for parties and the media to increase concern and worry rather than optimism and hope (Baumeister et al. 2001; Rozin and Royzman 2001; Soroka 2006).

To sum up, left-authoritarians’ concerns about each issue dimension may determine the weight of congruence on the economic and on the socio-cultural dimension in determining their electoral preferences. Therefore, we hypothesise that the electoral preferences of left-authoritarian voters are affected by party proximity on the two ideological dimensions, but that this impact is conditional on whether the individual is concerned about this issue dimension. More specifically, left-authoritarians concerned about economic issues will place more weight on proximity on the economic dimension and less weight on proximity on the socio-cultural dimension. In turn, left-authoritarians concerned about socio-cultural issues will place more weight on proximity on the socio-cultural dimension and less weight on proximity on the economic dimension.

In the next section we elaborate on how we operationalise these concepts in order to build a statistical model and test this proposition.

Data and Methodology

We will now present our measurement approach and the data we use. We will begin by describing how we estimate voter and party positions on the two dimensions and then discuss how we explain left-authoritarians’ electoral choices.

Voter and Party Positions

We begin by plotting voters and parties in a two-dimensional space; this enables us to confirm that there is a gap between ‘opinion packages’ at the voter and party levels in Western Europe. We thus first examine individual-level attitudes regarding the economic and socio-cultural dimensions using the EES dataset (van Egmond et al. 2011), which contains a good collection of relevant attitude items.⁴ We measure views on the two dimensions by constructing an index that includes specific issue items on economic and socio-cultural attitudes (see Appendix 1). Following Ansolabehere et al. (2008, see also Heath et al. 1994), we decide a priori which measures are linked to which scale (see also Benoit and Laver 2012).⁵ For economic matters, we choose all questions relating to the role of the state in the economy. For the socio-cultural dimension we include items that relate to immigration, respect for authority, and law and order (Flanagan and Lee 2003; Rydgren 2007).⁶ After re-scaling all items so that higher values indicate a more right-wing or authoritarian attitude, we calculate the mean response for every individual. At r = – 0.01, the correlation between the two independently constructed scales is very low, so there is evidence that the two attitudinal dimensions are indeed orthogonal among voters.⁷

Using these two attitudinal dimensions, we identify four fundamental groups: left-authoritarians, right-authoritarians, left-liberals and right-liberals. To qualify for any of these categories, voters should unambiguously hold such a package of attitudes; consequently, we rule out respondents who hold ambiguous or mixed views on either dimension.⁸

In the EES data, left-authoritarian views are common across all countries, as also shown by van der Brug and van Spanje (2009). In Table 1, we present the share of respondents we classify as left-authoritarian in each country. Greece, the UK, the Netherlands and Finland have a particularly large proportion of such respondents, while Germany and Denmark have a relatively low proportion. Across all countries, left-authoritarians are on average the second-largest group among those with a clear opinion on both dimensions: 22.3 per cent of the respondents in our sample are left-authoritarians, while 10.4 per cent are left-liberals, 22.8 per cent right-authoritarians and 7 per cent right-liberals (see Appendix 3 for more details).

Table 1. Proportion of left-authoritarian citizens by country

Country	Proportion of left-authoritarians (%)
Austria	13.8
Belgium (Flanders)	20.4
Belgium (Wallonia)	20.1
Denmark	7.0
Finland	27.8
France	23.1
Germany	8.8
Greece	36.2
Ireland	22.3
Italy	25.2
Netherlands	31.6
Portugal	25.7
Spain	25.7
Sweden	15.9
UK	28.4

Note: See Appendix 1 for questions used to calculate indicators.Source: EES 2009 (van Egmond et al. 2011).

We then turn to the 2006 Chapel Hill expert survey (Hooghe et al. 2010; Steenbergen and Marks 2007) to locate parties in the same two-dimensional space.⁹ We place specific parties on our two dimensions of conflict by using two indexes of economic and socio-cultural positions, respectively. Each index is made up of three items. Specifically, the economic index is composed of items on redistribution, deregulation, and improving public services versus reducing taxes, while the socio-cultural index is made up of items on immigration, multiculturalism and civil liberties (for details, see Appendix 2).¹⁰ To enhance the ease of comparability with the individual-level data, we recode both variables to range from 1 to 5 (instead of 0 to 10). Again, we use cut-off points to establish which parties can be seen as having neutral or centrist views.¹¹

Our results regarding the positions of parties confirm the findings of van der Brug and van Spanje (2009). Thus, there are very few parties in off-diagonal positions, i.e. in the left-authoritarian or the right-liberal sections. The other two main sections (left-liberalism and right-authoritarianism) usually contain parties. Most strikingly, there is not a single party that clearly falls into the left-authoritarian camp, although a considerable proportion of the electorate holds such views. The only exception is the Social Democrats in Denmark, which can almost be classified as a left-authoritarian party.¹² For full details on the positions of political parties in our sample of countries, see Appendix 4.

Explaining Left-Authoritarians’ Electoral Choices

This article’s aim is to understand the party preferences of left-authoritarians. To examine factors that affect the degree to which voters are attracted to a party in a cross-national setting, we use propensity-to-vote (ptv) scores as our outcome variable (van der Eijk and Franklin 1996; van der Eijk et al. 2006). An individual’s propensity to vote for a party is assessed by asking respondents how probable it is that that they would ever vote for each of a series of parties, using a scale from 0 to 10. This means that all survey respondents assess how electorally attractive each party is to them.

Our first key independent variable for explaining left-authoritarians’ degree of attraction to parties is their distance on the economic and socio-cultural dimensions. This is measured using the economic and socio-cultural indexes calculated based on the Chapel Hill expert survey.¹³ As mentioned above, the scale ranges from 1 to 5. We calculate policy distance on each dimension by calculating the absolute distance between the voter and each party. The maximum possible value of this variable is thus 4 (complete incongruence). The minimum is 0 (complete congruence).

Our core argument is that left-authoritarian voters will weight this policy distance differently depending on their current concerns. To measure this, we turn to a series of evaluations included in the EES (van Egmond et al. 2011).¹⁴ First, we code whether the respondent is concerned about the economic situation. These are respondents whose average assessment of the economy is negative, i.e. if the average of responses to the retrospective and prospective economic perceptions questions is worse than ‘stay the same’.¹⁵ This gives us a variable that is 1 if the respondent is concerned about the economy (i.e. believes it is getting worse) and 0 if not. Respondents who answer ‘don’t know’ or who refused to answer are also classified as 0.

Second, we code whether the respondent is concerned about the level of immigration. To capture this, we use two questions that ask whether the number of immigrants has increased and whether this is a good or a bad thing.¹⁶ So, respondents are coded as ‘concerned’ if they answer that immigration has increased and if they think this is a bad thing. In other words, we do not treat respondents as concerned if they have a negative view of immigration but think it is decreasing, as such individuals should not be as concerned about the issue. Respondents who think immigration is good (and increased or decreased), who answer ‘don’t know’ or who refused to answer are also coded as ‘not concerned’. This gives us a variable that is 1 if the respondent is concerned about immigration and 0 if not.

In our sample of countries and voters, about 43 per cent of left-authoritarians are coded as concerned about immigration. This ranges from 7 per cent of left-authoritarians in Germany to 79 per cent in Greece; 67 per cent of left-authoritarians are coded as concerned about the economy, with the lowest level of concern in Denmark (48 per cent) and the highest in Ireland (85 per cent).

An alternative measure of salience would have been responses to the questions asking for the ‘most important problem’. However, these questions are problematic (Johns 2010; Wlezien 2005). Nevertheless, we carried out robustness checks using these questions and present the results in the supplemental materials (Appendix 6). The results are largely consistent with those reported below.

To analyse ptv assessments, we stack our dataset so that each case is an individual × party unit; for details on this process, see for example Pardos-Prado and Dinas (2010). The structure of this transformed data is best described as hierarchical, with one lower level and two crossed higher levels: individual × party units nested in (1) individuals and (2) parties. Because of this, and due to the fact that the stacking procedure artificially increases the number of observations and may cause concerns over the independence of errors (Pardos-Prado and Dinas 2010), we run a hierarchical multiple linear regression with the ptv scores as the lower level and individuals and parties as crossed upper levels. Random intercepts for individuals account for unexplained individual-level differences in the average ptv score and for the potential violation of the assumption of independent errors across cases (Pardos-Prado and Dinas 2010). By including parties as a further second level, we also allow for party-specific differences in the mean ptv score for left-authoritarians. Our analyses are run using the xtmixed command in Stata 11. Since our goal is to explain the vote choices of left-authoritarians, we restrict our analyses to these voters; membership of this group is assessed using voter attitudes, as described above.

It is possible to control for further respondent-specific variables. Below, we therefore also present results when controlling for two very strong predictors of party choice: party identification and whether the party is seen as best at handling the most important problem. These can be easily stacked along with the ptvs, the outcome variable (see van der Eijk et al. 2006). Both additional controls are, of course, potentially endogenous to responses for vote choice, especially in a cross-sectional survey context. Finally, it is possible to include linear transformations of socio-demographic controls in stacked vote-choice models (for details see van der Eijk et al. 2006). While this means that their effects are no longer directly interpretable, this method allows us to check the robustness of our findings. Below, we therefore present additional models that include transformed versions of four socio-demographic characteristics: age, gender, religiosity and economic status (worker vs. non-worker).

Results

We begin our analysis of left-authoritarians’ voting preferences by providing a descriptive table of their voting intentions (Figure 2). This is operationalised using the EES question asking respondents which party they would vote for if an election were held on the following Sunday. The numbers shown are the percentage of left-authoritarians in each country voting for parties in each ideological section.

Figure 2 Distribution of electoral choices of left-authoritarian respondents in a two-dimensional policy space Note: cell entries are unweighted percentages of left-authoritarian voters’ party choice in each cell as their current voting intention. Source: EES 2009 (van Egmond et al. 2011).

There is no overall pattern in the type of party left-authoritarian voters prefer, that is, whether they privilege their socio-cultural or economic congruence. Indeed, aggregate voter behaviour very much differs by country. In some political systems, left-liberal parties are clearly favoured, for example Germany, Spain, Sweden and Wallonia. In others, right-authoritarian parties are more likely to get this group’s vote, with the most prominent cases being Flanders and the United Kingdom. Part of the explanation for this may of course lie in supply-side differences between countries, that is, the kinds of parties competing in each system.

Yet, in general, what is noticeable is that left-authoritarian voters in most countries tend to split their vote quite equally between the more left-liberal and the more right-authoritarian options. In other words, it is not the case that left-authoritarians generally prefer parties that are, for example, economically congruent with them. Instead, these voters sometimes opt for left-liberal and sometimes for right-authoritarian parties. Hence, the obvious question is what determines whether ‘cross-pressured’ left-authoritarian voters prefer economic or socio-cultural congruence. When do they privilege their opinions on the economic dimension, and when do they choose based on their views on the socio-cultural dimension?

To answer these questions, we turn to the multivariate analysis of ptv scores. The outcome variable is the ptv score for each party, as described above. Model 1 (Table 2) presents a simple analysis using just policy distance on the two dimensions to explain vote preferences. As noted above, we assume that left-authoritarians are guided by policy distance in their preferences. Indeed, we can see in Table 2 that, in the simplest model, policy distance on both dimensions affects ptv scores. The farther away a party is from left-authoritarians on either dimension, the less likely they are to find that party electorally attractive. For every one-unit increase in economic liberalism away from left-wing economic policies, we expect the ptv score to decline by 0.80 points, while a one-unit increase in socio-cultural liberalism is expected to decrease the ptv score by just 0.59 points.

Table 2. hierarchical linear regression results on stacked ptv data

	Model 1	Model 2	Model 3	Model 4
	Distance only	Distance + concern	Socio-demographic controls	Party preference controls
Party distance: economic dimension	–0.801^∗∗∗ (0.080)	–0.888^∗∗∗ (0.116)	–0.806^∗∗∗ (0.114)	–0.625^∗∗∗ (0.096)
Party distance: socio-cultural dimension	–0.586^∗∗∗ (0.058)	–0.752^∗∗∗ (0.085)	–0.663^∗∗∗ (0.083)	–0.460^∗∗∗ (0.072)
Concerned about economy		–0.544^∗∗∗ (0.164)	–0.605^∗∗∗ (0.162)	–0.305^∗ (0.142)
Concerned about immigration		–0.336^∗ (0.162)	–0.261 (0.160)	–0.160 (0.141)
Economic distance ^∗ economic concern		–0.185 (0.099)	–0.088 (0.097)	–0.126 (0.081)
Socio-cultural distance ^∗ immigration concern		–0.140 (0.072)	–0.141^∗ (0.071)	–0.133^∗ (0.060)
Socio-cultural distance ^∗ economic concern		0.360^∗∗∗ (0.073)	0.336^∗∗∗ (0.072)	0.194^∗∗ (0.061)
Economic distance ^∗ immigration concern		0.525^∗∗∗ (0.095)	0.453^∗∗∗ (0.094)	0.298^∗∗∗ (0.079)
Age			0.531^∗∗∗ (0.073)	0.389^∗∗∗ (0.063)
Female			0.841^∗∗∗ (0.132)	0.615^∗∗∗ (0.109)
Religious			0.841^∗∗∗ (0.047)	0.552^∗∗∗ (0.039)
Working class			0.815^∗∗∗ (0.051)	0.592^∗∗∗ (0.042)
Party ID				4.207^∗∗∗ (0.083)
Party best at most imp problem				2.385^∗∗∗ (0.089)
Intercept	5.312^∗∗∗ (0.162)	5.780^∗∗∗ (0.207)	5.607^∗∗∗ (0.206)	4.474^∗∗∗ (0.175)
Intercept variance, voters	0.674^∗∗∗ (0.067)	0.695^∗∗∗ (0.070)	0.670^∗∗∗ (0.068)	1.318^∗∗∗ (0.070)
Intercept variance, parties	1.190^∗∗∗ (0.199)	1.126^∗∗∗ (0.190)	1.143^∗∗∗ (0.191)	0.675^∗∗∗ (0.116)
Intercept variance, level 1 (voter∗party)	10.081^∗∗∗ (0.123)	10.045^∗∗∗ (0.127)	9.544^∗∗∗ (0.122)	6.325^∗∗∗ (0.082)
Observations	16330	15014	14720	14720
Log likelihood	–42626.74	–39189.25	–38059.77	–35602.01
AIC	85265.48	78402.5	76151.54	71240.01
BIC	85311.68	78493.9	76273.09	71376.76

Note: standard errors in parentheses; ^∗∗∗p < 0.001, ^∗∗p < 0.01, ^∗p < 0.05; model includes non-nested random intercepts for voters and parties.Sources: EE 2009 (van Egmond et al. 2011); Chapel Hill 2006 (Hooghe et al. 2010).

However, our hypothesis is that the extent to which policy distance matters should depend on the weight that the voter places on that dimension. This is tested in Model 2 (Table 2) by including interaction terms between voter concerns on the economy and immigration on the one hand and the policy distance variables on the other. Models 3 and 4 include further party- and voter-specific controls as robustness checks. These results show that findings from Model 2 are relatively robust. The size of the effects of policy distance and of the interaction terms decreases slightly in Model 3 and somewhat more in Model 4; this latter model includes two very strong and probably partly endogenous predictors of ptv scores. The two interaction terms which are strongly significant in Model 2 remain so in Models 3 and 4. Overall, this means that the nature of the patterns presented in Figures34 does not change much if control variables are included.

Figure 3 Marginal effect of policy distance on ptv scores Note: Coefficients calculated based on Model 2, Table 2. Bars indicate 95 per cent confidence interval around coefficient estimates. For coding of variables, see text.

Figure 4 Predicted ptv values for different levels of voter concern on the economy and immigrationSocio-cultural dimension Note: Graph shows predicted ptv values among four types of left-authoritarian voters: those worried about neither the economy nor immigration (box 1), those worried about the economy (box 2), those worried about immigration (box 3) and those worried about both issues (box 4). The numbers show how ptvs among left-authoritarian voters depend on party ideology, that is, the party positions on the economic dimension (x-axis) and the socio-cultural dimension (y-axis). The values are calculated based on Model 2, Table 2, for a voter who is at 1.5 on the economic dimension and 4.5 on the socio-cultural dimension. Party distance from left-authoritarians for each dimension is either 0 (complete congruence), 1.5 (moderate congruence, shown here as a position at 3) and 3 (low congruence, shown here as a position at 4.5 [economic dimension] and 1.5 [socio-cultural dimension]).

The main focus of our interpretation of the models is on the interaction effects. To understand their impact, we calculate both predicted marginal effects (Figure 3) as well as predicted values (Figure 4) based on Model 2 (Brambor et al. 2006; King et al. 2000). Turning first to the marginal effects, we can see that policy distance on the economic dimension matters more among voters who think the economy is not doing well: the predicted coefficient increases from 0.89 to 1.07, and this difference is statistically significant at the 0.1 level (p = 0.06). In contrast, for voters who think the economy is not doing well the predicted impact of policy distance on the socio-cultural dimension declines from 0.75 to 0.39 (p-value lower than 0.01).

A similar pattern is visible for voters concerned about immigration. Thus, policy distance on the economic dimension matters less among those who think that increased immigration is a problem. Here, the coefficient shrinks from 0.89 to 0.36, a difference significant at the 0.01 level. In turn, the impact of the dimension increases when voters think immigration is a problem: the coefficient increases from 0.75 to 0.89, though the p-value of this difference is only 0.05.

To gain a better grasp of how voter concern affects the impact of policy distance, we present predicted ptv values in Figure 4. Each box shows the estimated ptv for eight different parties, including right-authoritarian (top right corner of each box), left-liberal (bottom left corner) and right-liberal (bottom right). Each box represents a different type of left-authoritarian voter: the left-most box shows values for a left-authoritarian voter concerned neither about the economy nor about immigration; the second box for those concerned about the economy but not immigration; the third box for those concerned about immigration but not the economy; and the final box for those concerned about both issues. The predicted ptv values for right-authoritarian parties and left-liberal parties are highlighted to ease comparison.

When left-authoritarians are not worried about either topic, then they are predicted to be slightly more attracted to left-liberal parties than to right-authoritarian parties. The scores of 3.5 for left-liberal and 3.1 for right-authoritarian parties are nevertheless quite close to one another. Left-authoritarian voters concerned about the economy are a lot less attracted to right-authoritarian parties: the ptv value for the former group is 2, and the gap to left-liberal parties increases to 2.1 points. Left-authoritarian voters who are worried about immigration are attracted much less to left-liberal parties and much more to right-authoritarian parties: the two types of parties have predicted ptv scores of 2.8 and 4.4, respectively. Finally, among left-authoritarian voters concerned by both issues, left-liberal parties are just as attractive as right-authoritarian parties. Consistent with our theoretical expectations, right-liberal parties are relatively unattractive no matter which dimension voters are concerned about.

What these results show is that the impact of policy distance on the two dimensions depends on which issues voters are concerned about. In sum, we have solid evidence that the vote choice of left-authoritarian voters depends strongly on the issues with which they are concerned. When they are worried about the economy, they favour proximity on the economic dimension. When they are concerned by immigration, they are more attracted by a party close to them on the socio-cultural dimension.

Conclusion

For citizens who seek policy representation via elections, there may often be no clear-cut choice. Left-authoritarian views are held by many voters across Europe but find no direct correspondence at the party level. These voters face a particularly difficult task at the ballot box as no one party represents their views well. We argue that how they choose which type of congruence to prefer depends on the issues at the forefront of their mind, that is, which issues they are concerned about. Thus, the level of voter concerns about the economy and immigration has a strong influence on how proximity matters. The degree of left-authoritarians’ attraction to broadly right-authoritarian parties compared to their broadly left-liberal competitors therefore depends on the accessibility of each issue dimension.

Our study contributes to several theoretical debates. First, our arguments and findings about how voters choose between parties in situations where there is no fully congruent option are relevant to scholarship that assumes policy distance to be the key determinant of electoral preferences (Downs 1957). Second, this research relates to studies exploring the role of issue salience in electoral behaviour (e.g. Belanger and Meguid 2008; de Vries 2007; Green and Hobolt 2008; Singer 2011; van der Brug 2004). Drawing on the socio-psychological literature about accessibility and salience (e.g. Krosnick 1988; Zaller 1992; Zaller and Feldmann 1992), we extend this line of inquiry by applying it to a cross-pressure context. Third, these findings complement existing theoretical claims that radical-right parties are successful when they successfully mobilise voters’ grievances on immigration (e.g. Arzheimer 2009; Ivarsflaten 2008), but that they are vulnerable to the weight voters place on economic matters (Ivarsflaten 2005). As socio-cultural concerns linked to immigration and globalisation increase in importance, more left-authoritarians may privilege congruence on that dimension. Here, our findings stress the potential importance of media priming on electoral preferences. These findings also explain why, if they want to attract left-authoritarians, radical-right parties might want to take vague, imprecise positions on economic issues (Rovny 2012). Still, we found that the effect of economic views on party choice is stronger than that of socio-cultural views, and this is especially true when voters are concerned about the economy.

Given the current economic crisis, this has two implications. First, as long as economic concerns remain paramount, left-authoritarian voters should tend to vote for broadly left-liberal rather than broadly right-authoritarian parties. Second, the crisis presents a strategic opportunity for radical-right parties to adopt left-wing economic positions and therefore capture the left-authoritarian vote. Anti-austerity positions, coupled with anti-immigration rhetoric, could prove very attractive for this group of voters.

Our final point concerns how this research can be extended. To be sure, left-authoritarians are not the only group that may be particularly susceptible to cross-pressures. We found, for instance, that right-liberal voters are also often unrepresented by existing political parties. Hence, future research should consider how such ideological cross-pressures affect voters in general. For example, are they less satisfied with democracy? Are they less likely to turn out to vote? Do they tend to split their ticket across electoral arenas? Such questions could be explored in future work on left-authoritarians and other potentially cross-pressured groups of voters, whereby more attention is paid to supply-side differences between countries, i.e. in the kinds of parties competing in each system.

Acknowledgements

The research conducted for this paper has been financed by the Austrian Science Fund (FWF): S10902-G11 and I150-G14 under the auspices of the projects ‘Austrian Election Study’ (AUTNES) and ‘Representation and Policy Congruence in Europe’ (REPCONG) respectively. Zoe Lefkofridi also acknowledges support by the HumVib Eurocores Program of the European Science Foundation (ESF). We would like to thank Sylvia Kritzinger, our colleagues at the Dept. of Methods in the Social Sciences as well as Kostas Gemenis, Nathalie Giger, Monika Mühlböck and Agnieszka Walczak and two anonymous referees for valuable comments. An earlier version of this paper was presented at the 2011 ECPR Joint Sessions. All errors remain our own.

Notes

1. A related question that has received a lot of scholarly attention is why many working-class citizens and lower-income groups have left-wing economic views and conservative/authoritarian cultural views (e.g. Lipset 1959, 1966; for a more recent account, see Derks 2004). Unlike such sociological work, we concentrate here on issue-based representation and electoral choice for voters with such ideological preferences.

2. Given that the relationship between the two core dimensions under study (i.e. economic left–right and liberal/authoritarian) is different in the East compared to the West (e.g. Marks et al. 2006), our study confines itself to West European countries, i.e. the ‘old’ EU member states.

3. While the term ‘cross-pressures’ originally referred to individual membership of cross-cutting social groups (Lazarsfeld et al. 1960), we can also apply this term to issue- and policy-based cross-cutting appeals (Brader et al. 2013).

4. Ideally, we would be able to use two single-item summary dimensions indicating voter positions on economic and socio-cultural issues. To our knowledge, however, there is no cross-national dataset that places voters on such scales.

5. We do not use data reduction techniques to uncover dimensionality. In this decision, we follow the recommendation of Rovny and Marks (2011), who argue that the outcomes of such procedures depend heavily on the items included in the survey. Instead, we assume that West European political systems are at least partly structured by a socio-cultural and an economic dimension, so our approach is explicitly deductive.

6. We do not include items on European integration because this issue is related to both the economic and the socio-cultural dimensions (e.g. Hooghe et al. 2002).

7. The Cronbach’s alpha for the socio-cultural scale is a satisfactory 0.65 across all countries. For the economic scale, the Cronbach’s alpha is just 0.24, which is very low. To confirm that we were indeed reliably measuring voter’s economic views, we therefore also performed a principal component analysis (PCA) of the eight economic and socio-cultural items and extracted two varimax-rotated components. These components correlate with the two indices at r = 0.82 for the economic scale and at r = 0.95 for the socio-cultural scale. Our descriptive results do not differ substantively if we use this PCA-extracted component rather than the composite indicator. Note, finally, that these indicators are only used to classify voters and are not used in the subsequent regression analyses.

8. To exclude individuals with mixed or ambiguous views, we create cut-off points of 2.5 and 3.5 on the 1 to 5 scale. However, we need to acknowledge that responses to ‘agree–disagree’ questions may be coloured by acquiescence bias (Krosnick and Presser 2011). Although this is not an issue for the economic items, where there are two questions in each direction all four socio-cultural items point into the same (right-wing) direction. Hence, for the socio-cultural items, we adjust the cut-off points to 3 and 3.5. Doing so means that more voters are classified as left-liberal and right-liberal than would be the case if we left the cut-off point at 2.5; this coding approach is conservative in that it increases the size of these two groups relative to the number of left-authoritarians.

9. There are several advantages to using expert survey data rather than comparable data (e.g. based on coded manifestos or media sources) to assess party positions, as already argued by van der Brug and van Spanje (2009). Moreover, expert data tend to slightly underreport ideological changes of parties (see McDonald et al. 2006; van der Brug and van Spanje 2009), so matching 2006 party data to 2009 voter data should be valid. We also examined the EES candidate survey (Weßels 2011), which shows that there are also very few left-authoritarian candidates (less than 10 per cent) in Western Europe; details available from the authors on request.

10. The Cronbach’s alpha for the economic index is 0.97, for the socio-cultural scale 0.95. The economic index correlates with Hooghe et al.’s (2010) summary economic scale at 0.97; the socio-cultural index correlates with Hooghe et al.’s (2010) gal/tan scale at 0.86. Our results do not depend on the index or scale we use.

11. Acquiescence bias is less of a concern here as the questions are explicitly worded as dimensions with two endpoints.

12. Thomassen’s (2012) analysis of dynamic representation portrays the Dutch party system as self-correcting itself over time, i.e. as adjusting to voters’ preferences on the increasingly salient liberal/authoritarian dimension. Drawing on this insight, it is possible that some parties might have moved towards left-authoritarian positions since 2006. This should hold especially for countries with low electoral thresholds (Thomassen 2012). New parties might have formed that propagate such views, e.g. the Party for Freedom (PVV) founded by Geert Wilders. However the examination of movements on the supply side of democracy is beyond the scope of this article.

13. Ideally, we would also run our models using the distances of parties as perceived by voters themselves. Unfortunately, we know of no cross-national dataset that would include these perceptions on our two ideological dimensions.

14. These measures, on which we elaborate below, are also almost completely uncorrelated with respondents’ issue opinions on the two dimensions.

15. The full question texts are: (1) ‘What do you think about the economy? Compared to 12 months ago, do you think that the general economic situation in [COUNTRY] is a lot better, a little better, stayed the same, a little worse or a lot worse?’; and (2) ‘And over the next 12 months, how do you think the general economic situation in this country will be? Will it get a lot better, a little better, stay the same, a little worse or get a lot worse?

16. The full question texts are: (1) ‘And over the last 12 months, has immigration in Britain increased a lot, increased a little, stayed the same, decreased a little or decreased a lot?’; and (2) ‘In your opinion, is this a change for the better or the worse?’

Appendix 1. Attitude items forming the economic and socio-cultural indexes, EES 2009 (van Egmond et al. 2011)

The agree–disagree statements used to determine how many respondents are left-authoritarian are the following:

Table

Economic dimension
Q57	Private enterprise is the best way to solve [country’s] economic problems.
Q59	Major public services and industries ought to be in state ownership.
Q61	Politics should abstain from intervening in the economy.
Q63	Income and wealth should be redistributed towards ordinary people.
Socio-cultural dimension
Q56	Immigrants should be required to adapt to the customs of [country].
Q62	People who break the law should be given much harsher sentences than they are these days.
Q64	Schools must teach children to obey authority.
Q67	Immigration to Britain should be decreased significantly.

Appendix 2. Chapel Hill 2006, items used to construct the economic and socio-cultural indexes (Hooghe et al. 2010)

The positional assessments used to place parties on the two dimensions are the following:

Table

Economic dimension:

Q13: position on improving public services vs. reducing taxes (0: strongly favours improving public services, 10: strongly favours reducing taxes).

Q15: position on deregulation (0: strongly opposes deregulation of markets, 10: strongly favours deregulation of markets).

Q17: position on redistribution (0: strongly favours redistribution, 10: strongly opposes redistribution).

Socio-cultural dimension:

Q19: position on civil liberties versus law and order (0: strongly promotes civil liberties, 10: strongly supports tough measures to fight crime).

Q25: position on immigration policy (0: strongly opposes tough policy, 10: strongly supports tough policy).

Q27: position on integration of immigrants and asylum seekers (multiculturalism versus assimilation (0: strongly favours multiculturalism, 10: strongly opposes multiculturalism).

Appendix 3. Distribution of voters in a two-dimensional policy space

Notes: This Figure presents the proportion of survey respondents in each country who are left-authoritarian, right-authoritarian, left-liberal and right-liberal. The socio-cultural dimension is on the y-axis (1 = liberal, 5 = authoritarian), the economic dimension is on the x-axis (1 = economic left, 5 = economic right). The numbers do not add up to 100 per cent as in each country some respondents are located in the middle neutral sections (and hence are excluded from the four groups). See Appendix 1 for questions used to calculate indicators; the mean responses to four attitude questions form each dimension.

Source: EES 2009 (van Egmond et al. 2011).

Appendix 4. Distribution of parties in a two-dimensional policy space

Notes: This Figure presents the placement of political parties on the two dimensions. The socio-cultural dimension is on the y-axis (1 = liberal, 5 = authoritarian), the economic dimension on the x-axis (1 = economic left, 5 = economic right). See Appendix 2 for questions used to create the two dimensions. Expert survey values rescaled to range from 1 to 5 to match Appendix 3. Source: 2006 Chapel Hill Expert Survey (Hooghe et al. 2010).

Appendix 5. Robustness check using ‘most important problem’ questions

In the paper, we operationalise the ‘weight’ a voter attaches to an issue as the (lack of) concern the respondent expresses about recent developments related to that issue. As a robustness check, here we present a replication of the results using a measure constructed using open questions about what respondents see as the ‘most important problem’. This open question is asked at the start of the voter survey, and respondents are requested to name the three problems that they consider as most important in their country at the time the survey is conducted. Answers to these questions are primarily a measure of individual-level perceptions of contextual issue salience: a voter’s answers indicate whether in their eyes a problem is currently salient for the country as a whole (Johns 2010).

The responses were coded into broad categories by the EES (van Egmond et al. 2011); we simplified their coding by assigning categories to economic and socio-cultural concerns, if possible (see Appendix 6). We then created two indicator variables, each coded as 1 if the respondent names an economic or socio-cultural problem respectively as one of the most important concerns, 0 if not. These indicators were then interacted with policy distances as in Table 2.

Replicating Model 2 using this binary measure allows us to calculate the marginal effect of positional distance for different types of voters, similar to Figure 3 in the main text. We can see that:

• Among voters who see the economy as the most important issue, the effect of economic distance is greater. The coefficient changes from –0.73 to –0.88. This difference is not statistically significant at conventional levels.

• Among voters who see the economy as the most important issue, the effect of socio-cultural distance is also greater, which runs counter to our hypothesised interactive relationship. However, the difference in coefficients is far from statistically significant at conventional levels.

• Among voters who see socio-cultural topics as the most important issue, the effect of economic distance is smaller. The coefficient changes from –0.73 to –0.54. This difference is significant at p < 0.05.

• Among voters who see socio-cultural topics as the most important issue, the effect of socio-cultural distance is also greater. The coefficient changes from –0.51 to –0.74. This difference is significant at p < 0.01.

In sum, the results are very robust for the moderating effect of the salience of socio-cultural topics. The findings for the salience of economic topics are: socio-cultural distance are less strong, with no statistically significant interactions found and only one of these pointing in the hypothesised direction. Nevertheless, overall our results are reasonably robust to the use of this alternate (and in our view less well-suited) indicator.

Appendix 6. EES ‘most important problem’ codes assigned to the two dimensions

Economic dimension:

Table

Code	Issue
2	Environment
4	Economic conditions
51	Economic structure/policies/goals/conditions
52	Free enterprise capitalism
53	Private property rights
54	Government intervention/control over the economy (prices, wages rents)
56	Publicly-owned industry
57	Economic planning (of long-term economic planning, create of such a plan by authorities)
59	Government ownership, nationalisation in general (land. banks, etc.)
60	Privatisation (of government owned business or industry)
61	Corporatism (involvement of collaborations of employees and trade unions in the economic planning)
62	National policy on monopolies, trusts, consumer and small businesses protection
63	Incentives (references to tax and wage policies, financial incentives to start enterprises or stimulate investment)
64	Productivity (references to economic growth, the need to increase/facilitate production)
65	Technology and infrastructure (modernisation, development of industry, methods of transport, communication, research)
68	National media and ICT policy
69	Protectionism (as opposed to international cooperation, methods to protect national markets, economic growth)
70	Anti-growth economy (references to alternative economic planning e.g. Green politics)
71	Creating jobs (specifically)
74	Inflation
75	Unemployment
76	National employment policies
77	Stock market and its developments (shares, bonds, AEX, DAX, Dow Jones etc.)
78	Business (companies, banks, industry, mergers, manufacturing)
79	Bankruptcy of business, companies, banks (specifically)
80	Debt (public debt of a state, a community etc.)
81	Taxes
82	Trade (international trade), trade deficits
83	Wages and earnings
84	Effect of euro on the economy
85	Effects of financial crisis on domestic/ EU/ global economy
107	Labour groups (references to trade unions, unemployed, employees)

Socio-cultural dimension:

Table

Code	Issue
5	Immigration
23	Freedom and human rights
24	Civil rights, civil liberties, rights in general
25	Equality before law
30	Rule of law
99	Multi-socio-culturalism (socio-cultural diversity, socio-cultural plurality)
100	Law and order
101	Fight against terrorism
102	National crime prevention policy(ies)
103	Courts, trials, court decisions
104	National way of life (patriotism/nationalism, support for/opposition to established national ideas and/or values)
105	National immigration policy
110	Underprivileged minority groups
112	Homosexuals
113	Gay marriage
114	Ethnic minorities
115	Non-economic demographic groups