Exploring the neglected dimension of the economic vote: a global analysis of the positional economics thesis

Abstract

While much ink has been spilled as to how economics impacts election outcomes, existing scholarship has concentrated on the valence model, which focuses on assessments of incumbent administration’s handling of the economy. Recently, however, there has been an appreciation that economic voting is multidimensional. Nonetheless, the impact of positional economics – voters’ views of economic policy – on the vote remains less explored, especially from a cross-national perspective. In this contribution, we examine the impact of economic policy preferences regarding income redistribution and spending preferences on the vote in 32 states. Using the Comparative Study of Electoral Systems Module 4 data and hierarchical models, we show that voters’ economic policy preferences directly impact vote choice in many states. We also show that positional economic voting is more likely to take hold in mature democracies. However, support for the idea that ideological polarization contributes to macro-diversity concerning positional economic voting is mixed at best. Our research breaks new ground regarding positional economic voting and highlights how context impacts the extent of positional economic voting.

Introduction

Economic voting – for many, the concept brings forth the thought of the vengeful or rewarding voter – voters who punish outgoing governments for poor economic performance by voting against them but repay incumbents for a sound economy by reelecting them. The reward–punishment axiom, grounded in valence politics, has dominated studies of economic voting (e.g., Lewis-Beck and Stegmaier, 2000; Duch and Stevenson, 2008; Dassonneville and Lewis-Beck, 2019). Recently, there has been an appreciation that economic voting is multidimensional (Lewis-Beck and Nadeau, 2011; Lewis-Beck, Nadeau, and Foucault, 2013) with patrimonial and positional dimensions to the economic vote. Patrimonial voting focuses on the impact of asset ownership on the vote (e.g., Nadeau, Foucault, and Lewis-Beck, 2010; Lewis-Beck et al., 2013; Quinlan and Okolikj, 2019). Meanwhile, the third component of the trifecta, positional economic voting, remains understudied. Positional economic voting assumes that vote choice is driven by the voter’s selected position on economic policy – an issue motivated vote. Issue voting has a long pedigree in studies of electoral behavior (e.g., Nie, Verba, and Petrocik, 1979; Borre, 2001) and has its roots in rational choice models of electoral behavior (e.g., Downs, 1957), that is, voters are self-interested and will opt for candidates or parties whose positions on the issue in question are in sync with their own. Although a consensus is lacking regarding whether the stimuli driving issue voting are proximity (e.g., Downs, 1957; Blais, Nadeau, Gidengil, and Nevitte, 2001) or directionally (e.g., Rabinowitz and MacDonald, 1989; Adams, Bishin, and Dow, 2004) motivated, the critical point is that voters opt to vote for actors that in their judgment are most likely to implement the economic policies favored by them (Lewis-Beck and Nadeau, 2011). Thus, if the voter favors a specific policy to address unemployment, for example, they vote for the actor that is most likely to implement this policy, regardless of their incumbency status (see also Kiewiet, 1981).

The adoption of the positional perspective to studies of economic voting is relatively scant, at least vis-à-vis valence and patrimonial approaches. Early exceptions to this are Kiewiet (1981) and Bratton (1994). Nonetheless, a burgeoning literature is emerging (e.g., Lewis-Beck and Nadeau, 2009; Costa-Lobo, 2013; Lewis-Beck et al., 2013; Nezi and Katsanidou, 2014). However, most of these studies are country-specific, with a heavy concentration on the United States and Britain. Cross-national studies are rarer (exceptions are Van der Brug et al. 2007 and Papparo and Lewis-Beck, 2018). Moreover, existing analyses have relied on different measures of positional economic voting, some focusing on a single measure, raising the specter that these findings may be context-specific and/or measurement-specific. Our paper fills these voids by examining positional economic voting from a global (32 states) and multi-issue perspective. Our contribution has three aims: (1) to assess the extent of positional economic voting cross-nationally; (2) to explore whether positional economic voting varies by issue; and (3) the most novel of our goals on this subject, to account for possible macro-diversity in positional economic voting. Our starting point for this latter objective is that ‘context matters’, not an earth-shattering assumption as the literature on valence and patrimonial economic voting recognizes macro-heterogeneity (e.g., Powell Jnr. and Whitten, 1993; Singer, 2011; Quinlan and Okolikj, 2019). This contextual heterogeneity has also been observed in the issue voting literature (e.g., Kedar, 2005a, 2005b; Pardos-Prado and Dinas, 2010). We test whether positional economic voting varies in strength depending on two criteria. The first is a hypothesis grounded in democratization, where we test whether positional economic voting is more prevalent in mature democracies compared with newer ones. The second is a supply-side argument. Building on the literature above acknowledging macro-diversity concerning issue voting, we assume that positional voting will be more potent in states where political actors are more ideologically polarized compared with polities where differences between the parties and candidates are smaller.

Besides positional voting remaining the maverick of the trifecta of economic voting, our focus on it is propitious. With fewer voters anchored to political parties (Dalton, 2006) and some scholars arguing that traditional determinants of the vote have less influence than they once did (e.g., Franklin, 1992; Jansen, Evans, and de Graff, 2012), there is potential for the economy to take on higher saliency in shaping the vote. Some evidence implies economic considerations have become more critical for some voters (Kayser and Wleizen, 2011), although other literature suggests the valence component’s importance has remained stable over time (Dassonneville and Lewis-Beck, 2019). Our focus allows us to explore whether the economy’s influence on the vote has the potential to be more than merely performance-based and could be issue-driven too. An additional advantage of the positional economic thesis is that it factors in voters’ prospective judgments more comprehensively. The dominant valence economic paradigm falls short on this as it overlooks the possibility that attributing blame does not necessarily lead to vote switching (see Bratton, 1994, p. 280).

Using the Comparative Study of Electoral Systems (CSES) Module 4 data (2018), we demonstrate that voter positions on economic policy (income redistribution and unemployment and defense spending preferences, respectively) influence vote choice, despite a myriad of control variables. Hence, our study supports the idea that economic voting is multidimensional. However, the extent of positional economic voting varies by country and issue. We show that positional economic voting is more likely to take hold in mature democracies, a finding which holds across issues. The evidence that ideological polarization conditions positional economic voting is underwhelming, however, and lacks robustness. The article proceeds as follows. Next, we review the positional economic voting literature before devising a set of hypotheses for testing. Following this, we detail our research strategy. We then report our empirical analysis. Finally, we summarize our findings and outline remaining research frontiers in the area.

Theory and hypotheses

Positional economic voting: Intuition and the story so far

The former Australian Treasurer, Paul Keating, once quipped that, ‘I guarantee if you walk into any pet shop in Australia, the resident galah will be talking about microeconomic policy’ (‘The Recession We Had To Have’ 1993). Keating’s remark was designed to highlight the ubiquity of economic policy considerations among ordinary voters in Australia. However, the retort also gets to the nub of positional economic voting – voters have views on economic policy, these perspectives matter to the voter, and these views influence their vote. Positional economic voting is premised on a rational choice explanation of electoral behavior where citizens are assumed to be utility maximizers (Downs, 1957). Hence, they compare political actors’ positions on issues against their own and vote for the actor most likely to deliver the policy they desire.

While there is an agreement in the spatial model that electors evaluate actors on policy and that voters are utility maximizers, scholars hotly contest the motivations underlying voter utility. The traditional proximity model assumes voters support an actor closest to their position on the issue with utility at a maximum when the voter’s position overlaps with the actor’s position (e.g., Downs, 1957; Blais et al., 2001). However, the directional model understands utility as a two-step process, with voters first evaluating which side of the issue an actor is on and then voting for the actor with the most forceful stance on the issue (e.g., Rabinowitz and MacDonald, 1989; Adams, Bishin, et al., 2004).¹ Despite the spirited debate, research highlights how disentangling these two dimensions from one another is challenging (e.g., Lewis and King, 1999). More critically from our perspective, literature exists implying that both proximity and directional considerations can drive issue voting (e.g., Cho and Endersby, 2003; Merrill III, 1994). Additionally, Fazekas and Méder (2013) found that in three-quarters of cases, proximity and directional motivations result in indistinguishable predictions, a finding verified by Weber (2015, p. 523) who concluded that ‘with strong synergies between the three logics, pitting the models against each other no longer seems crucial.’ Hence, the absence of evidence that one model is superior over the other and literature implying that whichever motivation is assumed leads to near-identical vote choice, we do not assume that one particular model is the driver of positional economic voting (see Lachat 2008, p. 690 for similar reasoning). Thus, the intuition of position economic voting can be summarized as follows: it assumes that voters have positions on economic issues, that they know where political actors stand on these issues, and perceive differences between the actors on offer regarding the policy. In these circumstances, they vote for the actor they consider most likely to implement the policy, although the specific motivations underlying this choice can vary.

It would be a stretch to say economics has not featured in issue voting research. However, the tendency has been to look at the issue through the prism of the performance of incumbent administrations bringing it back to the valence economic model. Indeed, a 2011 review (Lewis-Beck and Nadeau, 2011, p. 289) of positional economic voting literature highlighted ‘the dearth of research in this area’, concluding ‘after conducting an extensive search for papers testing the policy-oriented, positional hypothesis of economic voting, we came up with only four candidates.’ Of the four studies noted in this pivotal review, Bratton (1994) explored voters’ perceptions concerning the budget deficit on vote choice in the 1988 US Presidential election. She concluded that while the issue dampened enthusiasm for the incumbent Republican Party, the deficit did not help the Democrats in the election as the Republicans were perceived to be better at managing it. Elsewhere, Alvarez and Nagler (1998) tapped the impact of voter economic policy preferences on the vote in the 1996 US Presidential election. They found that the further respondents’ sentiments concerning support for government services were from candidates Bill Clinton’s or Bob Dole’s economic positions influenced their electoral support. Niemi et al., (1995) demonstrated that increases in state taxes in the United States correlated with loss of votes for US governors (also see Kone and Winters, 1993). Another study overlooked in the original review but which incorporated aggregate positional economic dynamics was van der Brug et al. (2007). In their two-step model of vote choice where voters first determine their propensity to support a party, and having compared these predilections opt for the party they support most strongly (van der Brug et al., 2007, p. 46), their research showed that economic issue voting in 15 European states was relevant for opposition voters who judged political actors on a prospective basis (2007, pp. 99–102).

Since the 2011 proclamation of a scarcity of research in the area, more researchers have embraced the positional economics perspective. Nadeau, Lewis-Beck, and Foucault have been trailblazers in this regard, finding persuasive evidence that voters’ economic policy preferences affect the vote in the United States (Lewis-Beck and Nadeau, 2009; Lewis-Beck and Nadeau, 2011) and Britain (Lewis-Beck et al., 2013). Elsewhere, Nezi and Katsanidou (2014) found the attitudes of Greek voters to the EU-IMF bailout influenced vote choice in their 2012 general election. While Foucault et al. (2017) examined the impact of tax policy on the vote in 21 countries finding that it impacts left-wing parties’ support, with the relationship moderated by the extent of clarity of responsibility and aggregate economic conditions. Conversely, work by Costa-Lobo (2013, p. 464) found that voter attitudes on a range of economic policy items had no significant impact on vote choice in Portugal. More recently, research has emerged suggesting that positional economic voting is a more powerful determinant of vote choice than the traditional valence economic measure which has dominated studies of economic voting (e.g., Lewis-Beck et al., 2013, p. 256; Papparo and Lewis-Beck, 2018).

Two essential takeaways regarding positional economic voting transpire from our above overview. The first concerns the measurement of economic issue preferences. Many existing studies have relied on single items raising the possibility that we are capturing a salient issue in a particular election, which granted would still represent a positional economic dimension, but which might overstretch the importance of positional economic voting. Alternatively, relying on single-item measures to capture issues increases the potential of measurement error resulting in the effects of issue voting being downplayed (Achen, 1975; Ansolabehere, Rodden, and Snyder, 2008), leading Ansolabehere et al., (2008, p. 228) to ‘caution in drawing inferences from analyses of surveys that focus on single items introduced to capture issue preference’. Moreover, different studies have used various measures in testing the positional economic voting thesis. For it to stand on firmer ground, there is a need to incorporate multiple and consistent economic policy preferences into the mix. Our study responds by bringing forth three measures of positional economic voting (see next section).

The second takeaway is the positional economic thesis has been primarily tested in established democracies. Thus, the extent of variation between political systems is limited, allowing little study of macro-heterogeneity. Macro-diversity might be expected, considering that the other dimensions of economic voting are known to vary by context and we know that issue voting can be subject to contextual heterogeneity. Our study addresses this shortcoming by taking a comparative cross-national approach encompassing mature and newer democracies.

Positional economic voting: New frontiers and expectations

Our first new frontier is to put the positional economic voting thesis to a global test. We concentrate on three economic positions in this paper. Our first taps voter attitudes to the state’s role in income redistribution – what might be classified as the ‘go-to’ measure of positional economic voting (e.g., Lewis-Beck and Nadeau, 2009; Lewis-Beck et al., 2013). The other two positional economic issues are voter preferences regarding government spending on national defense and voter predilections for government spending on unemployment. The addition of the spending preferences is a new dimension to positional economic voting, although countless studies have explored how spending preferences and public opinion are connected (e.g., Wlezien, 1995; Soroka and Wlezien, 2010). The choice is advantageous as we have taken two policy issues, one traditionally associated with rightist and the other with leftist political actors. Ideological disposition is essential in understanding political behavior (Dalton, 2008) and is a critical heuristic for voters concerning the policies actors might be expected to implement if they gain power. Research shows that voters understand this can infer party positions from these stances and that it can map to vote choice (e.g., Kedar, 2005b; Fortunato and Stevenson, 2013). Consequently, we can have confidence that positional economic voters can understand which actors are more likely to implement their desired preference.

Thus, we have three initial hypotheses. We have specified our expectations considering the expected behavior of a voter on the issue and how this theoretically should map to voting for a rightist political actor. Thus, we suppose that:

Hypothesis 1a: As voters’ preferences for income redistribution decreases, the probability of voting for rightists will increase.

Hypothesis 1b: As voters’ preferences for unemployment spending decreases, the probability of voting for rightists will increase.

Hypothesis 1c: As voters’ preferences for defense spending increases, the probability of voting for rightists will increase.

Studies of economic voting are synonymous in the view that contextual circumstances influence the extent and magnitude of the economic vote. Macro-diversity is established concerning valence economic voting (e.g., Powell Jnr. and Whitten, 1993; Singer, 2011). Evidence suggests that macro-heterogeneity also translates to patrimonial voting (Quinlan and Okolikj, 2019). Besides this, a voluminous literature has identified contextual heterogeneity regarding issue voting (e.g., Kedar, 2005a, 2005b; Pardos-Prado and Dinas, 2010; Singh, 2010; Lachat, 2011). In what follows, we focus on two potential sources for this expected contextual heterogeneity regarding positional economic voting. We do not assume that these are the only sources of macro-variety. Rather, we suppose that these are two features that may be prominent in explaining the moderating effects of context concerning positional economic voting.

Our first assumption is that the maturity of democracy could matter as it partly determines political competition. We know that the emergence of political divides cross-nationally is heterogeneous. Competition in well-established democracies primarily developed around divisions evident in society in the 1920s, followed by ‘freezing’, with the basic structure of competition remaining relatively stable throughout the 20th century (Lipset and Rokkan, 1967; Bartolini and Mair, 1990; Golder, 2003).² Thus, these party systems tend to be based on ‘cleavages’ – deep and lasting divisions between groups around societal conflicts whether it is state vs. church, worker vs. the business classes or geographical divisions concerning the center vs. the periphery. This evolution of political competition is especially prominent in Western Europe and has resulted in party system institutionalization taking root. It is illustrated by voters enduring connections with political actors and has contributed to significantly less electoral volatility. However, this pattern of competition did not necessarily take hold in newer democracies in Latin America or Central and Eastern Europe (McFaul, 2002) where multiparty competition followed single-party dominance. Instead, democratization in many instances led to weak party development, less partisanship, patronage, control of the media, and a weakened civil society (Mainwaring, 1999; Carothers, 2002; Mainwaring and Torcal, 2006; Diamond, 2007). This is a phenomenon not specific to any region, with weak party institutionalization noted in Latin America (e.g., Mainwaring, 1999; Mainwaring and Torcal, 2006; Crabtree, 2010), South Africa, where single-party dominance persisted after democratization, thus complicating democratic consolidation (e.g., Giliomee, 1995; Randall and Svåsand, 2002), and in Central and Eastern Europe (e.g., Bielasiak, 2006; Tavits and Letki, 2009).

However, what are the implications of these different contextual circumstances concerning positional economic voting? We argue that leftists and rightists in newer democracies have the potential to deviate somewhat from the traditional economic policy stances associated with ‘their’ ideological side, at least vis-à-vis political actors in established democracies. Tavits and Letki (2009) have provided persuasive evidence that leftists in Fourth Wave democratization polities had stronger predilections for imposing fiscal discipline (the opposite to what we might expect of them theoretically). Conversely, rightists showed inclinations for higher spending to alleviate economic difficulties (again, contrary to what we might have expected). The driving force behind these deviations is that leftists in these states had stronger incentives to enact greater fiscal discipline to distance themselves from their Communist past. However, rightists – most of whom assumed power straight after the collapse of Communism – were compelled to spend more to alleviate the economic hardships they inherited on taking office. At face, this mechanism seems to hold well for the post-Communist countries. However, in Latin American states, evidence exists illustrating that the link between the political ideology of actors and the policies carried out when office was disrupted because of international pressure for the implementation of the Washington Consensus – a set of economic policy prescriptions including free trade, deregulation, privatization, and an adherence to fiscal discipline. In trying to deal with the fallout of the debt crisis of the 1980s in the region, many actors implemented these market-driven policies despite campaigning on a different platform and regardless of their conventional ideological label (Conaghan, 1996; Johnson and Crisp, 2003).

Consequently, while we expect individual voters in newer democracies are able to place political actors correctly on the left-right ideological continuum (see Tavits and Letki, 2009 for a similar assumption), they may associate the economic policy outputs of the respective actors differently, resulting in a deviation from the anticipated positional economic voting relationship in these states.³ Thus, a voter wishing to see more spending on welfare and trying to maximize his utility might express this by deviating from the traditional expectations we have in support of leftist and rightist actors – not because he cannot identify the actor’s ideological stance, but due to the economic policy outputs of actors in newer democracies being less conventionally clear-cut. Conversely, the situation in established democracies is expected to be more straightforward, given the relative stability of the system and political actors. Consequently, it is more likely that political actors will implement economic policy outputs associated with their traditional ideological position. Following this reasoning, we posit that:

Hypothesis 2a: In mature democracies where actors’ policy outputs are traditionally in sync with their ideological stances, the relationship between positional economic voting and support for rightists will be stronger.

Our second expectation is that ideological polarization within a system influences positional economic voting. By polarization, we mean a clear division in the political system between actors on issues, including economic ones. Ideological divergence is known to be an essential dimension in understanding a state’s political system (Dalton, 2008). For positional economic voting to work, a certain degree of differences between the actors’ positions on economic issues is essential. The theory assumes that voters form perceptions about which economic policies they would like to see implemented, connect these sentiments to the agendas of political actors, and thus vote for the actor best equipped to maximize their utility. At face, this path may appear straightforward if we assume that political parties’ agendas are constant. Sometimes this happens, but sometimes it does not, as parties can adjust their positions to attract support (e.g., Adams, Clarke, Ezrow, and Glasgow, 2004; Adams and Somer-Topcu, 2009). All of these considerations put the spotlight on the extent of ideological difference, which can be expected to vary cross-nationally.

Our expectations regarding polarization mattering are bolstered, as issue voting research has demonstrated that the extent of ideological differences influences proximity voting (e.g., Pardos-Prado and Dinas, 2010; Singh, 2010; Lachat, 2011). The conventional relationship is that the greater the policy differences between actors, the easier it becomes for a voter to cast a policy-motivated vote as clarity makes it easier to identify the actor more probable to implement the policy. Drawing on this literature, we posit that lower levels of ideological polarization will produce lower levels of positional economic voting, as in these circumstances the differences between actors on issues (including economic ones) will be minimal. In sum, if parties are too close to one another on policy, it is likely to reduce the positional discrepancies to near zero, and thus push the economic dimension to be a valence concern. Hence, we expect that:

Hypothesis 2b: The higher the ideological polarization between actors in a polity, the relationship between positional economic voting and support for rightists will be stronger.

Research Strategy

Data and case selection

Our data come from the CSES Module 4 survey (2018). The CSES is a cross-sectional comparative study ideal for testing micro-macro explanations of electoral behavior as identical questions are posed to survey respondents in different states. CSES Module 4 includes 45 election studies from 39 polities. Of these 39 states, we excluded Hong Kong and Thailand as they failed to pass the threshold of having democratic elections. We omit Argentina because voters’ spending preferences on defense and unemployment were unavailable. We excluded the Philippines because of lack of variation in the dependent variable and Kenya and Montenegro because of sizeable unit nonresponse to the ideology question, an essential control variable. We discount Taiwan because left and right divisions are not prominent (Jou, 2010). Our original micro number of observations is 40,682 respondents across 32 countries in 32 elections.⁴ In addition to our use of the CSES data, we obtained our polarization data from Dalton’s index (2017) from the CSES website.

Modeling strategy and variable operationalizations

In the CSES, observations are clustered, infringing on the assumption that they are independent. To account for this, we employ a multilevel multivariate estimation strategy. We define two levels of analysis: citizens (micro-level) that are nested in countries (macro-level).

Our dependent variable is whether a respondent reported voting for a right-wing actor, primarily in lower house elections (see Lewis-Beck et al., 2013 for a similar strategy).⁵ Guiding our ideological classification of actors into left and right, we use the expert cataloging of which party family the actor belongs to. When this measure was inconclusive, we relied, where data were available, on the CSES experts’ placement of parties on the left-right scale. For validity, we compared these judgments with voter opinions of actors by examining the mean placing of the parties in question by voters on the left-right scale. We detail our categorization of the parties/candidates in Appendix C.

We have three primary independent variables that tap voters’ positions on economic issues.⁶ The first is the standard positional economic voting question tapping attitudes to income redistribution. Survey respondents were asked to express their level of agreement on a five-point Likert scale to the following question: ‘The government should take measures to reduce differences in income levels.’ Low values indicate preferences favoring government interventions to reduce income levels, while high values indicate predilections for those who are against. The other two measures examine the respondent’s relative preferences on government spending (e.g., Wlezien, 1995; Soroka and Wlezien, 2010) regarding unemployment and defense.⁷ Respondents are asked whether they want more spending, less spending, or the same level as they currently perceive. High values indicate those who favor more spending and low values those who favor a reduction. Survey respondents are primed that providing the answer ‘more’ might require an increase in taxes, while a preference for ‘less’ spending could result in a reduction in the services.

We have two macro-level variables of interest. The first is the maturity of democracy captured here by how many consecutive years of democracy the state in question has had. This variable comes from the Polity IV project, which is included in the CSES macro-data component. Higher values indicate polities who have been democratic longer. Second is polarization, which we measure using the Dalton polarization index (2008, 2017). It uses citizen opinions of the left-right position of the actors and weights these estimates by the vote share achieved by each actor. It provides a consistent basis for measuring polarization cross-nationally using CSES data. Higher values on the scale imply more ideological polarization between actors.

Our estimation strategy involves three steps. First, we take a global perspective and estimate a multilevel model (what we refer to as ‘a global model’) examining the effect of each positional measure on the vote controlling for multiple micro- and macro-correlates. CSES Module 4 is uniquely suited to estimating a comprehensive economic voting model, as it includes measures of the trifecta of economic voting: positional economics (our primary interest here), valence economic perceptions, and patrimony. Our model controls for these two economic voting covariates, along with voters’ political identity as represented by the respondent’s self-reported placement on the left-right ideological scale,⁸ and the traditional sociodemographic correlates of age, gender, and education.⁹ We also include macro-level controls for Dalton polarization, age of democracy, whether a country used a proportional election system or not and the economic conditions of a country at the time of the election, the latter two traditional control variables at the macro-level.¹⁰ Second, we take a step back to the polity level and reestimate the global model for each individual polity in our sample to allow us to uncover the extent of positional economic voting by state (for an overview of the results, see Table 2).¹¹ Third, we return to our global multilevel analysis to test our expectations regarding macro-heterogeneity. To do this, we interact with our two macro-variables of interest, namely Dalton polarization and age of democracy, our positional economic items, creating a series of cross-level interactions (see Tables 3 and 4). To enable comparability across our models, we rescaled all coefficients to run on a scale from 0 to 1.¹²

Table 2. The effect of positional economic attitudes on the vote by country by item

Source of data: CSES (2018). Key: ☑ = significant direct effect on the vote

Please note: Positional economic attitudes are considered to have a significant effect on the vote if P < 0.05. The analysis based on logit models with positional economics regressed on the vote for rightist actors in each country with a series of controls being added in a stepwise fashion.

Table 3. Multilevel logit model examining the impact of voter economic positions on the probability of voting for rightists and macro-variability in elections in 32 countries 2011–2016

Source of data: CSES 2018.

Note: Variables are all standardized from 0 to 1. * P < 0.05 ** P < 0.01 *** P < 0.001.

Table 4. Multilevel logit model examining the impact of voter economic positions on the probability of voting for right and macro-variability in elections in 32 countries 2011–2016

Source of data: CSES 2018.

Note: Variables are all standardized from 0 to 1. * P < 0.05 ** P < 0.01 *** P < 0.001.

Empirical analysis

Positional economic sentiments by polity and bivariate relationship with the vote

We begin our analysis by exploring the distributions of our three measures and their bivariate relationship with vote. Figure 1 displays the distribution of opinion by country in favor of redistribution (left portion) and its relationship to vote (right portion). The left segment of Figure 1 shows that most voters favor the government taking action to reduce differences in income. Only in three countries does this opinion fail to get majority support (USA, Australia, and Ireland). Moreover, just in five nations does disagreement with the proposition top a quarter of voters (USA, Canada, Norway, Iceland, and the Czech Republic). In short, most voters in our sample favored redistribution. Concerning the vote, the results are encouraging for the positional voting concept (right-hand segment of Figure 1). Among respondents who said they strongly favored government action on redistribution, 39% report voting for rightist parties. It rises monotonically to 69% among those who vehemently disagree with the proposition.

Figure 1. Attitudes toward redistribution by country (left portion) and its impact on vote choice (right portion) in 32 countries 2011–2016. Source of data: CSES 2018. Base: Voters. N = 36,443. Note: Data in left portion is ordered by respondent disagreeing with the proposition in descending order. In Canada, no neutral category was offered to respondents.

Figure 2 turns the spotlight toward spending preferences. The two left quadrants show the distribution of opinion by country. On unemployment spending (top left quadrant), most voters said they preferred retaining the current level of spending (modal response in 14 polities) or increasing it (modal response in 18 countries). Only in Britain was there a majority support for reducing spending on unemployment. Preferences on defense spending were more mixed (bottom left quadrant). In seven countries (Britain, Brazil, Latvia, Peru, Romania, Serbia, and Turkey), a majority supported increasing defense spending. Conversely, in Austria, the Czech Republic, France, Germany, Greece, Slovenia, and Switzerland, majorities favored decreases. The modal response for the remaining state was to retain defense spending at current levels.

Figure 2. Attitudes toward public spending on unemployment (top left quadrant) and defense (bottom left quadrant) and their respective impact on vote choice (welfare top right quadrant and defense bottom right quadrant) in 32 countries 2011–2016. Source of data: CSES 2018. Base: Voters. N = 36,295 for unemployment; N = 35,876 for defense. Note: Data distributions are ordered by respondent favoring less spending on welfare and more spending on defense in descending order.

The right-hand segments of Figure 2 display the relationships with the vote. Again, the results are heartening for positional economic voting. The top right hand quadrant of Figure 2 shows that among those who prefer less spending on unemployment, support for the right is higher compared to those who prefer more public spending on the issue. As with redistribution, the relationship is monotonic, with a 27-point difference between the two ranges (63–36). Regarding the relationship between vote and defense spending (bottom right quadrant of Figure 2), those favoring more spending on defense are more likely to vote for the right. We see that voting for rightists steadily rises, the more one favors increased defense spending (from 35% among those who favor much less defense spending to 54% among those who favor somewhat more). However, the relationship tapers off among those who prefer much more spending. In sum, the preceding analysis suggests voters’ positions on economic issues maps on to vote. However, if the positional voting concept is to stand on firmer ground, a multivariate analysis is necessary.

Positional economic voting: hierarchical analyses

Table 1 displays our first set of hierarchical models. We find no matter what the positional economic measure we use, a robust relationship between voters’ economic policy preferences and vote choice exists, despite multiple controls. Taking Model I first, which measures attitudes toward income distribution, the coefficient is positive and statistically significant, implying that the more one disagrees with the proposition that the government should take measures to reduce differences in income, the higher the probability of voting for the right. It is in line with expectations. For unemployment spending, which we test in Model II, the coefficient is statistically significant (P < 0.001) but negative. It implies that the more one favors increased spending on unemployment, the less likely one is to vote for rightists, as expected. In Model III, the coefficient for defense spending preferences is positive (P < 0.001), signifying that those who favor more spending on defense are more likely to support rightists, again what we presumed.

Table 1. Multilevel logit model examining the impact of voter economic positions on the probability of voting for right in elections in 32 countries 2011–2016

Source of data: CSES (2018).

Note: Variables are all standardized from 0 to 1. * P < 0.05 ** P < 0.01 *** P < 0.001.

As the effects of logit regression coefficients are difficult to interpret, we estimate the marginal effects at the means (MEM) with covariates in the model held constant at their mean values.¹³ For income redistribution, we see that all other things being equal, the estimated likelihood of voting for a rightist party is about 27 points (95% CI: 24.2, 29.9) higher among individuals who vehemently disagree that the government should take measures to reduce differences in income levels vis-à-vis with those whose strongly favor government intervention. For unemployment spending, the likelihood of voting for the right is estimated to be 25.5 points lower (95% CI: 22.3, 28.6) among those who favor more spending on unemployment compared with those who favor spending less at present. For defense spending, the likelihood of voting for the right increases by an estimated 20.4 points (95% CI: 17.4, 23.3) for those who favor spending more on defense than present compared with those who favor spending much less than now. Hence, voter positions on economic issues influence the vote, offering support to Hypothesis 1a–c, even in the face of a series of controls.

Positional economic voting and macro-heterogeneity: polity level and hierarchical analysis with cross-level interactions

Our analysis to tease out the extent of and the reasons underlying macro-heterogeneity has two components. First, we turn to a polity-level analysis and reestimate our global models I–III from Table 1 at the country level to get a sense of how positional economic voting dynamics are playing out in each state. We summarize our results in Table 2. Column I details where attitudes toward income redistribution remained a statistically significant predictor (P < 0.05) of the vote for rightists. The second column tests whether sentiments regarding public spending on unemployment remained statistically potent, and the third column, whether defense spending preferences did.

As expected, we see a substantial degree of variation across polities and issues. In six states (Australia, Britain, Canada, New Zealand, Switzerland, and USA), all three positional economic issues influenced the vote in a meaningful way. In eight states (Austria, Bulgaria, France, Finland, Germany, Greece, Latvia, and Sweden), two of the three positional issues affected voting for rightists. Meanwhile, in eight other polities (Brazil, Iceland, Japan, Norway, Portugal, Romania, South Africa, and South Korea), only one issue meaningfully influenced the vote for the rightists. By issue, preferences toward income redistribution directly affect vote choice in 17 polities (i.e., 53% of macro-cases). Defense spending and welfare spending preferences had a significant impact on the vote in 13 and 12 (41% and 38% of macro-cases) polities, respectively. However, in the remaining 10 states in our sample, we found no evidence that voters’ economic positions influenced the vote. In sum, our evidence points to both macro and issue heterogeneity.

Our second step is to tease out what contributes to this macro-variability. For this, we return to our hierarchical model specification. We have two expectations about what might account for this macro-diversity. The first concerns the maturity of democracy. To test this explanation, we add a cross-level interaction between the age of democracy and our three positional economic measures. We expect that the relationship between positional economics and the vote will be stronger in mature democracies vis-à-vis newer democracies. We describe our results in Table 3.

Our proposition receives strong support across the three positional economic items. Taking income redistribution first (Model IV Table 3), the positive (and statistically significant) coefficient for the cross-level interaction between income redistribution preferences and democratic maturity implies voting for rightists is stronger in more established democracies. On unemployment spending preferences (model V, Table 3), the negative coefficient for the cross-level interaction infers that the relationship is also stronger in mature democracies. The positive coefficient for the cross-level interaction with defense spending (model VI, Table 3) signifies the same, bolstering our confidence on our supposition that positional economic voting is a phenomenon more likely to take root in mature democracies, which gives support for Hypothesis 2a.

Figure 3a–c plots the average difference in the probability of voting for rightists depending on the extent of a country’s democratic maturity. The square dots show the estimated change in voting for rightists given a one-unit change in preferences regarding redistribution or spending preferences for unemployment and defense, respectively, by democratic maturity. The vertical lines around the square dots represent the 95% CIs. Higher scores on the democratic maturity measure represent where democracy has been in place longer. What Figure 3a illustrates is that an increase of one on the income redistribution scale (i.e., moving toward a more negative stance regarding government actions on reducing income levels) gradually increases the likelihood of supporting a rightist, the more mature a democracy is. Figure 3b illustrates likewise except that the pattern is a negative relationship, as expected, as higher values on the spending scale imply wanting more government spending. Hence, Figure 3b implies that an increase of one (representing more positivity toward government spending on unemployment) gradually decreases the likelihood of supporting a rightist, the more mature a democracy is. Figure 3c plots the effect of preferences on defense spending by democratic maturity on voting for rightists. It shows that an increase of one on the defense spending preference (i.e., favoring more defense spending) gradually increases the likelihood of supporting rightists, the more mature a democracy is.

Figure 3. (a–c) Marginal effects plots estimating the impact of positional economics on the likelihood of supporting rightists depending on democratic maturity Source of data: CSES 2018. Base: Voters. Note: The shapes show estimated change in voting for rightists given a one-unit change in attitudes toward government redistribution (Fig 3a – triangles), a one-unit change in unemployment spending preferences (Fig 3b – circles), and one-unit change in defense spending preferences (Fig 3c – squares) by democratic maturity. The vertical lines around the shapes represent the estimated 95% confidence intervals. Estimates are based on the fixed components of the model using STATA’s margin (dydx) function.

Our second proposition concerning macro-heterogeneity tests whether positional economic voting varies depending on the extent of ideological polarization in a state. Again, we test this with a series of cross-level interactions between our three positional economic items and the Dalton polarization index being added to our hierarchical models (see Table 4). The evidence from Model VII suggests that sentiments regarding income redistribution and voting for rightists are not influenced by ideological polarization, with the cross-level interaction failing to attain conventional levels of statistical significance. For unemployment spending, the cross-level interaction between unemployment spending and the ideological index is negative and significant at P < 0.05. Thus, the higher the ideological polarization in the system, the likelihood of supporting a rightist party declines at a steeper rate the more one favors higher welfare spending, what we might have expected. However, for defense spending (Model IX), the coefficient for the cross-level interaction fails to reach statistical significance. Hence, the results for polarization mattering are weak, being relevant for spending on unemployment, and even then, the effect is relatively meek. In sum, there is little support for Hypothesis 2b.

Conclusion

Positional economic voting has traditionally assumed Cinderella status within the economic voting and issue voting literature. In this paper, we build on a burgeoning literature and break new ground by providing the first global and consistent multi-measure assessment of positional economic voting. Despite an array of controls, our study shows that voter positions on economic issues influence the vote in several polities. This finding holds across multiple measures of economic positions. It is a significant finding and demonstrates that economics influence on vote choice goes beyond voter judgments on the incumbent’s economic performance, the traditional means of exploring economic voting, and instead comprises multiple components.

Our contribution also explored macro-diversity regarding positional economic voting. Like its patrimonial and valence siblings, we establish that positional economic voting is subject to macro-heterogeneity. In our sample of 32 countries, we found evidence of meaningful positional economic voting in 22 polities (68.5% of macro-cases). While an impressive figure, we should not overlook that in 10 states, we uncovered no evidence of positional economic voting whatsoever. We found substantial evidence that positional economic voting was more likely to take root in more mature democracies compared with newer democracies. Also, while we uncovered some evidence that ideological polarization contributes to macro-variation on preferences for unemployment spending, the relationship is weak and less robust across specifications.

Frontiers remain to be conquered. One area that deserves attention is to explore the positional economic voting thesis over time. It would address a potential drawback with our study – namely that the positional economic measures at our disposal were all measured between 2011 and 2016 as many states recovered from the Global Financial Crisis. Consequently, this economic environment may have shaped voter economic preferences in a particular way. Additionally, exploring which voters are more likely to be positional economic voters would be a fruitful avenue to explore. Ours is unlikely to be the last word on the subject.