Volume 39, Issue S1 p. 279-304
SYMPOSIUM ON PARTISANSHIP
Open Access

The Nature of Party Categories in Two-Party and Multiparty Systems

Stephen P. Nicholson

Corresponding Author

Stephen P. Nicholson

University of California, Merced

Correspondence concerning this article should be addressed to Stephen P. Nicholson, University of California, Merced. E-mail: [email protected]Search for more papers by this author
Christopher J. Carman

Christopher J. Carman

Strathclyde University

Search for more papers by this author
Chelsea M. Coe

Chelsea M. Coe

University of California, Merced

Search for more papers by this author
Aidan Feeney

Aidan Feeney

Queen's University

Search for more papers by this author
Balázs Fehér

Balázs Fehér

University of Nebraska-Lincoln

Search for more papers by this author
Brett K. Hayes

Brett K. Hayes

University of New South Wales

Search for more papers by this author
Christopher Kam

Christopher Kam

University of British Columbia

Search for more papers by this author
Jeffrey A. Karp

Jeffrey A. Karp

Brunel University

Search for more papers by this author
Gergo Vaczi

Gergo Vaczi

Eotvos Lorand University

Search for more papers by this author
Evan Heit

Evan Heit

University of California, Merced

Search for more papers by this author
First published: 13 February 2018
Citations: 5

Abstract

Categories are one of the primary ways by which people make sense of complex environments. For political environments, parties are especially useful categories. By simplifying political life, party categories enable people to make sense of politics. A fundamental characteristic of party categories is that they minimize perceived differences of members within a party (e.g., two Democrats) and maximize perceived differences between members of different parties (e.g., a Republican and a Democrat). In two-party systems, politicians in leftist parties will often be perceived as highly differentiated from politicians in right-wing parties. Yet, in multiparty systems there is greater complexity and potential for confusion since there are often multiple parties on the left and/or right. Spatial models of political competition predict that ideologically close neighboring parties will be perceived as similar, yet a categorical perspective holds that the public will perceive parties on the same side of the ideological divide to be dissimilar. In the present article, we review a research program investigating how political parties are treated as categories and present new data from seven democracies showing that people perceive parties to be highly differentiated regardless of where parties are located in ideological space.

Making sense of politics can be hard. For the ordinary person, politics is a menagerie of political actors and policies, concepts and ideas, not to mention an abundance of political communications. Thankfully, political parties provide a great deal of insight, helping organize a great deal of this. In two-party systems, the primary choice is often between a party on the left and a party on the right, but in multiparty systems there are often multiple parties on each side of the ideological divide. Parties on the same side of the ideological divide may offer similar programs, some nearly identical, making it especially difficult for the public to differentiate between them. Coupled with the well-worn finding that mass publics are minimally informed about politics (Converse, 1964; Sniderman, 1993), how do citizens differentiate parties on the same side of the left-right divide, especially between close ideological neighbors? In other words, how do like-minded parties maintain boundaries? Why don't parties occupying (nearly) the same ideological space blend into each other?

Answers to these questions, we believe, are provided by research on parties as categories. Categories are one of the primary ways by which people make sense of their world. By placing people into categories, stereotypes simplify how we perceive others by ascribing attributes of the category to all its members, minimizing differences between them. Not only do people minimize differences within categories but they also maximize differences between categories (e.g., Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976). Thus, merely labeling entities as category members highlights differences between them. Such insights are important to how people understand parties and the politicians that inhabit them.

Political scientists have long recognized the importance of parties offering distinct choices, yet research on parties has not been able to illuminate how largely uninformed mass publics are able to differentiate parties, especially parties that are ideologically similar. Since people tend to maximize perceived differences between categories, this means not only do we expect to find that people rate politicians from parties on the left and right as highly atypical of the other party but that they also rate politicians from different parties on the same side of the left-right dimension as highly atypical of the other party. Consequently, the party labels attached to politicians make them appear highly differentiated regardless of the type of party system (two-party or multiparty system) or, perhaps more surprisingly, where parties are located on the left-right dimension (close to each other or far apart). By accentuating differences between politicians, categorical reasoning helps illuminate how parties maintain boundaries, an especially helpful cognitive mechanism for voters confronted with multiple parties on the left or right. Absent these boundaries, parties with similar policies might fade into each other, making it especially difficult for citizens to make left-right party placements.

The studies featured here advance a program of research (Heit & Nicholson, 2010, 2016; Nicholson, Coe, Martinsson, & Heit, 2016) that to now has focused on the role of party categorization in the American political system, a textbook example of a two-party system. As part of this larger research program, here we present collaborative research involving data collection from seven democracies featuring two- and multiparty systems: Australia, Canada, England, Hungary, Northern Ireland, the United States, and Scotland. As found in Heit and Nicholson (2010, 2016), we found that participants perceived politicians on the left and right as opposites. However, this pattern was not only found for parties on opposite sides of the ideological continuum in two-party and multiparty systems, but here we report similar findings for parties on the same side of the ideological divide in multiparty systems. The results for two-party systems provide novel, but consistent, support for empirical research showing that voters in two-party systems are likely to perceive parties as highly distinct. The results for multiparty systems, however, provide new insight since we find that voters in multiparty systems perceive parties as distinct regardless of whether parties are ideological opposites or ideological neighbors. In presenting these new studies, we place them in the larger context of research on parties as categories that includes a discussion of the nature of categorical thinking, focusing on conceptual and empirical foundations from a specific research program on party categories to related work from political science, psychology, and cognitive science.

Party Distinctiveness

For many scholars, understanding party distinctiveness involves knowing how much distance is between parties on the left-right ideological dimension (Downs, 1957). The premise behind spatial models of voting is that the greater the ideological distance between two parties, the more likely it is that voters perceive those parties as distinct. To our knowledge, all research using survey data on these questions asks respondents to place themselves and parties along a left-right dimension. Although many people can place parties and themselves on a left-right scale, critics have argued that it is not how most voters think about parties (Conover & Feldman, 1981; Converse, 1964). Issue space in multiparty systems is often more complex than a single left-right dimension (Benoit & Laver, 2006) including not only multiple policy dimensions but also nonpolicy dimensions (Adams, 2001). Yet, even if people think about parties in ideological terms, they may not be able to accurately estimate where parties are located in issue space because the ideological positions of parties vary by policy area (Benoit & Laver, 2006) or because of perceptual biases (e.g., Granberg & Holmberg, 1988).

Despite these shortcomings, many scholars nevertheless use the left-right dimension because it travels relatively well across time and space (see Dalton, Farrell, & McAllister, 2011). Furthermore, many voters can place themselves and parties on a left-right scale (Inglehart & Klingemann, 1976) and use these placements (relative to their own) to inform their vote choice (Adams, Ezrow, & Somer-Topcu, 2011; Dalton et al., 2011). Taken together, the public's grasp of ideology appears to be relatively thin (at least compared to elites) but nevertheless meaningful in helping them understand party politics and make informed vote choices.

Missing from these accounts is the question of party distinctiveness. It is often assumed in studies of party systems using the left-right dimension that the further apart two parties are from each other, the more distinct voters perceive them to be. Successfully placing parties on the left-right continuum does not specifically address whether parties are perceived as distinctive. A person might judge two parties to be ideologically close to each other yet, at the same time, see them as highly differentiated. For example, someone might consider the categories “cats” and “dogs” as very similar (e.g., both are common household pets, have four legs, and fur) but also see them as highly distinctive (e.g., there is no debate whether a particular animal is a cat or a dog). Similarly, one might view two parties as very similar ideologically but also perceive them as highly distinctive. Categorical perception lies at the heart of this insight.

Parties as Categories

Our approach, grounded in research from cognitive psychology, begins from the premise that we make sense of the world by identifying objects as members or nonmembers of categories (see Murphy, 2002). Social categorization, also referred to as stereotyping, is the process whereby people classify people into categories based on a characteristic. Some of the most well-known research on stereotyping concerns racial prejudice (Allport, 1954). In Allport's (1954) highly influential research, he begins from the premise that “The human mind must think with the aid of categories…Once formed, categories are the basis for normal prejudgment” (p. 20). He concludes that, “We cannot possibly avoid this process. Orderly living depends upon it” (p. 20).

Rather than treating each new entity a person encounters as unique, people place them into categories (see Murphy, 2002). Categories help simplify understanding, allowing people to make inferences about the unique entity without having to learn everything anew. Upon encountering a unique dog, for instance, a person will make inferences about that dog based on what they already know about dogs generally (or categorically). Although we take such encounters with dogs for granted, our automatically doing this for many other entities, including other people, is often hugely beneficial. We understand the different dogs we meet, or dogs that we have never seen before, by our knowledge of the category of dogs. In placing all new dogs we see, or even just hear about, into a category, we minimize differences between dogs. The process of categorization also enhances perceived differences between objects. Despite the fact that cats and dogs are both domesticated, four-legged, furry animals that are common household pets, invoking the category labels of “cat” and “dog” maximizes perceived differences between the two. Similarly, a group of people attending a baseball game have a great deal in common, but when we identify one half as Yankees fans and the other half as Red Sox fans, we perceive them as highly differentiated.

Political parties are no different. Parties are social groups, and when people are categorized as members of a party, the attributes that define the category are ascribed to those individuals. As with other stereotypes, the attributes that define party categories allow people to make inferences about the individuals that belong to them. For example, in the American case, Rahn (1993) found that participants asked to identify candidates' policy positions largely ignored individuating information about candidates holding positions that were inconsistent with party stereotypes. Stereotypical thinking about a category can also lead to a caricature (a highly distorted image of category) rather than prototypical representations (the most typical member). Ahler and Sood (2016) found evidence of such distortions in perceptions of the social groups inhabiting political parties. For example, they found that the public thinks 32% of Democrats are LGBT (in reality it is only 6%) and that 38% of Republicans make over $250,000 a year (in reality it is only 2%). The attributes that define parties are also highly stable (Nicholson & Segura, 2012; Petrocik, 1996) but can, over time, change due to political circumstances (Bowler, Nicholson, & Segura, 2006; Carmines & Stimson, 1989).

Regardless of whether we refer to party stereotypes (Rahn, 1993), party images (Brewer & Stonecash, 2007), party brands (Aldrich, 1995; Cox & McCubbins, 1993), issue (Egan, 2013; Nicholson, 2005; Petrocik, 1996) and trait (Hayes, 2005) ownership, or party reputations (Goggin & Theodoridis, 2017; Sniderman & Stiglitz, 2012), the public is well-aware of the attributes that define and differentiate party categories (Abramowitz & Saunders, 2008; Carsey & Layman, 2006; Heit & Nicholson, 2016; Hetherington, 2001; Levendusky, 2009). Parties are perhaps the best-known categories inhabiting politics.

Essential to the study of parties as categories is research on categorical perception and minimal groups. Categorical perception, a phenomenon from perceptual psychology, refers to how category labels lead within-category differences to be minimized and between-category differences to be maximized (Goldstone & Hendrickson, 2010; Kelly & Heit, 2017). For example, suppose that X, Y, and Z are three patches of color, equally spaced in terms of an objective measure of color, namely wavelength of light. If X and Y are called “green,” and Z is called “blue,” then Y will seem more similar to X than to Z, despite, objectively speaking, Y being equally similar to X and Z. In terms of politics, merely labeling individuals as party members will make politicians within a party seem more similar while enhancing differences between politicians from different parties. In social cognition research, a related finding is known as minimal groups, in which grouping and labeling persons based on arbitrary and even trivial criteria, for example, random assignment to the X group or the Y group, still leads to strong beliefs about between-group differences, for example, ingroup favoritism and outgroup discrimination (Tajfel & Turner, 1979).

Note that categorical perception and minimal groups are relatively local, low-level phenomena, involving accentuation of boundaries between pairs of categories, and not necessarily a global restructuring of all knowledge. So, for example, if there are sharp perceived differences between politicians from two ideologically similar parties, that does not necessarily mean perceptions of other public figures will be affected. Revisiting the categorical perception phenomenon, if some colors are typical of green but not blue, and others are typical of blue but not green, there will also be colors, such as shades of red, that are considered typical of neither. In terms of political parties, some political figures will be typical of one party or another, but many public figures can be considered typical of neither.

There are additional psychological reasons to expect strongly perceived differences between parties. Davis and Love (2010) have shown that category differences are accentuated if a person chooses repeatedly between competing categories. Thus, if voters are frequently tasked with choosing between two parties on the right, they will perceive politicians from different right-wing parties as less similar. Having to make a choice between two similar options is at the heart of electoral politics. According to Bowler (1990), “party competition will take place and/or be at its fiercest between adjacent parties rather than between ostensibly opposing parties such as Christian Democrats and Communists” (p. 69).

In a series of articles, Heit and Nicholson and colleagues (Heit & Nicholson, 2010, 2016: Nicholson et al., 2016) apply insights from categorization research to how people perceive political parties. Applying the insight about how people maximize perceived differences between members of different categories, Heit and Nicholson (2010) found that the more a person (e.g., Barack Obama) is perceived to be typical of one party (e.g., Democrat), the less he or she is perceived to be typical of the other party (e.g., Republican). In their study of party categories in the United States, respondents rated public figures either on typicality as a Democrat or typicality as a Republican. The relation between the two sets of ratings was strong, negative, and linear, with a correlation of −0.9957. There was a complete trade-off between the two parties: To the extent that a politician was less typical of one party, the politician was equally more typical of the other party. Indeed, polarization was so extreme that the Democrat and Republican parties were treated as mirror opposites.

In a related article, Heit and Nicholson (2016) examined the question of how well voters understand party categories in the United States. Using data from the American National Election Study (ANES), they created profiles of hypothetical individuals that varied in terms of demographic characteristics (race, gender, number of children) and issue positions (social welfare spending and abortion). By varying the different attributes, they created profiles that varied (in increments of approximately 10%) in terms of objective probability of being a Democrat. They found that the public, including relatively low-knowledge voters, appears to be able to appreciate how different mixtures of policy issues and demographic characteristics alter the probability of a person's partisanship, even when the party cue itself is omitted. In two studies, partisans were able to accurately estimate how likely a candidate would be to belong to their own party versus the opposite party and also to make successful voting decisions about candidates, in terms of their own partisan interests.

In sum, the Heit and Nicholson (2010, 2016) articles demonstrated that in the United States two-party system, voters have a highly structured and systematic representation of the Democratic and Republican parties; indeed, one party is seen as the mirror image of the other. Even when party cues are omitted, such as in a nonpartisan election, voters are able to infer party membership accurately based on demographic and issue-based information, and vote according to the party line. At least in a two-party system, parties are well-described as categories that support a variety of political judgments and decisions.

Another critical feature of categories involves the level of categorization (see Murphy, 2002). Parties are basic-level categories, the type of category most frequently invoked when encountering some entity. Basic-level categories are the level of category taught to children. For instance, “dog” is a basic-level category. Upon seeing one, a child is taught to identify it as a “dog” rather than a “mammal” or a “Shih Tzu.” Mammals, of course, are higher level or superordinate categories whereas Shih Tzus are lower-order categories. The insight from categorization research is that as the level of categorization moves higher, it becomes more inclusive of a greater range of members. In looking at mammals, a superordinate category, cats, dolphins, bears, and rats, along with dogs, are viewed as much more alike than different, but as one moves to lower or more specific category levels, those same animals that were perceived to be more similar are now perceived to be more different.

Borrowing these insights, Nicholson et al. (2016) examined how the public perceives politicians depending on levels of categorization moving from the most specific, political party, to the most general, ideology. And for whichever party is in government, coalition is a level of category between party and ideology. The insight from categorization research they investigate is that as higher levels of categories are invoked, for example, moving from party to ideology, so are perceptions of similarity between politicians from different parties. As with many cognitive biases, they found that political categories help the public comprehend and navigate complicated party systems, but they also distort perceptions of how similar or different politicians are to each other. To test these insights, they conducted a survey experiment in Sweden, a crowded, multiparty system. They found that as higher-order categories are invoked such as “ideology,” the public views politicians from different parties on the same side of the ideological divide as similar whereas when lower-order categories such as “parties” are invoked, those same politicians are viewed as highly distinct. For example, politicians from the Sweden Democrats and Christian Democrats, both right-wing parties, are viewed as similar when evaluated in terms of the superordinate category of “right-wing politicians” but when viewed as parties, politicians from these same parties were viewed as highly differentiated.1

Taken together, this research program suggests that party categories are highly differentiated (Heit & Nicholson, 2010); the public is highly aware of what goes into defining party categories (Heit & Nicholson, 2016) and that in multiparty systems, people perceive politicians differently (more alike or dissimilar) depending on the level of categorization (Nicholson et al., 2016).

Here, we advance this research by examining whether, and how, the categorization of parties varies across party systems. In effect, we investigate how well the previous research in this program replicates in other world democracies. Is the United States unique in terms of the “mirror” phenomenon, or do citizens in other party systems also view their parties as opposites? Comparatively, the United States is uncommon in terms of having a long-standing two-party system. Would the polarization of party categories found in the United States manifest in multiparty systems? In particular, do perceptions of politicians in multiparty systems vary by whether their parties are on the same or opposite side of the ideological continuum? Comparing perceptions across parties provides insight into whether people view parties as similar or opposites or whether they perceive them to be altogether different.

In light of the preceding, we posit the following hypotheses.

H1: When looking at politicians from a pair of parties, politicians who are highly typical of one party will be highly atypical of the other.

In other words, across a set of politicians from two respective parties, there will be a strong negative correlation between typicality in one party and typicality in the other. Note that Hypothesis 1 predicts strong negative correlations regardless of the degree or type of ideological difference between parties. For example, there should still be a strong negative correlation when considering politicians from two parties on the left. Additionally, we also expect to find a strong negative correlation between left-right parties and niche parties, parties that do not compete on the left-right dimension but instead champion issues regarding nationalism, regionalism, or ethnicity (Adams, Clark, Ezrow, & Glasgow, 2006).

We also introduce a rival hypothesis, derived from Downs' (1957) theory of spatial party competition. In his theory, the public makes judgments about how alike two parties are relative to how close they are to each other in ideological space.

H2: The closer (farther away) parties are to each other on the ideological continuum, the more similar (dissimilar) the public finds them to be.

Setting aside the circumstance in which parties have converged toward the middle, for example, one center-left and one center-right, Downs predicts that voters will evaluate parties on opposite sides of the ideological continuum as dissimilar. This expectation is the same as Hypotheses 1, namely that typicality ratings of politicians in parties on the left and right, will be negatively related. In contrast to Hypotheses 1, however, Hypotheses 2 holds that the public will judge politicians from ideological neighbors as similar. In other words, Hypotheses 2 predicts a positive relationship between politicians from parties that are ideological neighbors whereas Hypotheses 1 predicts a negative relationship. Again, important to Hypotheses 2 is the assumption that parties on the same side of the ideological divide reside closer to each other than parties on opposite sides of the ideological divide (e.g., they do not converge toward the middle). This assumption is supported by the spatial locations of parties in our study and is a well-recognized empirical finding in the literature (Adams, 1999; Adams, Merrill, & Grofman, 2005; Grofman, 2004). If we observed two very closely situated centrist parties, one on each side of the ideological divide, the Downsian perspective would hold that they would be perceived as more similar to each other compared to extreme parties on the same side.

We also examine correlations between respective pairs of parties that include politicians from other parties as well as nonpoliticians. For example, to examine Hypotheses 1 and 2 for England, we look at the correlation between the Labour and Conservative parties only including politicians from each of those parties. In subsequent analyses for those parties, however, we not only include politicians from both parties but also Liberal Democrat politicians and nonpoliticians. The purpose of including politicians from other parties and nonpoliticians is to observe whether they fall into the same typicality structure. For example, if Liberal Democrats are more typical of the Labour party, are they less typical of the Conservative party? Or are they simply atypical of both? Likewise, as found in Heit and Nicholson (2010), we generally included nonpoliticians as stimuli to be rated, again with the purpose of observing whether nonpoliticians fall into the same typicality structure as politicians or are simply atypical of the political parties. Heit and Nicholson (2010) found in the United States that including nonpoliticians in analyses sometimes weakened the overall correlations because these individuals were seen as outside the party system. Here, we expect that including politicians from other parties, as well as nonpoliticians, is likely to weaken overall correlations in the analyses.

Party Experiments Around the World

For the analyses to follow, we conducted experiments in Australia, Canada, England, Hungary, Northern Ireland, the United States, and Scotland. The selected nations were chosen based on the variability of party-system characteristics such as the number of parties and types of issue cleavages as well as access to research participants. In other ways, the democracies are largely comparable (for example, five are parliamentary systems in the British tradition). All participants were students enrolled at universities in their respective countries and were given course credit for participation.

We begin by presenting results from the United States and Australia to test our hypotheses regarding the polarization of party categories in two-party systems. We also present results from Canada, England, Scotland, Northern Ireland, and Hungary to further examine expectations about the relationship between politicians from parties on the left and right in multiparty systems expecting to find the same strong, negative correlations found in two-party systems. In addition, as mentioned in the discussion of hypotheses, the multiparty systems allow for a novel test of whether politicians in parties that are close neighbors on the same side of the ideological divide are perceived as different (predicted by categorization theory) or similar (predicted by proximity theory) to each other.

General Method

A standard approach for understanding the nature of categories from cognitive psychology involves people rating how typical or representative an instance is of a given category ranging from high to low (Murphy, 2002). When rating the typicality of birds, for example, a person would likely rate a cardinal as a good example and a bat as a poor example. Similarly, when rating the typicality of “Democrat,” a person in the United States would likely rate Hillary Clinton as a good example and Donald Trump as a poor example. Studying party perception from the perspective of categorization allows people to decide which political figures are typical, or not, of a political party without imposing a standard of evaluation. All that is required is that individuals recognize public figures and rate how typical the person is of a party category without having to explain why.

The experiments involved participants rating how typical a list of public figures are of a given political party in the participant's country using a 7-point scale ranging from “very poor example” to “very good example” (see Appendix B for question wording). For example, in the U.S. experiment, we asked half the participants “How typical is George W. Bush of the Republican Party?” and the other half “How typical is George W. Bush of the Democratic Party?” The public figures included prominent party politicians as well as familiar nonpoliticians, primarily with recognizable political sympathies. Since participants were to rate how typical each public figure was of a given party, it was essential that we used highly recognizable public figures even if it meant excluding minor parties.

We use a between-subjects research design, for example, only one rating per public figure, with the aim of collecting high-quality data while minimizing potential fatigue or carry-over effects. All participants rated the same public figures, but different groups of participants were randomly assigned to provide ratings for each party. This means that we asked participants to rate how typical a given politician is not only of his or her own party but all other viable parties. In England, for example, all participants rated the same names including politicians from the Conservative Party, the Labour Party, and the Liberal Democrat Party. However, if they rated typicality in the Labour Party, they did not rate typicality in other parties. Participants also rated nonpoliticians, many of whom have political leanings. The nonpoliticians allow us to compare whether the results we see for politicians extend beyond party figures, offering insight into whether party categories encompass a wide swath of society or whether they only encompass politicians.

The data analysis relies on simple correlations between ratings of politicians for pairs of parties. Although hundreds of participants were involved, the N for each analysis is the number of political figures rated. For example, in the U.S. study, 97 participants rated how typical 15 political figures were of Democrats, and 98 different participants rated the same 15 political figures for Republican typicality. However, since the unit of analysis is the mean rating of each politician, N = 15, the number of public-figure ratings. Any responses by participants that were blank or otherwise did not follow instructions were treated as missing data and not entered into calculation of item means. An IRB or ethics committee in each country approved the studies, and informed consent was obtained from all participants (see Appendix A).

To aid the comparison with left-right placements, Figure 1 depicts citizen placements of parties on the left-right scale, based on survey data. From this scale, the public appears adept at making left-right distinctions, locating the Democrats in the United States on the left and the Conservatives in Great Britain on the right.

Details are in the caption following the image

Citizen placement of parties on the left-right scale (mean values). Data for Australia is from the 2010 Australian Election Study (AES); data for Canada is from Module 3 (2008) of the Comparative Study of Electoral Systems (CSES); data for Hungary and the United Kingdom is from the 2009 European Election Studies (EES); data for Northern Ireland is from the 2010 ESRC Northern Ireland Westminster Election Survey; and data for the United States is from the 2012 American National Election Study (ANES). Data for the LMP for Hungary are from the 2010 Manifesto Data collection since at the time of this study no nationally representative survey had included left-right placement questions for this party.

Using the left-right placements from Figure 1 as a basis for comparison, in the analyses that follow we present typicality ratings of politicians for parties. Recall that both the categorical approach and an ideological proximity model both predict negative correlations between politicians in left and right parties in either type of party system (two- or multiparty systems). In Hungary, for example, Fidesz, a party on the right, and the LMP (a Green party), a party on the left, will be viewed as highly dissimilar. In multiparty systems, however, an ideological proximity approach predicts a positive correlation between politicians from parties in close ideological proximity whereas Hypothesis 1, our categorical hypothesis, predicts a negative correlation. Returning to the example of Hungary, an ideological proximity approach would hold that typicality ratings of Hungary's leftist politicians in the MSZP and the LMP would be positively associated whereas a categorical approach holds that they will be negatively related. In other words, for the categorical approach, parties are always negatively related regardless of where they are located on the ideological continuum.

Two-Party Systems

To evaluate the hypotheses that people will judge parties as opposites, we replicated Heit and Nicholson's (2010) study of the United States and extend the investigation to Australia. However, we acknowledge that Australia is less of a pure case of a two-party system than the United States. For the United States, we present the methods used for illustrative purposes, but for Australia and all subsequent analyses, details such as the names of politicians are included in Appendix C and, where possible, in figures.

Study 1

United States

The participants were 185 students at the University of California, Merced surveyed during the spring of 2012. Ninety-seven participants made typicality ratings with respect to Democrat (the party on the left), and 98 participants made typicality ratings with respect to Republican (the party on the right). We used some of the same public figures from Heit and Nicholson (2010) but also updated a few. For example, we dropped John McCain, the GOP front-runner from 2008 and replaced him with Mitt Romney, the hands-down favorite to win the 2012 Republican nomination. In the present study, we asked participants to rate five Democratic politicians (Joe Biden, Jerry Brown, Bill Clinton, Hillary Clinton, Barack Obama), seven Republican politicians (George W. Bush, Herman Cain, Newt Gingrich, Sarah Palin, Ron Paul, Mitt Romney, Rick Santorum), and three nonpoliticians with recognizable political sympathies (Rush Limbaugh, Jon Stewart, Oprah Winfrey).

Figure 2 depicts the relation between mean typicality ratings for the Democratic and Republican parties (N = 15). To calculate the correlations, we take the mean typicality ratings of public figures from the Republican condition and correlate them with ratings from the Democratic condition. This yields a correlation matrix using mean typicality ratings from the respective parties. Providing support for Hypothesis 1 and Hypothesis 2 (recall they predict the same outcome in two-party systems), there is a strong negative relationship between typicality in one party versus the other, with an overall correlation of −.9869. Omitting the three nonpoliticians leaves the correlation essentially the same, −.9893. Overall, as found in Heit and Nicholson (2010), the results suggest that parties in the United States are viewed as opposite, polarized categories.

Details are in the caption following the image

Party typicality ratings for public figures in the United States. One hundred and eighty-two students participated in this study at the University of California, Merced in the spring of 2012. [Color figure can be viewed at wileyonlinelibrary.com]

Study 2

Australia

Given the polarization in contemporary U.S. politics, it is possible that these results do not generalize to other two-party systems. For this reason, we ran a comparable study in Australia. The participants were 257 psychology students at the University of New South Wales surveyed in the spring of 2011.

Figure 3 depicts a strong negative relationship between typicality ratings in the Coalition (right) versus the Labor party (left). In an analysis looking only at the 11 politicians, the correlation is −.9427, providing support for Hypothesis 1 and Hypothesis 2. The strong, negative correlation suggests that the Australian party system is very similar to the United States with regard to how people view the major left and right parties as opposites. Yet, in contrast to the United States, the Australia data shows that the relation between typicality in the Coalition and typicality in Labor is not as strongly linear. Most of the data points fall below the main diagonal. Further, the correlation including all 16 public figures in Australia (politicians and nonpoliticians) is −.8160, which is still strongly negative, although not as much as the sharply polarized United States. Note that for Australia, there were five nonpoliticians compared to three in the United States.

Details are in the caption following the image

Party typicality ratings for public figures in Australia. Two hundred and fifty-seven students participated in this study at the University of New South Wales in the spring of 2011. [Color figure can be viewed at wileyonlinelibrary.com]

Multiparty Systems

Our expectations for multiparty systems are similar to two-party systems with regard to parties on the opposite sides of the ideological divide. However, when looking at multiparty systems, ideological proximity models would hold that politicians from ideologically proximate parties (e.g., two neighboring leftist parties) would be generally viewed as similar (Hypothesis 2). In other words, Hypothesis 2 holds that correlations between typicality ratings for politicians in neighboring parties on the same side of the left-right continuum will be positive (e.g., two relatively close parties on the left). Alternatively, according to Hypothesis 1, the categorization hypothesis, we expect correlations for ideologically similar parties to be negative.

The other difference, as an empirical matter, for multiparty systems concerns correlations across all public figures, including politicians in other parties as well as nonpoliticians. For the United States and Australia, we were only able to include nonpoliticians. For multiparty systems, we had the opportunity, crucially, to include politicians from other parties in addition to nonpoliticians (except Hungary). Since we look at five multiparty democracies, we present each separately beginning with Canada.

Study 3

Canada

The participants were 242 political science students from the University of British Columbia surveyed in the autumn of 2012. In Canada, the Conservative Party is located on the right whereas the NDP and the Liberal Party are located on the left and center-left, respectively (see Figure 1). The relations between parties are shown in Figure 4, and pairwise correlations are shown in Table 1 for politicians in respective pairs of parties and for all 20 public figures. When looking at Table 1A, for the seven politicians who belong to either the Conservative party or NDP, the correlation in Table 1A is very strong, −.9658. This result is comparable to the results from the United States and Australia and provides support for Hypothesis 1 that people view parties on the opposite side of the ideological continuum as dissimilar. Furthermore, as mentioned, it is consistent with a Downsian approach (Hypothesis 2) since it also predicts a negative relationship. However, the correlations between adjacent parties (Conservative-Liberal: −.9276, and Liberal-NDP: −.9281) are also strongly negative providing support for Hypothesis 1 but not Hypothesis 2.

Details are in the caption following the image

Party typicality ratings for public figures in Canada. Two hundred and forty-two students participated in this study at the University of British Columbia in the autumn of 2012. [Color figure can be viewed at wileyonlinelibrary.com]

Table 1. Party Correlations for Public Figures in Canada
A. Correlations Across Politicians From Respective Parties
Conservative (right) Liberal (center-left) NDP (left)
Conservative (right)
Liberal (center-left) −.9276 (N = 8)
NDP (left) −.965 (N = 7) −.9281 (N = 7)
B. Correlations Across All 20 Public Figures
Conservative (right) Liberal (center-left) NDP (left)
Conservative (right)
Liberal (center-left) −.2437 (N = 20)
NDP (left) −.6083 (N = 20) −.1917 (N = 20)

Next, we conducted analyses including nonpoliticians as well as politicians from a third party. Inspection of Figure 4 for the Conservative versus NDP comparison suggests two groups of data points. First, there are data points for politicians from the Conservative and NDP parties, falling on the main diagonal in two groups. Second, there are politicians from other parties, and nonpoliticians, falling below the main diagonal. Indeed, a similar pattern holds for the other pairwise comparisons. The figure shows one set of data points on the main diagonal with the remaining data points falling below the main diagonal. Including all 20 public figures in the correlation between the Conservative party and NDP, the correlation remains negative, −.6083, but weaker compared to the correlation only including politicians from respective pairs of parties. Indeed, the correlations across all 20 public figures are particularly weak when comparing typicality in ideologically adjacent parties (Conservative-Liberal: −.2437, and Liberal-NDP: −.1917). For example, knowing that Stephen Harper is highly atypical of NDP makes it easy to predict that he is highly typical of Conservatives. But knowing that Gilles Duceppe (a retired Bloc Quebecois politician) is atypical of the NDP has unclear implications for whether he is typical of the Conservatives. In Canada, there does not seem to be a single structural dimension to represent all politicians as well as nonpoliticians.

Study 4

England

England is represented in the Parliament of the United Kingdom. At the time of the survey, the Conservative and Liberal Democrat parties formed a coalition government. Labour and regional parties also held seats. Since the regional parties are based in other parts of the United Kingdom (Scotland, Northern Ireland, and Wales), we focused on the three main parties for England. As seen in Figure 1, the Tories are on the right, Labour is on the left, and the Lib-Dems are slightly left of center.

The participants were 171 political science students from the University of Exeter, surveyed in March 2011. The categorical relations between parties are shown in Figure 5 whereas pairwise correlations are shown in Table 2 for politicians in respective pairs of parties and for all 25 public figures. When looking at Table 2A, for the nine politicians who belong to either the Conservative or Labour Party, the correlation is very strongly negative, −.9833. This result is comparable to the results from the United States and Australia. However, despite being in a coalition together, the correlation between Conservative and Liberal Democrat (−.9366) politicians is remarkably strong and negative. Finally, the correlation between Labour and Liberal Democrat, two left-of-center parties, is −.9522 providing support for Hypothesis 1 but not Hypothesis 2.

Details are in the caption following the image

Party typicality ratings for public figures in England. One hundred and seventy-one students participated in this study at the University of Exeter in the spring of 2012. [Color figure can be viewed at wileyonlinelibrary.com]

Table 2. Party Correlations for Public Figures in England
A. Correlations Across Politicians From Respective Parties
Conservative (right) Labour (left) Lib-Dem (left)
Conservative (right)
Labour (left) −.9833 (N = 11)
Lib-Dem (left) −.9366 (N = 12) −.9522 (N = 11)
B. Correlations Across All 25 Public Figures
Conservative (right) Labour (left) Lib-Dem (left)
Conservative (right)
Labour (left) −.8349 (N = 25)
Lib-Dem (left) −.3209 (N = 25) .0455 (N = 25)

Next, we conducted analyses including nonpoliticians and politicians from other parties. In Figure 5, the Conservative versus Labour comparison shows some data points for politicians from these parties falling on the main diagonal in two groups. Then there are politicians from other parties, and nonpoliticians, falling below the main diagonal. A similar pattern holds for the other pairwise comparisons. Including all 25 public figures, the correlation between the Conservative and Labour parties remains negative, −.8349, but weaker compared to the correlation only including the politicians from these parties (See Table 2B). In England, as in Canada, there does not appear to be a single structural dimension to represent all politicians as well as nonpoliticians which suggests people make clear distinctions between category and noncategory members.

Study 5

Scotland

Although part of the United Kingdom, Scotland has had its own national parliament since 1998. As shown in Figure 1, Scottish parties largely conform to the left-right dimension. The Scottish National Party (SNP), a left-of-center party devoted to Scottish independence, currently holds the majority in the Scottish Parliament. The left-wing Labour party, the centrist Liberal Democrats, and the right-wing Conservative party also hold seats, along with smaller parties. Although the SNP is a left-of-center party, its identity is clearly associated with promoting Scottish independence. Despite the fact that the SNP holds a majority of seats, it is most appropriately considered a niche party because its major identity is related to nationalism (Adams et al., 2006).

The participants were 334 political science students from the University of Strathclyde and the University of Glasgow in Scotland in June 2011. The relationships between Scottish parties are shown in Figure 6, and pairwise correlations are shown in Table 3, both for politicians in respective pairs of parties and for all 25 public figures. The Conservative versus Labour correlation in Table 3A for respective pairs of parties is most directly comparable to England. Here, the correlation is also extremely strong and negative, −.9900. However, all of the pairwise correlations in Table 3A are very strong, even between ideologically adjacent parties (Labour and Liberal Democrat), and even between the niche party, SNP, and the other parties. Hence, these results provide further support for Hypothesis 1, that members of different parties are perceived as opposites when party labels are invoked.

Details are in the caption following the image

Party typicality ratings for public figures in Scotland. Three hundred and thirty-four students participated in this study at the University of Strathclyde and the University of Glasgow in Scotland in spring 2012. [Color figure can be viewed at wileyonlinelibrary.com]

Table 3. Party Correlations for Public Figures in Scotland
A. Correlations Across Politicians From Respective Parties
Conservative (right) Labour (left) Lib-Dem (center-left) SNP (left)
Conservative (right)
Labour (left) −.9900 (N = 9)
Lib-Dem (center-left) −.8946 (N = 10) −.9382 (N = 9)
SNP (left) −.9787 (N = 8) −.9829(N = 7) −.9488 (N = 8)
B. Correlations Across All 25 Public Figures
Conservative (right) Labour (left) Lib-Dem (center-left) SNP (left)
Conservative (right)
Labour (left) −.7043 (N = 25)
Lib-Dem (center-left) −.3026 (N = 25) .1118 (N = 25)
SNP (left) −.5683 (N = 25) .0225 (N = 25) .0039 (N = 25)

The Scottish study also includes politicians from several different parties and a relatively high number of nonpoliticians. When focusing on all 25 public figures, the strength of the correlation for Conservative versus Labour drops to −.7043. When looking across all 25 public figures, the correlations depicted in Table 3B between typicality in one party and typicality in the other party are generally modest. With the exception of the moderately negative Labour-Conservative correlation—similar to England—most of these correlations are weakly negative or near-zero.

The pairwise comparisons between Conservative, Labour, and Liberal Democrat in Figure 6 resemble the patterns for England in Figure 5. Politicians from respective pairs of parties largely fall on the main diagonal, and other public figures fall below the main diagonal, being relatively atypical of both parties. The scatter plots for SNP versus the other three parties are particularly interesting. With the exception of SNP politicians and Sean Connery (a strong SNP supporter), no other public figures are considered typical of SNP. This finding is consistent with the notion that the SNP occupies a niche rather than falling on an ideological spectrum. As found in the other multiparty countries, parties in Scotland are differentiated, but this structure is not entirely consistent. For example, knowing that David Cameron is very atypical of Labour makes it easy to predict that he is very typical of Conservative. But knowing that Sean Connery or Alex Salmond are atypical of Labour has unclear implications for whether they are typical of Conservative. As found in the other multiparty systems in our study, people make clear distinctions between category and noncategory members in Scotland since each party has its own typicality structure, largely unrelated to typicality in the other party.

Study 6

Northern Ireland

Although part of the United Kingdom, Northern Ireland has had its own national assembly since 1998. There are four major parties, two Unionist (Democratic Unionist and Ulster Unionist) and two Nationalist (Sinn Fein and the Social Democratic and Labour Party or SDLP). As shown in Figure 1, Sinn Fein and SDLP are on the left and the unionist party is on the right. Despite these placements, the major division between unionists and nationalists is that unionists desire Northern Ireland to remain a part of the United Kingdom whereas nationalists desire a united Ireland. Mapped onto this dimension is a religious cleavage in which most Protestants identify as unionist and most Catholics identify as nationalist.

The participants were 335 psychology students from Queens University. In April of 2011, 59 participants made typicality ratings with respect to Sinn Fein, 57 participants made typicality ratings with respect to the Ulster Unionist party, 56 participants made typicality ratings with respect to the Democratic Unionist party, and 56 participants made typicality ratings with respect to the SDLP. As elaborated below, given findings from the initial study, we ran a subsequent study in the autumn of 2012 to examine typicality judgments of the superordinate party categories, unionist and nationalist. In this later study, 56 participants rated typicality with respect to nationalists, and 51 rated typicality with respect to unionists.

The pairwise relations between parties are shown in Figure 7, and pairwise correlations are shown in Table 4, both for politicians in respective pairs of parties and for all 19 individuals. First, we computed correlations for politicians from respective pairs of parties. As seen in Table 4A, most of the correlations are strongly negative, both across the ideological divide, for example, Sinn Fein versus DUP (−.9841), and for adjacent parties, for example, Sinn Fein versus SDLP (−.9837). The sole exception is the UUP versus DUP correlation, which is moderately positive, .3873. Hence the results mostly support Hypothesis 1, but there is some support for a left-right interpretation, which can explain the results for UUP versus DUP but not the results for Sinn Fein versus SDLP. Participants saw Sinn Fein politicians as the opposite of SDLP politicians but still found politicians who are typical of one unionist party to be somewhat typical of the other party. One possible explanation is that Catholics tend to be overrepresented at the university where this study was run, relative to the general population. Hence, nationalist parties would be seen as the ingroup, and more heterogeneous, and unionist parties would be seen as the outgroup, and more homogenous (Nicholson, 2012; Quattrone & Jones, 1980).

Details are in the caption following the image

Party typicality ratings for public figures in Northern Ireland. Three hundred and thirty-five students participated in this study at Queens University in spring of 2011. [Color figure can be viewed at wileyonlinelibrary.com]

Table 4. Party Correlations for Public Figures in Northern Ireland
A. Correlations Across Politicians From Respective Parties
Sinn Fein (left) SDLP (left) UUP (right) DUP (right)
Sinn Fein (left)
SDLP (left) −.9837 (N = 7)
UUP (right) −.9638 (N = 6) −.7512 (N = 7)
DUP (right) −.9841 (N = 7) −.9628 (N = 8) .3873 (N = 7)
B. Correlations Across All 19 Public Figures
Sinn Fein (left) SDLP (left) UUP (right) DUP (right)
Sinn Fein (left)
SDLP (left) .2026 (N = 19)
UUP (right) −.7463 (N = 19) −.3972 (N = 19)
DUP (right) −.6785 (N = 19) −.4782 (N = 19) .9294 (N = 19)

Turning to the analysis, including all 19 public figures in Table 4B, the correlations are less strongly negative than corresponding correlations in Table 4A. For example, the Sinn Fein versus DUP correlation, across all public figures, drops in magnitude to −.6785, and the Sinn Fein versus SDLP correlation is now mildly positive, .2026. Interestingly, there is now a strong positive correlation between typicality in UUP and typicality in DUP, .9294. It appears that there is a cognitive dimension of being a unionist that runs across all public figures, including nonpoliticians. This trend also appeared for nationalists, but more weakly.

To further examine the notion that there are overall cognitive dimensions of unionist and nationalist, or more likely, a single dimension spanning both categories, we conducted a subsequent study in which different participants in the spring of 2012 made typicality ratings in the superordinate categories of nationalist and unionist, that is, each public figure was judged on typicality either as a nationalist or a unionist. Figure 8 depicts the results. When restricting the analysis to politicians exclusively, the correlation is −.9713, a strongly negative association. Across all 19 individuals, there was still a strongly negative correlation between unionist and nationalist, weakening somewhat to −.8948. In this regard, Northern Ireland seems to better resemble the United States and Australia, the two-party countries in our study.

Details are in the caption following the image

Typicality ratings for Northern Ireland superordinate party categories. One hundred and nine students participated in this study at Queens University in spring of 2012. [Color figure can be viewed at wileyonlinelibrary.com]

Study 7

Hungary

Hungary was reestablished as a multiparty democracy in 1990, following the peaceful transition from state socialism to a free-market system. In 2010, after eight years of social-liberal governance, Fidesz and its partner, the Christian Democratic People's Party (KDNP), won a supermajority in the Parliament. In addition, two new, anti-status quo parties, the green Politics Can Be Different (LMP), and the radical nationalist Jobbik, won seats in Parliament. Following a pretest, we only include four parties in our study, leaving out the DK and the KDNP because participants were unable to correctly identify any of their politicians. Given that we wanted to keep the number of public figures rated roughly consistent across nations, we did not include nonpoliticians. In terms of the left-right dimension, Figure 1 shows Fidesz and Jobbik on the far right and MSZP and LMP on the left.

The participants were 207 political science students from Eotvos Lorand University in June of 2012. The pairwise relations between parties are depicted in Figure 9, and pairwise correlations are shown in Table 5, both for all 28 individuals, and just for politicians in respective pairs of parties. Beginning with Table 5A, pairwise comparisons between respective pairs of parties, the correlations are all strongly negative. As expected, the correlation between the two left-right parties, Fidesz on the right and MSZP on the left, is strongly negative, −.9093, a finding similar to other party systems. In addition, the pairwise comparisons between the two niche parties, LMP and Jobbik, are strongly negative, −.9436, suggesting that participants also viewed politicians in these parties as contrasting categories. Lastly, the correlations between ideological neighbors are also strongly negative between the two conservative parties, Fidesz and Jobbik −.8431, and the two liberal parties, LMP and MSZP, −.8468. The results support Hypothesis 1, but not Hypothesis 2, since all pairs of parties, not just those with opposite ideologies, have strong negative correlations.

Details are in the caption following the image

Party typicality ratings for public figures in Hungary. Two hundred and seven students participated in this study at Eotvos Lorand University in summer 2012. [Color figure can be viewed at wileyonlinelibrary.com]

Table 5. Party Correlations for Public Figures in Hungary
A. Correlations Across Politicians From Respective Parties
Fidesz (right) LMP (left) Jobbik (right) MSZP (left)
Fidesz (right)
LMP (left) −.9415 (N = 15)
Jobbik (right) −.8431 (N = 16) −.9436 (N = 7)
MSZP (left) −.9093 (N = 21) −.8468 (N = 12) −.9031 (N = 13)
B. Correlations Across All 28 Public Figures
Fidesz (right) LMP (left) Jobbik (right) MSZP (left)
Fidesz (right)
LMP (left) −.5197 (N = 28)
Jobbik (right) −.0123 (N = 28) −.1403 (N = 28)
MSZP (left) −.6558 (N = 28) .0203 (N = 28) −.4647 (N = 28)

Table 5B shows the same pairwise correlations, but including all 28 politicians. By including politicians from other parties in the pairwise party comparisons, we again are able to examine the nature of party categories. The results in Table 5B show considerably weaker correlations than those in Table 5A reported for respective pairs of parties. However, the most strongly negative pairwise correlation, −.6558, happens between a party on the left, MSZP, and a party on the right, Fidesz. The weakest correlations are between parties on the same side of the ideological divide. The correlation between the left parties, MSZP and LMP, is .0203 whereas the correlation between the conservative parties, Fidesz and Jobbik, is −.0123. The weak correlations suggest that participants largely view these categories as noncontrasting when evaluating politicians in all parties.

Figure 9 tells the same story as the correlations in Table 5. Due to the large number of politicians rated in each party, we only include the names of party leaders when the data are highly clustered. The typicality ratings of the politicians in these parties are not included in the correlation. In general, when looking at politicians in pairs of parties, being typical of one party is predictive of being atypical of the other. In each of the six graphs comparing typicality in pairs of parties, the remaining politicians who are not of those two parties are considered atypical of both, as if they do not fall on that dimension at all. Moreover, the two major parties, MSZP and Fidesz, do not seem to have a single dimension running across politicians from both parties. This point is most evident in the bottom-left panel of Figure 9 where we plot typicality in MSZP versus Fidesz. Politicians in these two parties do not fall on the main diagonal. MSZP politicians vary in typicality but hardly vary in typicality within Fidesz. Likewise, Fidesz politicians vary in typicality, but all are about equally atypical of MSZP. Hence, there does not seem to be a single structure corresponding to an ideological dimension across the two main parties. Instead, each party has its own typicality structure, largely unrelated to typicality in the other party.

Conclusion

People make sense of the world through categories. The mere act of identifying persons as members of categories accentuates differences between those individuals. We proposed that political party categories are no different. Regardless of where parties are located in ideological space, we predicted that people would view politicians from one party as highly distinct from another. Although our categorical perception hypothesis dovetails with Downs' theory when parties are on opposite sides of the left-right dimension (assuming that parties do not converge toward the center), it significantly departs from his prediction that citizens will view ideological neighbors as similar.

To examine whether and how people view parties as categories, Heit and Nicholson have carried out a line of work including several previous studies (Heit & Nicholson, 2010, 2016; Nicholson et al., 2016). Two of these articles (Heit & Nicholson, 2010, 2016) focused on the United States and showed that voters possess a highly structured representation of the two main political parties as opposites. Another article (Nicholson et al., 2016) investigates a multiparty system (Sweden) showing that the phenomenon of whether parties are treated as opposites depends on the taxonomic level of description. In the present work, we conducted experiments in seven democracies, aiming to replicate and extend the previous studies. In most cases, we found that the results were remarkably uniform, although there were a few notable deviations. The results from the United States and Australia, countries with two-party systems, suggest that people view parties as opposites. Although the correlation was somewhat stronger in the United States, the two Australian parties were largely viewed as mirror opposites. The more substantial difference involved nonpolitical figures. In the United States, whatever made a person more a Democrat made him or her less a Republican, and this held for political and nonpolitical figures alike. In Australia, this was somewhat less the case. Although we can only speculate, the difference may reflect the greater polarization of American politics (see Dalton, 2008) or the types of public figures chosen.

In multiparty systems, we also found that parties on the left and right were viewed as opposites. In England, Canada, Scotland, Northern Ireland, and Hungary, we found a strong, negative relationship between party figures on the left and right. This finding supports the notion that the public views parties as categories and is consistent with Downs' notion that parties that are on opposites sides of the ideological continuum would be perceived as highly differentiated (assuming parties do not converge toward the middle). However, when looking at ideological neighbors, we found that, contrary to Downs, participants in our study also largely viewed parties as opposites (with the exception of one pair in Northern Ireland). According to Downs, the public will view politicians from parties on the same side of the ideological continuum as similar given that they share many policy positions. Yet, with one exception, in the analyses restricted to politicians, the correlation for any two-party dyad was strongly negative regardless of where parties were located in ideological space.

Although our results contradict Downs' prediction about ideological neighbors being viewed as similar, they may provide insight into why the left-right continuum “works” despite an ill-informed public. Our results help explain why voters are able to distinguish parties in multiparty systems even under the most demanding of circumstances such as when parties are tightly bunched in ideological space. We believe that categorical reasoning is key to understanding the “boundedness” of parties, particularly why ideologically proximate parties can coexist and potentially thrive among publics that are not ideologically minded. Our results may also help explain why parties that have entered into a coalition government do not lose their identity to a broader, superordinate category. Coalition parties are perceived to be ideologically similar (Fortunato & Stevenson, 2013), but instead of blending into each other, they remain distinct because party labels encourage people to maximize between category differences even when parties offer identical policy.

Aside from the United States, we generally found that when analyses included nonpoliticians, or politicians from other parties, the correlations were weaker. This finding was robust despite differences in the types of nonpoliticians selected for each country, especially with regard to whether they had well-known political sympathies. Visual inspection of the figures shows that these individuals literally did not fall on the same line as politicians from a pair of parties being compared. When comparing parties X and Y as politicians from these parties are considered more typical of one party, they are considered less typical of the other. But this structure does not apply to politicians from party Z, or to nonpoliticians, who can be typical of neither party. The results from the United States are an anomaly and might reflect its sharply polarized politics but may also depend on the type of nonpoliticians we included, a group with identifiable partisan sympathies. We are reminded of the distinction, from cognitive psychology research, between local and global coherence (e.g., McKoon & Ratcliff, 1992). Consider the following sentences: “The apple is to the left of the banana. The carrot is to the right of the apple. The dandelion is in front of the banana.” Here, local coherence refers to getting pairs of relations correct, for example, where the carrot is relative to the apple. Global coherence refers to developing a consistent, multidimensional model for the whole system, resolving ambiguities and putting each entity in a specific location in relation to all other entities in terms of every dimension. We speculate that in multiparty systems, local coherence (maintaining a coherent representation of politicians deriving from contrasting pairs of parties) is possible, but achieving global coherence (maintaining a coherent representation of all individuals across multiple parties) is likely more difficult due to greater complexity.

An area for future research might explore the implications of categorical party thinking with regard to mass opinion formation. One of the central tenets of public opinion research is that elite discourse often shapes public opinion (Carmines & Stimson, 1989; Zaller, 1992). In a two-party system, the logic is simple: Elite consensus on an issue produces a public consensus whereas elite dissensus produces a divide in public opinion with liberals taking the position of politicians on the left and conservatives taking the position of politicians on the right. However, in a multiparty system, the public has a greater choice of elite cue-givers. In the absence of a cue from his or her own party, a person can take a cue on a policy issue from a politician from an ideologically compatible party. This logic would be consistent with Downs. Our data, however, suggests that since people view parties as highly distinct categories, regardless of where parties are located in ideological space, people would be likely to reject a cue from a politician not of their preferred party even if the politician was from an ideologically similar party. The other wrinkle would be whether a person would take a cue from an out-party politician who is in a coalition government with that person's preferred party.

The implications for democratic accountability are potentially substantial. One of the fundamental tenets of the “responsible party government” model is that voters are able to perceive clear differences between parties. In the absence of distinctive party choices, voters are rendered incapable of holding elected leaders accountable. If people perceive parties as categories, however, the target of accountability is in clear focus. Politicians cannot easily duck and cover, even if they are not to blame for their party's actions since people minimize differences within a category. However, categorical differences might potentially obscure accountability if the public holds parties, not ideologically like-minded politicians, accountable. Politicians from a party not in power, who nevertheless supported a disastrous ideological shift in policy, would likely not be held accountable since the public views parties, even close ideological neighbors, as highly distinct.

In sum, the public finds parties to be highly distinctive entities regardless of the type of party system or where parties happen to be located in ideological space. As institutions whose primary purpose is to link citizens to government, well-defined party boundaries are necessary for the public to hold politicians accountable, and our findings suggest that parties largely succeed by this measure.

Acknowledgments

Earlier versions of this article were presented at the 2013 annual meeting of the European Political Science Association, Barcelona, Spain and the 2011 annual meeting of the American Political Science Association, Seattle, Washington. We thank Jim Adams, Pablo Barberá, John Bullock, Tom Brunell, Dominik Duell, David Fortunato, Simon Jackman, Maria Newmen, Oksana Newmen, Peter Newmen, Shauna Reilly, Zeynep Somer-Topcu, and Alex Whalley for helpful comments and suggestions. We thank Ally Anderson for research assistance. This material includes work by Evan Heit while serving at the U.S. National Science Foundation. Any opinion, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. Correspondence concerning this article should be addressed to Stephen P. Nicholson, University of California, Merced. E-mail: [email protected]

    Footnote

  1. 1 Collins (2010) proposes that differences in the number of categories voters use for evaluating candidates has implications for whether they engage in directional (on my side) versus proximity (choosing the candidate closest to my preferences) voting. As the number of categories grow, he shows that the voter can make finer distinctions between candidates in a way that approximates proximity voting. Alternatively, if the number of categories reduces to two, the voter is likely to make either/or distinctions approximating directional voting.
    • The full text of this article hosted at iucr.org is unavailable due to technical difficulties.