Populism Is a Recurring Phenomenon in Democratic Politics and Has Been Studied for Many

Total Page:16

File Type:pdf, Size:1020Kb

Populism Is a Recurring Phenomenon in Democratic Politics and Has Been Studied for Many

Populist Attitudes and their Correlates among Citizens: Survey Evidence from the Americas

Kirk A. Hawkins Scott Riding Assistant Professor Brigham Young University Brigham Young University

In this paper we set forth a novel, survey-based measure of populist attitudes, and we use this data to explore populism at the individual level. The data are drawn from the 2008 AmericasBarometer surveys by the Latin American Public Opinion Project at Vanderbilt University, the 2008 Cooperative Congressional Elections Studies; and the 2008 Utah Colleges Exit Poll. We find that a coherent set of populist attitudes does exist among average citizens and is rather common. These attitudes are distinct from a variety of other measures of issue positions and attitudes, including support for strong chief executives. We also find that our measures correlate in predictable ways with such individual-level attributes as ideology, partisanship, education, wealth, and positions on immigration. However, populist attitudes fail to vary widely across countries with very dissimilar histories and contexts, and in a subsequent multilevel analysis, we fail to find strong results for any country-level predictors. This supports a theory suggesting that populist attitudes measured here are only a latent disposition, and that activation in the form of populist movements is the product of macro-level factors and is better gauged by other tools.

Draft only; please do not cite without permission of the authors.1 Prepared for the ECPR Workshop “Disassembling Populism (and Putting It Back Together Again): Collaborative

1 Correspondence may be addressed to the first author, at [email protected], 782 SWKT, Provo UT 84602 2

Empirical Research on Interactions among Populism’s Attributes,” March 22-25, 2010, Muenster, Germany. 3

As the Tea Party movement and reactions to the banking bailout in the United States remind us, populism is a recurring phenomenon in democratic politics. Scholars of both the developing and advanced industrial democracies have studied populism for years in an attempt to understand its causes and better define it (c.f. Ionescu and Gellner 1969; Betz 1994; Roberts 1995; Weyland 2001). Most studies assess populism at the aggregate level, by studying whole movements and their macro-level causes and consequences, or more rarely by studying key leaders. However, no one has ever attempted to study populist attitudes quantitatively among the general public. In this paper we develop a survey-based measure of populist attitudes and use the resulting data to explore both the concept of populism and its correlates at the individual level. Our measure reflects an ideational or discursive definition of populism, defined as the degree to which the public sees the political world as a struggle between a putative will of the people representing the side of Good and a conspiring elite representing the side of Evil (see de la Torre 2000; Mudde 2004; Laclau 2005). Here we present the results of surveys conducted specifically in Utah (where we work), the United States, and a large sample of countries in the Americas. We find that a coherent set of populist attitudes does exist among average citizens and is rather common. These attitudes are distinct from a variety of other measures of issue positions and attitudes, including a set of measures of support for strong executives. We also find that our measures correlate in predictable ways with such individual-level attributes as ideology, partisanship, education, wealth, positions on immigration, and support for the incumbent. Specifically, populist attitudes consistently diminish with wealth and higher education, and they are strongly associated with ideological extremism (rightism in the United States, leftism in most of Latin America), support for some third parties, support for incumbents (especially those who are populist), and (in the United States) anti-immigrant policies.

Definition and Measurement In this paper we treat populism as a set of ideas best characterized as a worldview or mindset: a set of basic assumptions about the world that individuals are rarely aware they hold. Specifically, populism is a Manichaean outlook that identifies Good with a unified will of the people and Evil with a conspiring elite (de la Torre 2000; Mudde 2004; Laclau 2005; Hawkins 2009). Seeing populism as a set of ideas has implications for measurement. If populism is a fundamentally a set of ideas, then it can be measured using techniques such as textual analysis and surveys. However, populism is not a set of conscious issue positions or an ideology, but a worldview or mindset: a set of attitudes so fundamental that most people are unaware that they hold it. How can we gauge a subconscious attitude? Here postmodernism provides us with an important insight. A number of scholars who analyze populism in ideational terms see it as a discourse, or a characteristic language and conversation intimately bound up in political action. For these scholars, populism is not simply a set of ideas held in the abstract, but a set of ideas held and communicated in words and other actions (de la Torre 2000; Panizza 2005; Laclau 2005). The important point is that discourses in this postmodernist sense are largely discerned through subtle cues of language and diffuse attributes of text such as tone and metaphor. If we are doing survey research on populism, this means we cannot rely on neutral statements of its core ideas but must instead develop questions that incorporate both the ideas and the language in which they are expressed. Based on these insights, we developed two slightly different measures of populist attitudes across three surveys or sets of surveys. The underlying design is the same in each: we 4 present a series of populist-sounding statements that roughly capture key ideas of populism, then ask respondents to what degree they agree with them. But the specific statements we use change somewhat across the surveys, due partly to considerations of language (English language versus non-English language surveys) as well as our attempts to improve the measures; in the end, though, all of our measures yield very similar results. First, working together with several colleagues at the Latin American Public Opinion Project (LAPOP) at Vanderbilt University, we developed a series of six statements. After testing and refining these measures in pilot surveys over the latter part of 2007 and early 2008, LAPOP deployed them in the 2008 round of the AmericasBarometer (hereafter, AB), a regional social science survey conducted approximately every other year. This version of the AB was conducted in 21 countries of the Americas, including the United States, Canada, Haiti, and Jamaica, as well as the familiar Spanish- and Portuguese speaking countries of the region. However, the full set of populism questions was not asked in Canada, and we are forced to drop this country from the analysis. The English version of the questions is as follows:

POP106. Our presidents [prime ministers] must follow the will of the people because what the people want is always right. How much do you agree or disagree with that view? POP107. The people should govern directly and not through elected representatives. How much do you agree or disagree? POP109. In today’s world there is a battle between good and evil, and people must choose between one of the two. How much do you agree or disagree that such a battle between good and evil exists? POP110. Once the people decide what is right, we must prevent opposition from a minority. How much do you agree or disagree with that view? POP112. The biggest obstacle to progress in our country is the dominant class or oligarchy that takes advantage of the people. How much do you agree or disagree with that view? POP113. Those who disagree with the majority represent a threat to the interests of the country. How much do you agree or disagree with that view?

In both surveys, respondents were asked to indicate whether they agree or disagreed with these statements using on a scale from one to seven, where one means “strongly disagree” and seven means “strongly agree.” These were not the only relevant questions in the survey. The AB 2008 included three others that gauge an alternative approach to populism that has gained traction among scholars over the past decade, a political-institutional one that sees populism as a political “style” combining charismatic leadership, poorly institutionalized organization, low esteem for institutions of representative democracy, and an emphasis on support from large numbers of voters—in short, a movement with charismatic leadership (Weyland 2001; Roberts 2003; Barr 2009).2 In the questionnaire, these questions appear immediately before our own. While we (the authors) did not participate in the design of these questions, they provide an important point of comparison for our initial descriptive analysis, and we include their results in the next section. They use the same 7-point scale of agreement and are worded as follows:

2 Technically, Barr (2009) and a few others (see for example Deegan-Krause and Haughton 2009) combine populist ideas with charismatic leadership in their definitions. 5

POP101. It is necessary for the progress of this country that our president [prime minister] limits the voice and vote of opposition parties. How much do you agree or disagree with that view? POP102. When the Congress hinders the work of our government, our presidents [prime ministers] should govern without the Congress. How much do you agree or disagree with that view? POP103. When the Supreme Court [Constitutional Tribunal] hinders the work of our government, it should be ignored by our presidents [prime ministers]. How much do you agree or disagree with that view?

While the AmericasBarometer questions yielded sensible results that we describe in the next section, we felt dissatisfied with the feel of some of them, especially in English. Working with our colleagues at Brigham Young University, we redesigned a shorter series of four questions with stronger populist language. We tested and refined these measures in pilot surveys over the latter part of 2007 and early 2008, then deployed them in two large, representative surveys conducted around the U.S. general election in November 2008: the Utah Colleges Exit Poll (hereafter UCEP), and the Cooperative Congressional Elections Survey (hereafter CCES). The statements are the same in both surveys:

POP1 Politics is ultimately a struggle between good and evil. POP2 The politicians in Congress need to follow the will of the people. POP3 The power of a few special interests prevents our country from making progress. POP4 The people, not the politicians, should make the most important policy decisions.

In both surveys, respondents were asked to indicate whether they agreed or disagreed with these statements. In the UCEP, respondents were given a 5-point Likert-type scale with labels only for extremes: 1=strongly disagree, 5=strongly agree; in the CCES, for reasons given below, a 4- point scale was used instead with labels for each response. In order to gauge the distinctness of populist discourse, in these U.S. surveys we developed and/or tested two additional, related modules of questions. The first was a series of three questions designed to gauge pluralist discourse. While the results were instructive and help strengthen our confidence in the populism questions, this was an experimental series included only in the Utah survey. The statements are as follows:

PLU1 Democracy is about achieving compromise among differing viewpoints. PLU2 When our opposition presents new and challenging viewpoints, there is something we can learn by listening. PLU3 Freedom depends on diversity.

In the CCES, we instead deployed a set of four questions originally designed by Hibbing and Theiss-Morse (2002) to gauge “stealth democracy” in the United States. Their concept is highly analogous to populism and has received considerable attention in the study of U.S. politics. Hibbing and Theiss-Morse argue that Americans display strong anti-government tendencies that seem to justify participatory democracy, but that these attitudes are paired with a 6 preference for technocratic decisionmaking by disinterested experts and political outsiders, rather than personal involvement by citizens. The stealth democracy questions are as follows:

SD1 Elected officials would help the country more if they would stop talking and just take action on important problems. SD2 What people call ‘compromise’ in politics is really just selling out on one’s principles. SD3 Our government would run better if decisions were left up to successful business people. SD4 Our government would run better if decisions were left up to non-elected, independent experts rather than politicians or the people.

We share the concerns of some of Hibbing and Theiss-Morse’s critics that the coherence of this set of ideas have not been rigorously tested (Hibbing and Theiss-Morse fail to conduct any factor analysis of their results, for example), and, as scholars trained in comparative politics, we suspect that much of what they measure is tapping into populist discourse. Specifically, we expect that the first two of their questions—which incorporate remarkably populist language— will load onto the same underlying dimension as ours, while the latter two will capture a distinct set of attitudes. In order to make our data from the CCES as comparable as possible with Hibbing and Theiss-Morse’s reported results, we used their original four-point scale of agreement (1=strongly disagree, 2=disagree, 3=agree, 4=strongly agree), rather than the 5-point Likert-type scale found in the Utah poll. As we show below, however, our results across the two surveys turn out to be very similar and allow us to make sensible comparisons with moderate rescaling.

Descriptive Results While we reserve a discussion of some of our causal predictions for the next section, we want to preview one of the most important predictions right now because the descriptive results will immediately present readers with relevant data. Specifically, we expect to find high aggregate levels of populist attitudes among the general public. This hypothesis may seem counterintuitive because successful populist movements are rare, even in Latin American countries. However, theorists who study populism as a set of ideas see it as an inevitable facet of democratic values everywhere (Canovan 1999; Panizza 2005). According to this view, populism is merely the redemptive counterpart to our pragmatic, pluralist view of democracy that normally prevails in public discourse. Populism rarely manifests itself because it depends on the interaction of personality with systemic phenomenon, such as the presence of moral threat and the availability of credible populist leadership. But the populist template is stored in most of us, and in the contrived context of a survey of public opinion that gauges dispositions and attitudes, it should be easy to find.

AmericasBarometer Now the results. The AmericasBarometer 2008 was conducted using nationwide, nationally representative samples of 1,500 respondents (in a few countries, 3,000 respondents). It is a face-to-face survey done at the home of respondents. The survey is long (slightly under an hour) and includes questions on a variety of political and sociodemographic attributes and experiences. The bulk of the surveys were conducted in the first few months of 2008. Survey 7 workers were nearly always nationals of the country hired by private contractors, but they were closely trained and monitored on-site by LAPOP representatives. Table 1 reports average values for each of the two populism modules after rescaling them to a 100-point scale where 100 indicates the highest level of agreement and 0 the lowest; this scale will facilitate comparison across surveys. While the three political-institutional measures generate only moderate support across the entire sample, always below 40, most of our own questions generate much stronger support. With the exception of POP107, all of our measures generate average scores over 40, and in most cases over 50. Given the roughly unimodal, symmetric distributions of all of these questions, this means that a majority of respondents in each country can be characterized as having populist attitudes. This confirms our first hypothesis about the prevalence of populist attitudes. By way of note, the standard deviations are essentially identical across all of the questions, in the range of 31.0-33.0, suggesting quite a bit of spread across the entire multicountry sample. As we will show at several points, most of this variance is actually within countries rather than across the region.

Table 1 Mean Responses to AmericasBarometer Populism Modules (rescaled 0-100)

Variable Mean POP101 (limit opposition parties) 39.0 POP102 (govern without Congress) 37.3 POP103 (ignore Supreme Court) 35.0 POP106 (follow will of people) 67.2 POP107 (people govern directly) 39.1 POP109 (choose good or evil) 68.2 POP110 (prevent opposition) 52.1 POP112 (dominant class) 61.5 POP113 (disagreement is threat) 45.2

Notes: All countries weighted equally. N ranges from 33,981 to 38,207, and standard deviation from 31.7 to 33.4. Source: AmericasBarometer 2008.

In order to examine the concomitance of all of the survey questions, both our ideational module and the political-institutional module, we performed a factor analysis on the entire populism series. The results are reported in Table 2. The first apparent characteristic of these questions is that they sort into two underlying factors, as illustrated by the highlighted fields in the table. Five of our six ideational questions (POP106, POP 109-POP113) consistently load together on one factor, while the three political- institutional measures (POP101-POP103) all load heavily on a separate factor and are complemented weakly by POP107 (“the people should govern directly”). Both factors load evenly across questions, suggesting that no one instrument dominates either sequence.

Table 2 Principal Component Analysis of Populism Series, 2008

Variable Factor 1 Factor 2 Uniqueness POP101 0.74 0.06 0.45 8

POP102 0.84 0.07 0.31 POP103 0.82 0.06 0.33 POP106 0.14 0.60 0.62 POP107 0.47 0.29 0.69 POP109 -0.06 0.68 0.53 POP110 0.27 0.63 0.52 POP112 -0.01 0.67 0.55 POP113 0.31 0.56 0.58

Eigenvalue 2.86 1.54

Notes: Principal component factor analysis with an orthogonal varimax rotation. 17 parameters, 2 retained factors, n = 26,704. Majority loadings are highlighted for readability. Source: Americas Barometer by LAPOP.

This factor classification gives us confidence in the coherence of our measures and suggests that populist discourse captures an interrelated set of ideas. It also suggests that the support of unilateral executive power associated with the political-institutional definition is not an essential aspect of populism, at least not at this individual level. Some people with strong populist attitudes also support unilateral executive power, while others with strong populist attitudes do not. The presence of charismatic leadership in successful populist movements is not a function of populist attitudes. Interestingly, POP107 has an unsubstantial relationship with the ideational populism factor and only a weak connection to the executive supremacy factor. In individual country-level principal component analyses (not shown), POP107 fluctuates between the two factors, but never strongly associates itself with either. And as we saw earlier, it is also the question in our module with the weakest average support. This question was added to our set much later in the design of the survey and was meant to be a crossover between ideational and political-institutional perspectives. The results here suggest that the question ended up capturing the latter better than the former; consequently, we drop it from further analyses. Because of the large number of questions and countries, we avoid presenting average results across individual questions/countries. However, using the factor analysis as a reference, we produce an additive populism index from the second factor. The index is simply the mean value across all five of the remaining ideational populism questions ([POP106 + POP109 + POP110 + POP112 + POP113]/5). Figure 1 is a cross-national comparison of the resulting statistic showing the average value for each country. [Authors’ note: our apologies for not replicating this graph with the 100-point scale; this is the original 7-point scale from the AB survey.] Excluding the cases of Argentina and the United States, which have distinctly low values (3.5), there are moderately high levels of populist discourse in all of our countries in the sample. As already mentioned, the distribution within each of our countries is roughly normal and ranges across the scale, with an average standard deviation of 1.2. Again, this suggests that populist attitudes are common everywhere at the individual level, though not perfectly ubiquitous.

Figure 1 Comparison of Populism Index across Countries, 2008 9

The pattern of variance across countries also tells an interesting story. Excluding again the cases of Argentina and the United States, the cross-country variance is rather minor. All remaining 19 countries fall between 4 (Haiti) and 5 (Dominican Republic) on a scale of 1-7. While the average within-country standard deviation is 1.2, the average standard deviation of the scores across countries is just 0.4. The weak pattern that does exist is rather surprising. Many of the rankings are intuitive, as in the case of Bolivia under Morales. But other countries defy what we have suggested are the expectations of traditional approaches, such as the (relatively) low populism among the general public in Venezuela under Hugo Chavez, or the even lower populism in Argentina with its long history of populist leaders. These data lead to two implications: 1) the presence of a populist leader does not a priori predict strong populist sentiment among the voters, and 2) populist tendencies among voters may not a priori predict the election of a populist leader. Again, these findings seem to support the discursive expectation, which sees populist discourse as a fairly universal phenomenon that in the abstract context of a survey may not vary significantly across countries; the important difference is between latent and mobilized sentiment, which we cannot really test here. At the very least, the results reaffirm that as we try to explain variance in populist discourse across respondents, individual-level determinants are likely to prevail.

U.S. surveys 10

Like the AB, the UCEP is a face-to-face survey, but it is considerably shorter (5-10 minutes) and is administered as an exit poll just outside voting area. Respondents are chosen by student survey workers stationed at the site and are asked to fill out their own survey form by hand. The 2008 survey consisted of a statewide sample of 950 respondents and was conducted during the November 2008 general elections. In contrast, the CCES is an Internet survey designed by a consortium of political scientists from across the United States and conducted by Polimetrix. The survey is administered to over 20,000 respondents who receive a common core of questions; nationally representative subsamples of 1,000 respondents each receive additional questions designed by contributing teams of scholars. The populism and stealth democracy questions were administered to one of these subsamples. Table 3 reports average values for each question after rescaling them to our common 100- point scale, where 100 indicates the highest level of agreement and 0 the lowest. The results are fairly similar across the two samples, showing that the bulk of respondents agrees (strongly or mildly) with all four populism questions. In fact, the level of agreement is higher than what we found in the AmericasBarometer. In both surveys, the agreement is strongest for the latter three questions in the module (POP2-POP4), between 66.9 and 83.7; it is lowest for POP1 (“politics is ultimately a struggle between good and evil”), around 43.9-51.6. We find it remarkable that a set of questions with stronger populist language generates such a high level of agreement, especially in a context where successful populist movements are relatively rare. The fact that these basic results are similar across both the Utah and U.S. surveys gives us added confidence in their reliability. The results of individual questions vary somewhat across samples: Utah respondents appear slightly more populist on the second and fourth questions, while the national sample is more populist on the first and third questions. Thus, Utah respondents are not especially different from the rest of the country, suggesting—as we saw in the AmericasBarometer—that variance across samples is much less than variance within samples. Table 3 also presents mean values for the questions in our other modules: the pluralism questions and the stealth democracy questions. Again, these have been converted to a 100-point scale (0=strongly disagree and 100=strongly agree). The respondents (Utah only) generally agree with the pluralist statements, yielding average scores between 68.6 and 82.3, about the same levels as we found for the populist statements in these surveys. This suggests support for both pluralism and populism, although we still need to test the correspondence between these sets of responses.

Table 3 Mean Responses to U.S. Surveys (rescaled 0-100)

Variable UCEP 2008 CCES 2008 POP1 (good and evil) 43.9 51.6 POP2 (will of the people) 83.7 79.1 POP3 (special interests) 71.5 75.5 POP4 (people make decisions) 71.0 66.9 PLU1 (democracy is compromise) 74.6 -- PLU2 (listen to opposition) 82.3 -- PLU3 (freedom is diversity) 68.6 -- SD1(stop talking/take action) -- 72.0 11

SD2 (compromise is selling out) -- 55.0 SD3 (successful business people) -- 38.1 SD4 (independent experts) -- 40.0

Notes: N ranges from 911 to 928 in the UCEP and is 1,000 in the CCES; standard deviations range from.20.6 to 31.1 in the UCEP and XX in the CCES. Source: 2008 Utah Colleges Exit Poll; 2008 CCES.

In the case of the stealth democracy questions (CCES only), we see results suggesting that our guesses about the relationship between this concept and populism are correct. While respondents give an average score of about 72.0 for the first statement (“stop talking/take action”), and a somewhat smaller 55.0 for the second statement (“compromise is selling out”), they score much lower—38.1 and 40.0 respectively—on the last two statements. This suggests that the first two and latter two questions are indeed tapping into different sets of attitudes. This pattern of responses is remarkably similar to what Hibbing and Theiss-Morse found in their original study using national survey data from 1998. They reported average levels of support for these statements of 69.3, 55.3, 42.3, and 41.7 on a 100-point scale.3 As we did with the AmericasBarometer questions, our next step is to run factor analyses of the populism module in Utah and the United States, together with the pluralism questions in the UCEP and the stealth democracy questions in the CCES. For good measure, in the CCES analysis we include four additional questions gauging issue stances on abortion (CC310, higher numbers are more pro-choice), protecting the environment (CC311, higher numbers mean favoring jobs over environment), privatizing social security (CC312, higher numbers mean opposition to privatization), and affirmative action (CC313, higher numbers mean opposition to affirmative action). Knowing that the populism questions load onto a different factor from these issue stances will give us added confidence that our measures are capturing a truly distinct set of attitudes. The results of these analyses are in Tables 4 and 5. To begin with, both analyses again show that our populism questions hang together. With the partial exception of our first question (POP1, “good versus evil”), the populism questions all load heavily onto the same factor. In the CCES, this factor clearly stands apart from a separate factor gauging salient issue stances. This repeats the findings from the AmericasBarometer and provides added reassurance that our measures are getting at a coherent underlying set of ideas and language.

Table 4 Factor Analysis of Populism Series, UCEP 2008

Variable Factor 1 Factor 2 Uniqueness POP1 -0.32 0.75 0.34 POP2 0.41 0.57 0.51 POP3 0.36 0.55 0.57 POP4 0.33 0.67 0.44 PLU1 0.68 0.17 0.51

3 According to their original 4-point scale of agreement, Hibbing and Theiss-Morse find that the percent who strongly agree and agree with these statements is 86, 60, 32, and 31 percent. Using this same scale, we have results of 83, 56, 33, and 32 percent. 12

PLU2 0.75 0.37 0.44 PLU3 0.74 0.01 0.46

Eigenvalue 2.46 1.27

Notes: Principal component factor analysis with an orthogonal varimax rotation. 13 parameters, 2 retained factors, n = 882. Majority loadings are highlighted for readability. Source: 2008 Utah Colleges Exit Poll.

Table 5 Factor Analysis of Populism Series, CCES 2008

Variable Factor 1 Factor 2 Factor 3 Uniqueness POP1 0.45 -0.23 -0.40 0.59 POP2 0.79 -0.09 0.02 0.37 POP3 0.63 0.22 -0.04 0.55 POP4 0.74 -0.13 -0.08 0.42 SD1 -0.55 0.07 0.28 0.62 SD2 -0.51 0.34 0.36 0.50 SD3 -0.03 0.39 0.69 0.38 SD4 -0.08 -0.14 0.83 0.28 CC310 (abortion) -0.04 0.67 0.01 0.55 CC311(environment) 0.14 -0.68 -0.05 0.51 CC312 (Soc. Security) 0.03 0.70 0.12 0.49 CC313 (affirm. action) 0.17 -0.66 0.05 0.53

Eigenvalue 2.38 2.27 1.55

Notes: Principal component factor analysis with an orthogonal varimax rotation. 33 parameters, 3 retained factors, n = 943. Majority loadings are highlighted for readability. Source: 2008 CCES.

Beyond this basic finding, the analysis of the UCEP data (Table 4) presents a somewhat complicated picture. While the populism questions load most heavily onto a separate factor, they also load somewhat onto a factor that brings together the three pluralism questions as well. This discounts any notion that populism and pluralism are incompatible, showing that there is at least some overlap between these two sets of ideas in the minds of many citizens. However, the fact that the populism questions load more heavily on a separate factor says that these two sets of ideas are still largely distinct. Thus, populism and pluralism apparently coexist in the minds of individual citizens, but they are not the same thing. The results of the CCES factor analysis (Table 5) are much clearer. The populism questions partly overlap with the stealth democracy questions in the way we earlier predicted: with the partial exception of POP1, populism loads heavily on the same factor as the first two stealth democracy questions, while the latter two stealth democracy questions constitute their own factor. The concept of stealth democracy does not seem to reference a coherent set of ideas 13 for Americans, but may partially capture populist sentiments that are common among the public. We think this shows that populism is the better concept for categorizing attitudes towards democracy.

Correlates at the Individual Level Now that we have what appear to be three, related sets of individual-level data on populist attitudes, we can turn to our second objective: analyzing the correlates of populism at the individual level. Here we face two challenges. While we are interested in identifying likely causes of populist attitudes, we need to recognize that some putative causal relationships are very complicated, and in some instances we can only talk about expected associations. Our statistical models include a mixture of sociodemographic and partisan attributes that are not entirely exogenous. Thus, we stop short of providing a clean, comprehensive causal model. Second, with some important exceptions we note below, most theories of populism are aggregate-level theories. In thinking about how these translate into individual-level behavior and attitudes, we must beware the ecological fallacy. Just because the charismatic leader of a successful populist movement has a particular discourse does not mean that all of his followers share this; they may follow him for unrelated reasons (de la Torre 2007). Likewise, the systemic factors that spark populist movements may not plant new populist ideas in the heads of citizens, but merely arouse existing sentiments shaped more by personality and socialization. Let us suggest several individual-level correlates and a bit of the reasoning behind them. First, as already mentioned and tested, we expect to find high levels of populist attitudes among the general public. The idea is that the activation of populist attitudes in the form of actual discourse and mobilization by a populist movement is rare, but that the basic attitudes are fairly common and innate to democracy. This naturally leads us to ask what those triggers might be. In keeping with our earlier arguments (Hawkins 2010) as well as those of political psychologists who study authoritarian attitudes (Stenner 2005), we suggest that populist movements arise as a response to normative threat: a sense that one’s fundamental identity and values are being challenged, whether by increasing diversity of values within the community or by the corruption of public officials. Unfortunately, because our survey measures gauge underlying attitudes— something close to a disposition—rather than actual everyday talk and action of respondents, they provide poor tests of ideational or discursive theories of populist movements. Underlying dispositions are unlikely to correlate much with aggregate trends or country-level attributes, such as perceived corruption, levels of ethnic pluralism, or economic crisis, because they simply do not vary much with short-term situations or external factors. Explaining the activation of populist attitudes requires a different research design. Our second prediction is that populist attitudes should be more prevalent among the poor and less educated. This is of course is an old expectation in the literature on populism. For scholars more inclined to view populism positively, the reason for the correlation is rational and instrumental. The populist message defends a defrauded majority and exalts the virtue of common folk. Being the majority, the common folk are by definition not the best educated or the most wealthy; in turn, the wealthy and well-educated know they are potential targets of populist wrath and unwilling to condemn themselves. For political psychologists and scholars who take a negative view of populism, the explanation is more cognitive. For these, populism is essentially a result of ignorance. According to traditional approaches, it is only the poor and the ignorant that would be willing to support short-sighted economic policies and cast themselves at the feet of politicians who claim to embody their will. At best populist leaders are demagogues without real 14 ideologies; at worst they are charlatans (di Tella 1965). Indeed, studies by political psychologists show that low education and poverty are strong predictors of authoritarian personality, a trait closely related to the Manichaean outlook of populism (Stenner 2005; other source here). And recent experimental research and surveys suggest that these attitudes are closely related to low cognitive capacity (Stenner 2005). Populism should also be associated with specific ideological tendencies. To wit, populism should be associated with ideological extremism, and its ideological stripe will be correlated with the level of economic development in the polity. Populism is a worldview that can accommodate numerous ideologies, but certain ideologies have greater affinity for the discourse. Populism is of course the language of radical change and embodies a demand for systemic transformation of the polity—revolution, not policy change. On any left/right, conservative/liberal ideological scale, those with the strongest populist attitudes are more likely to be on the extremes. Yet, where will populist tend to fall, on the right or on the left? Here our explanation is more sociological. In economically developed countries, the middle class—a property owning class—is likely to be the largest segment of the population; these are stakeholders in the economy (usually a capitalist one) and much less likely to seek radical reforms of property rights in their attempts to revolutionize the system. In contrast, in poor countries the poorest groups are also the largest; these are generally not property holders and are thus much more likely to be in favor of radical redistribution of wealth in order to achieve a more equitable, just system. Consequently, we argue that populist attitudes are more likely to be on the right in countries where levels of economic development are high (e.g., the United States), and they are likely to be on the left in countries where levels of economic development are low (for example, most of Latin America). Regarding partisan affiliations, the Latin American sample of the AmericasBarometer is too diverse to make or test any predictions. Partisan identity is often low in each country, and our area-specific knowledge is too limited for us to make predictions in more than a few countries. However, with the U.S. surveys we can make some predictions. While populist attitudes should be associated with greater support for ideological extremes, in terms of partisan tendencies this does not necessarily imply support for traditional parties of the extreme right/left. Populism condemns the existing elite as corrupt, and it often assumes the form of third-party movements that can play the role of outsiders. Thus, in the American context with its two-party system, individuals with populist attitudes are more likely to support (indeed, to constitute) third-party or outsider movements. We can also say something about specific issue positions. Again, the Latin American sample in the AmericasBarometer is too diverse for us to say much here, but in the United States we have a strong expectation that populist attitudes will be strongly associated with anti- immigrant stances. This is a fairly commonplace assumption in the United States and especially in Europe, where populism is almost equated with the radical right. Using survey data, Ivarsflaten (2008) documents the strong association of anti-immigrant positions with support for right populist parties in Europe. In fact, anti-immigration stances are the one issue that seems to unite all right-populists in Western Europe. Here, though, we can test whether anti-immigration stances are associated with populist attitudes, rather than just partisan affinity. Because variance in populist attitudes across countries is so low, we do not expect to find that country-level predictors make much difference. Indeed, several tests with country level- predictors yielded negative results, although we omit most of the results for the sake of brevity. 15

The one exception to this is the interaction between populist attitudes and the socioeconomic context (developed, less-developed), where we include a country-level measure that turns out to have a large effect. However, we do have a few individual-level measures that are analogous to the country- level causes of populist movements, and we bring them in to better identify how these relationships might operate at the individual level. All of these are measures are available only in the AmericasBarometer. First we suspect that populist attitudes will have some relationship to each individual’s experience with crime and corruption; those individuals who have the worst experiences with failed government may also find themselves willing to believe populism’s message of a conspiring, morally debased elite. However, we cannot really be certain about the direction of causality at this level of analysis. It may be things work the other way, and that individuals with populist attitudes are more likely to perceive corruption and ineffective government where they do not objectively exist. If populist attitudes are more basic and the driving force here, then we expect to find a strong correlation between populist attitudes and perceived corruption and crime, but not with experienced corruption and crime. Finally, in keeping with claims in some of the populism literature that populism arises as a response to economic crisis and/or ineffective responses to the challenges of globalization, it may be that populist attitudes are associated with negative evaluations of the economy. Again, though, this may be because hard times evoke populist responses, or it may be because people with preexisting populist attitudes have a naturally negative assessment of everything the government does.

U.S. surveys Since the analysis of the U.S. surveys is relatively simple, we reverse the order and begin here. For each of our surveys we construct a similar set of OLS models (robust standard errors) using a simple additive index of our four populism questions as our dependent variable. Because of the different response scales used for this module in each survey, we again use the 100-point index created above and then calculate the average value for each respondent across all four questions; thus, higher values indicate stronger populist attitudes. Wherever possible we use similar independent variables across the two surveys, the UCEP and the CCES. Our first indicator is a measure of ideology/extremism, specifically, how the respondent places themselves on a conservative/liberal scale. We group the responses into a set of binary variables: strongly conservative, moderately conservative, moderate, moderately liberal, and strongly liberal; we omit the “moderate” category and use this as our baseline. We expect the strongly ideological respondents to be more populist than the baseline or moderate categories, with conservatives being more populist than liberals. Our second indicator is a measure of education. We group responses into four categories: high school graduate or less, some college, college graduate, and postgraduate. We expect education to be negatively associated with populism. Our next indicator is a measure of household income, which we use to gauge wealth. Both surveys use a non-linear measure, with smaller increments of income at the lower ends of the scale, in order to generate a roughly normal distribution of responses. The UCEP uses 8 categories, while the CCES uses 14. Thus, the results should be seen as gauging the leap across different percentiles rather than as a continuous effect of income. In any case, we expect to find a negative association between income and populism, similar to what we find with education. 16

Because income and education tend to be correlated, one or the other effect may be diminished when we include both of these indicators in the model. Our fourth indicator is a 4-point scale of partisan identity, which we break down into a series of binary variables for each category: Democrat, independent, Republican, and “other.” We use this rather than a sophisticated 7-point scale (indicating strength of partisan attachment) largely because we get the same results and prefer to simplify the presentation of results. While we may find some small association between Republican and populism (because populism tends to be a rightist phenomenon), we think it more likely that the most populist respondents will be those in the “other” category, because they likely support third parties opposed to the traditional system. In the model for the UCEP, we also include a measure for stance on immigration reform. The original question uses a 4-point, ordinal scale to gauge an increasingly friendly stance towards immigrants, but we get similar results and have an easier time presenting them if we reverse the scale and treat it as a continuous variable. Thus, we expect to find a positive association between stance on (anti-)immigration and populism. Finally, we control for age and sex of respondents. Sex is coded so that 1=male, 0=female. Although we ran some initial models with age coded as one continuous variable, we ultimately decided to group responses into four cohorts: pre-Boomer (born before 1946), Boomer (between 1946 and 1964), Generation X (1965 and 1976), and Generation Y (after 1976). We omit the Generation Y cohort as our baseline. We lack strong expectations about the association between either of these controls and populism. The results of these three models are presented in Table 6. Because attitudes towards immigration reform tend to be strongly associated with some of our other indicators, we present a version of the UCEP model without this variable (Model A); the resulting specification is essentially the same as the CCES one (Model C). In looking over these, remember that the scale for the dependent variable is 0-100, and so the coefficients for many of these variables can be interpreted roughly as percentage point shifts. We generally confirm our initial expectations and find similar results across the two surveys, with some important exceptions. First consider ideological extremism. In both the UCEP and CCES (Models A and B), conservatives are more populist than liberals, but extremists are much more populist than moderates. Strong conservatives in particular are 4-7 points more populist than respondents in the middle of the scale, and they are probably more populist than moderate conservatives or moderate liberals (the latter coefficients are in the right direction but not always statistically significant at the p<.10 level or better, so we cannot calculate a meaningful difference). Likewise, strong liberals seem to be more populist than moderate liberals (see the reversal of signs across the two variables), although the difference is not statistically significant at standard levels. (We test this by rerunning the model using moderate liberals as the reference category and find that the coefficient for strong liberal is positive but significant at only the p<.424 level).

Table 6 Regression Results for U.S. Surveys

Model A Model B Model C UCEP 2008 UCEP 2008 CCES 2008 (without immigration) (with immigration) 17

Strongly conservative 4.46** 3.39 6.86* (2.02) (2.05) (2.57) (Moderately) conservative 1.81 1.44 3.20 (1.67) (1.66) (1.63) (Moderately) liberal -1.22 -0.82 -2.68 (2.09) (2.15) (1.76) Strongly liberal 1.68 3.40 -0.53 (2.91) (2.91) (2.26) Some college 0.58 0.27 -0.90 (1.98) (1.99) (1.55) College graduate -4.08** -3.76 -3.05** (2.02) (2.04) (1.44) Post graduate -6.56* -5.89** -7.38* (2.27) (2.30) (2.64) Democrat 0.91 1.15 0.14 (1.95) (1.96) (1.58) Independent -1.82 -1.14 2.06 (1.92) (1.94) (1.70) Third party 9.12* 8.41* -10.42* (2.51) (2.57) (3.58) Pre-boomer (before 1946) 11.97* 11.25* 2.68 (2.37) (2.45) (2.04) Boomer (1946-1964) 2.56 2.71 2.88 (1.53) (1.52) (1.90) Generation X (1965- 1976) 4.49* 4.51* 2.32 (1.43) (1.43) (2.19) Male -0.87 -1.56 -2.27 (1.15) (1.17) (1.23) Position on immigration - 3.04* - - (0.69) - Income - - -0.71* - - (0.20) Constant 65.11* 57.23* 73.61* (2.41) (3.24) (2.72) Observations 879 853 874 R-squared 0.08 0.10 0.08 Adjusted R-squared 0.06 0.08 0.06 Robust standard errors in parentheses ** significant at 5%; * significant at 1%

Looking at education, we find an even stronger, consistent result across both surveys (Models A and C). In both, higher levels of education are associated with lower populism. The effect is strongest among postgraduates—6-7 points—although it is also statistically significant among college graduates. In the case of income, we find a clearly dissimilar effect that partly confirms our expectations. In the CCES the effect is quite statistically significant and in the direction we 18 expected—on average, a 0.7-point reduction in populism for each percentile shift in income. In the UCEP, the effect is not statistically significant at standard levels, although the coefficient is in the right direction. We do not yet know why these outcomes differ so much; the average incomes across the two samples are very similar, and their scales really only differ in terms of the number of intervals. The effect of partisanship diverges even more sharply across the two samples. In all models, of course, we find that traditional partisan identities matter very little: The statistically insignificant results suggest that Democrats, Republicans, and independents are essentially identical in their level of populism.4 This itself is a significant finding and is largely what we expected. However, third-party supporters are either much more populist (in the UCEP) or much less populist (in the CCES); in the UCEP, the difference is 8-9 points, while in the CCES it is over 10 points. Although the number of respondents in each of these subsets is rather small in both surveys, the effect is still quite strong. On closer examination, this result makes sense. In Utah, third party voters are usually from the radical right and have a strong patriotic/nationalist emphasis (Constitutional Party, American Party, etc.); these are parties we would expect to have a populist discourse. In the national sample, in contrast, respondents indicating “other” are often Libertarians, a party responding to a classic liberal ideology that seems less compatible with populist assumptions about the dualistic, Manichaean nature of politics or the need to respect a common will. Thus, not all outsider parties are populist, but people with populist attitudes seem much more likely to support outsider parties. Finally, one of the strongest associations we find—and only in the UCEP model (Model B), because the question was not included in the CCES—is between populism and position on immigration reform. The coefficient, which is statistically significant at the p<.000 level, indicates that a one-standard-deviation shift in views on immigration is associated with a 3 point shift in populist attitudes. However, the scale on immigration reform runs from 1-4, and the most extreme response is actually the second most-common one; this response is associated with a shift of about 12 points. This confirms the common perception of populism as a xenophobic phenomenon associated with the radical right in the industrial democracies. Note that including the immigration questions tends to depress the coefficients on all other variables. As for our controls, sex of the respondent has no discernible relationship to populism in either model. Age does have an association, but only in the UCEP model, where populist attitudes seem to strengthen with age. The effect is nonlinear: those in the oldest cohort (pre- Boomer) are far more likely to be populist than those in the next cohort (Baby Boomers) or the others, a difference of an astonishing 11 points with the youngest cohort. We lack an explanation for this effect or why it is stronger in the Utah sample.

AmericasBarometer Because the AmericasBarometer includes data for over 20 countries, we have an opportunity to consider country-level effects as well as their interaction with individual-level attributes. We do this by using a hierarchical or multi-level model. Hierarchical models are especially useful when statistically examining cross-national data such as the AmericasBarometer, because they include mechanisms that allow individual-level variables to be compared against country-level variables. In this way we can consider respondent demographics, attitudes, and contexts jointly. As a statistical bonus, hierarchical models use a complex disturbance term that helps us understand

4 Remember that these partisan categories are dummy variables, with Republican as the reference category; coefficients show the difference between these categories and a typical Republican respondent. 19 how much variance is explained by each level of the model (Singer and Willett 2003), giving us added insight into its predictive value. Our model isolates two levels. First, it considers several sets of individual-level or intracountry factors derived from questions on the AmericasBarometer. The bulk of these are measures of individual sociodemographic and partisan attributes like those in the U.S. surveys, namely, education, gender, age, ideology, and wealth. We recoded level of education into a series of four binary variables for no education (omitted from the model as the baseline), primary, secondary, and higher education. Gender is also a binary variable (1=women, 0=men). Age is a continuous variable that ranges from 18 to 105 years. Ideology is coded at the individual level as a series of dummies indicating position on a ten-point leftist to rightist scale. Respondents were grouped by dividing the scale into five categories: strong left (1 or 2 on the scale), moderate left (3 or 4), moderate (5 or 6), moderate right (7 or 8), and strong right (9 or 10); the moderate category is omitted as the baseline. Wealth is an additive index ranging from 1 to 12 gauging the number of material possessions of the respondent (as established by Córdova 2008). Though this measure provides a large amount of variance in Latin American countries, its utility breaks down in the United States, where almost every respondent would score a twelve. Thus, this series was not included in the U.S. questionnaire, and respondents from this country are dropped from the analysis whenever the wealth variable is included. Because the AmericasBarometer is longer and richer than the U.S. surveys, we are also able to gauge individual attitudes that are analogous to the country-level factors often associated with the emergence of successful populist movements. These consist of perceptions of corruption, experienced corruption, crime victimization, sociotropic evaluation of the economy, and pocketbook evaluation of the economy. Reported perception of corruption and both the sociotropic and pocketbook economic evaluations are measured on ordinal scales, ranging from 1 to 5. Higher values indicate greater perceived corruption, and positive economic evaluations. Experienced corruption and crime victimization are binary variables, a 1 indicating that the respondent has been victimized in one or more crimes or has ever been solicited for a bribe from a government official, at work, or in receiving public services. Finally, we include the rightist ideological index a second time, now coded as a single continuous variable using the entire ten-point scale, rather than as a series of categorical, binary variables. We include this measure a second time for the purposes of testing the interaction between the ideological stripe of populism and the level of economic development. By including both versions of the same indicator, we can explore the intercountry effects of development on ideology, but this arrangement diminishes the extreme right ideology variable's coefficient, as we will highlight later. The second level of the model captures intercountry factors that vary between national contexts and sometimes interact with individual-level variables of interest. Here there are only two variables: level of development and average experience with democracy. Level of development is operationalized as the 2007 logged per capita GDP, and ranges from ___ to ___ [get numbers]; data come from the World Bank’s World Development Indicators database. Experience with democracy is the average score since independence of the democracy variable in the Polity IV index. We have to include the level of development in order to gauge its interaction with ideology, but we also include both of these country-level variables separately so that readers can see the (negligible) country-level effects when we analyze populist attitudes. The various levels of the model can be expressed arithmetically for ease of interpretation. Level one abstracts as follows: 20

Populismi = π0i + π1iX1i + β1X2 + ... βxXy+ εi The first term, π0i, is essentially the slope term of the model, but rather than being a constant as in OLS regression, it is a function of country-level attributes that we will shortly model. The second set of terms, π1i X1i, captures the interaction between ideology and level of economic development. The first of these terms, π1i, is a country-level function of economic development that we will shortly specify; and X1i is a term representing the rightist ideological scale as it varies across the pooled individuals in the sample. β1i + ... βx are regression coefficients representing the populism trajectory of each country i. These coefficients are interpreted with respect to X2 ... Xy, which represent the individual-level variables (here, the demographic measures and the governmental attitude and experience indicators) that vary across individuals without regard for country. And εi is an error term. In the second level of the model, π0 and π1 are further specified, allowing us to investigate interactive effects of country-level contextual factors (level of development and experience with democracy) on both the overall levels of populism and on the individual-level variable of interest (the previously mentioned rightist ideological scale). We illustrate this with the following set of equations:

π0i = α00 + α10Xi + ζ0i π1i = α01 + α11Xi + ζ1i where α00 is the average constant across all the countries, α10 is the average effect of country- level predictors Xi on π0i, and ζ0i is an error term. The equation for π1i is similar: α01 is the average slope with respect to X1i from the equation in level one, and is adjusted by the effect α11 resulting from some country-level predictor Xi (here, the level of economic development) with an error term ζ1i. The resulting model is reported in Table 7 in four specifications. Model A is an unconditional means model, which provides us a baseline of model statistics that allow us to estimate how much variance is being explained by the distinct levels of our model. Extracting the variance components of this model, we calculate a rho-statistic of .10, which indicates that roughly 10 percent of the variance in the populism index can be explained by country-level contextual factors. This reaffirms what we already know, that most of the variance is among individuals and not due to any distinct national characteristic. Model B extends the unconditional means model by exploring the rightist ideology variable with respect to development, while including the separate ideological categories and controlling for sociodemographic measures. Explaining these results is difficult, because we have to take into account multiple coefficients at once in order to properly gauge the interaction. [In the future you will see a graph that talks about average expected effects.] In a nutshell, we find that the effect of rightist ideology on populism does vary with the context. When development is low, the average effect of ideology on populism drops below zero, but as development increases, the effect of rightist ideology reverses and becomes positive. This supports our development hypothesis: countries with high levels of development tend to have right-populist attitudes, while countries low levels of development tend to have left-populist attitudes. Additionally, this model clarified the modest tendency we found in the U.S. survey results for populism to be associated with ideological extremism. When the ideological 21 categories are included, extreme left ideology predicts higher levels of populism; by itself, this result does not particularly support our hypothesis. But as mentioned before, the continuous ideological measure absorbs some of the effect of the extreme right ideology variable. When it is omitted (not reported here), the extreme right ideology binary measure shows an effect nearly identical to that of extreme left ideology, both substantively and statistically. Thus, in a larger sample of countries where both right and left populisms are prevalent, we find that populism is still consistently radical. Model C adds almost all of the remaining individual-level and country-level variables, leaving off wealth. Three factors dampen the levels of populist attitudes, besides the ideological and developmental factors already noted. As we saw in the U.S. surveys, education has a negative effect; higher education especially is associated with low levels of populist attitudes. Likewise, favorable economic evaluations of the country, both sociotropic and pocketbook, predict lower populism with a maximum effect equivalent to that of higher education. And higher levels of development are also associated with reduced levels of populism. The first two of these findings confirms what we predicted, although the results alone tell us little about the direction of causality; the third of these results would seem to go against our assertion that country-level factors generally fail to predict populist attitudes, but we will see in a moment that this result is spurious. Three other factors predict higher populism in a country. As in Model B, ideological extremism has a positive effect. A high perception of corruption in government also predicts higher levels of populism. The remaining variables, experienced corruption, crime victimization, and experience with democracy, have no statistical effect. The fact that populist attitudes go with perceived corruption, but not experienced corruption or crime, is especially noteworthy. It suggests that at this individual level the association of corruption with populism is not really because corruption causes populism, but because people with populist attitudes tend to see corruption around them. The final specification, Model D, provides a convincing test of robustness for our model and clarifies the relationship of our country-level variables to populist attitudes. We add the wealth index to the model, which excludes the observations from the United States from the calculations. The findings of the previous models hold steady, though the effect of higher education is lost due to a high correlation between education and material possessions. Note that even without the United States, the joint effect of development and ideology on populism is robust across the relatively homogeneous Latin American countries; this is an impressive confirmation of our prediction about the ideological direction of populism. Yet we also find the direct effect of development on populism washes out once we control for wealth at the individual level, suggesting that our previously positive finding for the level of development was spurious. Thus, in the end none of our country-level variables have any direct effect on populist attitudes.

Discussion and Conclusion In this paper we engage in a novel attempt to measure populist attitudes and their correlates at the individual level. Our measures, which capture an ideational or discursive definition of populism, hold up well in a number of ways. Even when using different wording in different contexts and surveys, responses to our questions tend to cluster together on a separate dimension from other attitudes and issue positions. Thus, elements of populism such as a Manichaean cosmology, reified notion of a popular will, and belief in a conspiring elite typically go together. This confirms work elsewhere showing that elements of populist discourse cohere very tightly at 22 the level of party elites (Jagers and Walgrave 2007). However, our finding that these ideas are generally separate from support for strong executive leadership leads us to question recent efforts to define populism in political-institutional terms. We suspect that if populist movements have charismatic leaders, they do so for other functional reasons. Our analysis of the correlates of populist attitudes generally supports a view of populism as a kind of latent worldview or disposition that is activated under circumstances of moral threat. We find that populist attitudes are very common in the population of dissimilar countries, and that variation in populist attitudes is much greater within than across these countries. Populist attitudes correlate in predictable ways with basic sociodemographic and partisan attributes, especially low education and poverty, ideological extremism, support for third parties, and attitudes towards government performance. The specific ideological stripe of populist attitudes in any country seems to be the result of an interaction of populist attitudes with sociological context. However, country-level differences, such as level of economic development or cumulative democratic experience, have little direct effect on the incidence of populist attitudes in the population. These findings strongly suggest that we have tapped into basic features of personality rather than a set of conscious ideas that are the product of socialization. If populism is somehow rooted in a basic set of attitudes or a widespread disposition, then this has implications for how populist movements are mobilized. We cannot assume that attitudes translate seamlessly into action, nor, ironically, can we treat actual populist movements as the mere coincidence of willing followers and capable leaders. A large pool of potential followers always exists. This means that different frequencies of populism across countries are primarily the result of macro-level factors and possibly the availability of charismatic leadership to provide the movement with focus and direction. If Latin American countries have more populism, it is the result of their different capacities for democratic governance, not some difference in their political culture. Yet populism is only an empty box into which many ideologies fit. Individual participation in any movement is also contingent on the fit between the ideology of the movement and that of the potential activists; some people with populist attitudes are mobilized into the movement, while others with different ideologies may mobilize counter-movements. This magnifies the polarization that naturally results from application of a Manichaean interpretation of the political world. However, because the elite often becomes the object of populist outrage , and because populist attitudes are less common among the wealthy and well educated, the opposition is less likely to embody a fully populist response. Testing this theory requires a research design different than ordinary surveys with closed- ended questions like the ones we have here. We need to be able to distinguish between populist attitudes and actual populist practice, and if we find a difference, then we need to see if we can actually arouse the latter. We are currently exploring experimental methods to see if this is possible. 23

Table 7 Regression Results for AmericasBarometer

Fixed effects Model

A B C D

On levels of populism Intercept 4.46*** 3.33*** 3.13*** 3.92***

(0.09) (0.32) (0.31) (0.30)

Higher education -0.13*** -0.16*** -0.09

(0.05) (0.05) (0.06)

Secondary education 0.08 0.04 0.06

(0.05) (0.05) (0.05)

Primary education 0.06 0.01 0.02

(0.05) (0.05) (0.05)

Woman 0.01 0.02 -0.01

(0.01) (0.02) (0.02)

Age 0.00 0.00 0.00

(0.00) (0.00) (0.00)

Extreme left ideology 0.30*** 0.34*** 0.41***

(0.07) (0.07) (0.07)

Moderate left ideology 0.07 0.09 0.15***

(0.04) (0.04) (0.04)

Moderate right ideology -0.05 -0.06 -0.04

(0.04) (0.04) (0.04)

Extreme right ideology 0.06 0.04 0.06

(0.07) (0.07) (0.08)

Support for the Executive 0.16*** 0.13***

(0.01) (0.01)

Perception of corruption 0.08*** 0.07***

(0.00) (0.00) 24

Experienced corruption 0.01 0.01

(0.02) (0.02)

Crime victimization 0.00 0.00

(0.00) (0.00)

Sociotropic economic eval. -0.03*** -0.03***

(0.01) (0.01)

Pocketbook economic eval. -0.07*** -0.05***

(0.00) (0.00)

Experience with democracy -0.01 -0.01

(0.02) (0.02)

Level of development -2.03*** -1.75* -0.26

(0.75) (0.72) (0.67)

On effects of ideology Intercept 0.14*** 0.14*** 0.08***

(0.03) (0.03) (0.03)

Level of development 0.27*** 0.26*** 0.11*

(0.07) (0.06) (0.05)

Variance components

Level 1 Within-country 1.46*** 1.40*** 1.41*** 1.37***

(0.01) (0.01) (0.01) (0.01)

Level 2 In overall levels 0.16* 0.30* 0.24* 0.16*

(0.05) (0.09) (0.08) (0.05)

In effects of ideology 0.00*** 0.00*** 0.00***

(0.00) (0.00) (0.00)

Covariance -0.02*** -0.02*** -0.01

(0.01) (0.01) (0.00) 25

Summary statistics

n 31380 26115 23060 21531

Log likelihood 50486.0 41537.2 36719.6 34000.3

AIC 100978 83108.3 73487.2 68050.6

BIC 101003 83247.2 73680.3 68250 26

References

Barr, Robert. 2009. Populists, outsiders and anti-establishment politics. Party Politics. 15 (1):

29-48.

Betz, Hans-Georg. 1994. Radical Right-Wing Populism in Western Europe. Basingstoke:

Macmillan.

Canovan, Margaret. 1999. Trust the people! Populism and the two faces of democracy. Political

Studies. 47 (1): 2-16.

Cardoso, Fernando Enrique, and and Enzo Faletto. 1979. Dependency and development in Latin

America. Berkeley: University of California Press.

Córdova, Abby. 2008. "Methodological note: Measuring relative wealth using household asset

indicators." LAPOP Insight Series, edited by M. Seligson: Vanderbilt University. de la Torre, Carlos. 2000. Populist Seduction in Latin America: The Ecuadorian Experience.

Athens, Ohio: Ohio University Center for International Studies. de la Torre, Carlos. 2007. The resurgence of radical populism in Latin America. Constellations.

14 (3): 384-97.

Deegan-Krause, Kevin, and Tim Haughton. 2009. Toward A More Useful Conceptualization of

Populism: Types and Degrees of Populist Appeals in the Case of Slovakia. Politics &

Policy. 37 (4): 821-41.

Di Tella, Torcuato S. 1965. Populism and reform in Latin America. Chapter in Obstacles to

Change in Latin America, ed. Claudio Veliz, pp. 47-74. London: Oxford University

Press.

Hawkins, Kirk A. 2009. Is Chávez populist? Measuring populist discourse in comparative

perspective. Comparative Political Studies. 42 (8): 1040-67. 27

Hawkins, Kirk A. 2010. Venezuela’s Chavismo and Populism in Comparative Perspective.

Cambridge: Cambridge University Press.

Hibbing John R., and Elizabeth Theiss-Morse. 2002. Stealth Democracy: Americans’ Beliefs

about How Government Should Work. New York: Cambridge University Press.

Ionescu, Ghita and Ernest Gellner, eds. 1969. Populism: Its Meaning and Characteristics.

London: The Macmillan Company.

Ivarsflaten, Elisabeth. 2008. What Unites Right-Wing Populists in Western Europe? Re-

Examining Grievance Mobilization Models in Seven Successful Cases. Comparative

Political Studies. 41 (3): 3-23.

Jagers, Jan and Stefaan Walgrave. 2007. Populism as political communication style: An

empirical study of political parties’ discourse in Belgium. European Journal of Political

Research. 46: 319-45.

Laclau, Ernesto. 2005. On Populist Reason. London: Verso.

Mudde, Cas. 2004. The populist Zeitgeist. Government and Opposition. 39 (Fall): 541-63.

Panizza, Francisco, ed. 2005. Populism and the Mirror of Democracy. London: Verso.

Roberts, Kenneth M. 1995. Neoliberalism and the transformation of populism in Latin America:

The Peruvian case. World Politics. 48 (1): 82-116.

Roberts, Kenneth M. 2003. Social correlates of party system demise and populist resurgence in

Venezuela. Latin American Politics and Society. 45 (3): 35-57.

Singer, Judith D. and John B Willett. 2003. Applied Longitudinal Data Analysis: Modeling

Change and Event Occurrence. Oxford: Oxford University Press.

Stenner, Karen. 2005. The Authoritarian Dynamic. Cambridge: Cambridge University Press. 28

Weyland, Kurt. 2001. Clarifying a contested concept: Populism in the study of Latin American

politics. Comparative Politics. 34 (1): 1-22.

Weyland, Kurt. 2003. Economic voting reconsidered: Crisis and charisma in the election of

Hugo Chávez. Comparative Political Studies. 36 (7): 822-48.

Recommended publications