The Ideational Approach to Populism

Although the study of populism has traditionally been the domain of Latin Americanists, research here has become increasingly comparative. One of the most important payoffs of this comparative work is conceptual. Rather than defining populism in structuralist, economic, or political-strategic terms, a growing number of scholars around the world are using an ideational conceptualization that draws heavily from earlier discursive theories. By employing the ideational approach, scholars have been able to provide empirical measures of populist discourse. In this article we explain and show the advantages of this ideational approach to a Latin American audience by presenting a new historical dataset measuring the discourse of Argentine, Chilean, and Peruvian presidents across the twentieth century. Our main intent is to clarify the ideational approach as well as to enliven the conceptual debate. While we are critical of alternative definitions, we acknowledge and reassess their theoretical insights.

Although the study of populism has traditionally been the domain of Latin Americanists, research here has become increasingly comparative. One of the most important payoffs of this comparative work is conceptual. Rather than defining populism in structuralist, economic, or political-strategic terms, a growing number of scholars around the world are using an ideational conceptualization that draws heavily from earlier discursive theories. By employing the ideational approach, scholars have been able to provide empirical measures of populist discourse. In this article we explain and show the advantages of this ideational approach to a Latin American audience by presenting a new historical dataset measuring the discourse of Argentine, Chilean, and Peruvian presidents across the twentieth century. Our main intent is to clarify the ideational approach as well as to enliven the conceptual debate. While we are critical of alternative definitions, we acknowledge and reassess their theoretical insights.
For decades the study of populism was the domain of Latin Americanists. Aside from the frequently cited volume edited by Ionescu and Gellner (1969), most of the conceptual, theoretical, and empirical work on the phenomenon came from scholars studying Latin America, indeed from Latin Americans. 1 However, the center of gravity has begun to shift. The study of populism has grown dramatically in Europe, where the rise of populist radical right parties-and more recently also populist radical leftist parties-have captured the attention of policy makers and scholars (Mudde 2007;Stavrakakis and Katsambekis 2014). And while a small undercurrent of historical scholarship in the United States has always focused on populism (Hofstadter 1965;Kazin 1998), the emergence of the Tea Party and Occupy movements in the 2000s, followed by the tumultuous 2016 president campaign, have brought a resurgent interest in the topic (Oliver and Rahn 2016;Skocpol and Williamson 2013). Now, scholars are seeking to understand populist forces in countries ranging from Thailand to India and from Zambia to Australia (Kuhonta 2008;Moffitt and Tormey 2013;Resnick 2014;Subramanian 2007). Thus, work on populism is becoming increasingly comparative.
This new comparative research has yielded a number of scholarly payoffs, one of the most important of which is conceptual. Over the years, Latin Americanists defined populism in a number of ways: structuralist, economic, political-strategic, and discursive. The disagreement among scholars yielded rich case studies but little in the way of consistent, replicable measurement or causal analysis. Skeptics despaired, suggesting that populism was not simply a "contested concept," but one that should be abandoned (e.g., Roxborough 1984).
However, out of the new comparative research is emerging an argument that populism should be defined in ideational terms very similar to the discursive definition used among some Latin Americanists. This definition reduces populism to a common, minimal core, seeing it as a political discourse that posits a cosmic struggle between a reified "will of the people" and a conspiring elite. Populism in this ideational sense exists as part of a larger typology of discursive frameworks, including pluralism (in which people reject a Manichaean outlook and see the good in their political opponents) and elitism (which inverts populism's outlook by celebrating the virtues of the elite and the fallibility of the masses). According to this ideational conceptualization, the organizational features of populist movements and their structural preconditions are contingent aspects, which sometimes but not always accompany populism. Of course, the ideational approach to populism stays in close contact with the work of the Argentine political theorist Ernesto Laclau, who was a pioneer in the study of Latin American populism (Laclau 1977) and later elaborated a sophisticated political theory on populism (Laclau 2005), but it moves beyond his work in important ways.
The ideational approach has been highly productive. On the one hand, by assuming that populism is a set of ideas that can be combined with other ideological features, new research has been able to identify different subtypes of populist forces across the world (de la Torre 2013; Mudde 2007; Mudde and Rovira Kaltwasser 2013;Stavrakakis and Katsambekis 2014). On the other hand, a number of scholars have applied techniques of textual analysis and survey research to the measurement of populist ideas, not just in one or two countries (Armony and Armony 2005; Jagers and Walgrave 2007) but in regional and cross-regional studies (Akkerman, Mudde, and Zaslove 2014;Hawkins 2009;Rooduijn and Pauwels 2011). These works are beginning to examine the causes and consequences of populism, including its impact on policy and democratic institutions (Hawkins 2010;Mudde and Rovira Kaltwasser 2012;Mudde and Rovira Kaltwasser 2013;Mudde and Rovira Kaltwasser 2017).
In this article we seek to speak to a Latin Americanist audience by showing the ideational definition in action within the region. We first do so empirically, by showing the results of a new historical dataset that measures populist discourse of presidents across the twentieth century in three countries: Argentina, Chile, and Peru. This by itself is a remarkable achievement. Studies of populism using any definition, let alone an ideational one, have not engaged in any kind of quantitative historical comparison (qualitative studies abound). Our analysis points out some of the challenges and pitfalls of constructing this kind of data.
Second, we try to leverage this data to say something about alternative definitions and their merits. In this regard, we feel that Latin America still has something to offer to the comparative study of populism; the definitions developed in this region spring from an appreciation of actual movements and trends that are unusually strong. While we are critical of most of these definitions, we do not feel that they should be forgotten but reassessed for their theoretical insights. In the latter half of the article we provide this critique, and in the conclusion we revisit these definitions and show how they point us to important directions for future research.

The Ideational Approach
In referring to an ideational approach to populism, we suggest not only that it has a particular conceptualization but that this conceptualization comes with some rough theoretical arguments about the importance of ideas for causal analysis. Among those who adhere to the ideational approach, there is some debate about what exactly populism is a type of: a discourse, a thin-centered ideology, or something else. Given that we see these differences as minor, we use the terms "discourse," "ideology," "outlook," and "worldview" somewhat interchangeably. Populism is one instance of a particular level or kind of ideas. While this type is more specific and articulated than a mere set of personality traits, it is not quite as conscious and programmatic as an ideology. Rather, it always attaches itself to some "host" ideology, as can be seen in comparing right-wing, "exclusionary" forms of populism in Europe with left-wing, "inclusionary" populism in much of Latin America (Mudde and Rovira Kaltwasser 2013).
What is populism's unique content? We explain this in terms of Dryzek and Berejikian's (1993) grammar of political discourse. To begin with, populism has a cosmology, or a general belief about how the political universe operates. Populism sees politics as a struggle between the forces of good and the forces of a knowing, diabolical evil-hence, it is Manichaean or dualistic. At the extreme it appears as a kind of political paranoia (Hofstadter 1965). Second, it has an ontology, or a belief about the naturally existing set of political actors, the amount of agency associated with each of these entities (essentially, whether they are objects or subjects), and these actors' motives. Specifically, populism talks about the reified will of the ordinary folk who constitute the bulk of the citizenry; these are seen as the embodiment of democratic virtue. It juxtaposes this notion of the people with an equally reified group of elites, whom it sees as secretly aiming to subvert the popular will for selfish purposes. Because populism seeks the immediate political expression of the popular will and sees its opposition in such diabolical terms, it tends to adopt an "anything goes" attitude toward democratic institutions and minority rights.
Scholars who adopt an ideational definition have identified at least two opposites: elitism and pluralism (Hawkins 2009;Mudde 2004;Plattner 2010). Elitism shares populism's Manichaean distinction between the people and the elite but adopts an opposite view regarding the morality of each of these entities. It assumes that the people should be conceived as a dangerous mob while depicting the elite as a reduced group of actors who, due to their intellectual and moral superiority, should be in charge of the government. In contrast, pluralism rejects the Manichaean distinction between the people and the elite. It adopts a more universalistic view of human nature, one that takes for granted the diversity of ideas and interests in society and treats evils as impersonal challenges. Thus, those who adhere to pluralism are normally inclined to think of popular sovereignty as a dynamic and open-ended process rather than a fixed and unified will of the people (Ochoa Espejo 2011).
Although a growing number of Latin Americanists adhere to an ideational approach (Barros 2006;de la Torre and Arnson 2013;Hawkins 2010;Panizza 2005;Rovira Kaltwasser 2014), the traditional tendency has been to use one of three other definitions. The first is an economic one that sees populism as a set of shortsighted macroeconomic policies adopted for electoral purposes and that end up generating more harm than good (Dornbusch and Edwards 1991;Edwards 2010). This definition sees these policies as interconnected parts of a larger approach to economic policy making that results in a cycle of demand overstimulation, inflation, and structural adjustment. It is worth noting that this economic definition is common among journalists and policy makers who use "populist" as a pejorative.
The second approach is a structuralist one that sees populism as a type of regime tied to the backward development of Latin American countries, one in which a cross-class, charismatic movement motivated by an "anti-status quo ideology" pursues import-substituting industrialization (Di Tella 1965;Germani 1978). This definition focuses on not only the politics behind economic policies but also the structural forces behind class formation. For instance, Germani (1978) argued that the rapid rural-urban migration experienced by some Latin American countries facilitated the rise of large swaths of population who felt orphaned and thus could be easily mobilized by a populist leader. While the structuralist approach was strongly associated with dependency theorists (cf. Cardoso and Faletto 1979), it continues among scholars more inclined toward political sociology (Oxhorn 1998;Vilas 1992).
Finally, a newer political-strategic approach associates populism with mass movements led by charismatic outsiders using an anti-elite discourse and governing in a top-down fashion (Barr 2009;Weyland 2001). This definition is similar to the older structuralist one in its emphasis on the political, but encompasses movements without cross-class appeals or the economic policies of import-substituting industrialization. It typically sees the adoption of populist discourse as intentional and insincere and hence insufficient as a tool for understanding populist parties and movements.
Although there are apparent ontological differences among these three approaches-economic definitions are presumably about economic policy, structuralist definitions are about political regimes, and the newer political-strategic approaches are about political strategies-it is striking that all of them incorporate some notion of an underlying set of ideas. Thus, populism is "an approach to economics" (Dornbusch and Edwards 1991, 9), is motivated by an "anti-status quo ideology" (Di Tella 1965, 53), or involves an "anti-establishment discourse" (Barr 2009, 31). Where the ideational definition differs is in its emphasis. It sees populist ideas as the main driving force behind other material features of populist movements. Many of the features highlighted by other definitions are in fact a product of these ideas, contingent on socioeconomic and sociopolitical context. For example, populist ideas can push organizations in highly participatory directions, but the precise form of movement organization depends on the available material and human capital. As political skills, communications technology, and social networks diffuse among the population, populist movements may acquire a grassroots character that no longer depends as much on top-down leadership.
The ideational definition is closely related to the pioneering discursive approach of Laclau (1977Laclau ( , 2005. Both approaches locate populism in the realm of ideas and highlight the central place of a so-called popular identity as well as its antagonistic relationship with a putative, vilified elite that stands as the anti-people. However, there are some important differences. Perhaps the most important is that Laclau's work has a normative impetus to transform politics and break with the liberal status quo. By contrast, the ideational approach is more inclined to facilitate the generation of empirical knowledge and avoids making normative judgments, to the point that the impact of populism on democracy is seen as an empirical question (Mudde and Rovira Kaltwasser 2012). Second, and in close relationship with the previous point, the ideational approach sees the possibility of many types of redemptive discourse. Laclau tends to see populism as the only democratic discourse capable of unifying and inspiring large majorities around a transformative project, a "logic of equivalence," set against a "logic of difference" that lowers the sights of citizens to their individual material concerns and eliminates the most politically meaningful questions. Nevertheless, the reasons for limiting the discursive terrain this way are not clear; empirical research reveals a wide variety of normatively charged discourses.
Third, because Laclau's notion of "discourse" blends the substance of populist ideas with how they actually play out in the political sphere (discourse implies a conversation or exchange), Laclau tends to limit populism to movements that attract a numerical majority, such as left-populist movements in Latin America. In contrast, the ideational approach analytically separates the existence of populist language from its effect on politics. This allows it to test propositions about the conditions under which populist rhetoric succeeds in its political goals, and means that a greater variety of movements and parties can be included under the populist umbrella, including minoritarian radical-right ones that may lack charismatic leadership. 2 An important insight provided by Laclau concerns the relationship of populism to Marxism. Traditionally, leftist scholars (especially those drawing on structuralist approaches) were opposed to the idea that populism could be combined with socialism. The type of deep revolutionary transformation required by Marxism-Leninism was inimical to the supposedly halfhearted reforms untaken by classical Latin American populists, which retained capitalism while engaging in import-substituting industrialization (Cardoso and Faletto 1979;Weffort 1978). However, in his earliest work, Laclau suggested that many instances of socialist revolutionary experiences were populist, using the same antagonistic discourse that distinguished traditional populist movements and facilitated their electoral success (Laclau 1977). The connection was not a necessary one. In fact, the Leninist position that communist parties should present themselves as the vanguard of the working class made it likely that parties would adopt an elitist stance limiting their capacity to mobilize those who were being exploited by the capitalist system. Not by chance, in his work with Chantal Mouffe (Laclau and Mouffe 1985), Laclau argued that the way ahead for the Marxist left consists in generating a discourse that is capable of gathering a set of different and divergent demands under the same political umbrella. And in his book On Populist Reason (Laclau 2005), he comes full circle and argues that populism is the ideal discursive practice for achieving this logic of equivalence.

Expectations and Method
In order to demonstrate the utility of the ideational definition, we present a new, historical dataset of elite-level populist discourse in Latin America. This dataset is derived from a textual analysis of presidential speeches from 1900 to 2000 in three countries: Argentina, Chile, and Peru. Our decision to focus on presidential discourse is largely pragmatic: most populism studies focus first on the politicians who lead these parties and movements, and presidents are the most relevant elite-level actors. Unsurprisingly, presidential speeches are also the most readily available texts for a historical analysis. The choice of countries is driven by three similar considerations. First, collecting political speeches for Latin American countries across time is a very difficult task, but we have been able to do this with great success for the countries in question because they have better historical records. Nevertheless, this was an enormous effort that took over a year for each country. We worked with a small team of assistants who first combed through research libraries using the Online Catalog of the Library of Congress, then traveled to the capital of each country and continued searching for documents in national archives, university libraries, and (where they existed) presidential foundations. Assistants took digital photographs of each speech, then created a comprehensive catalog.
Second, given that the historical and contemporary literature on these three countries highlights several populist leaders who have gained power (e.g., Juan Domingo Perón for Argentina, Carlos Ibáñez del Campo for Chile, and Alberto Fujimori for Peru), our data should allow us to validate the claims of the ideational approach empirically. After all, one of the main goals of this contribution lies in providing empirical measures of populist discourse and analyzing these empirical results in light of the rich scholarly debate on Latin American populism by reassessing both the validity and value of alternative definitions. Third, by comparing Argentina, Chile, and Peru we have the opportunity of examining three countries that have experienced quite different trajectories, particularly in terms of the formation and institutionalization of political parties across the twentieth century. This is an important feature for our analysis because various scholars have argued that a strong party system seriously affects the likelihood of populism (e.g., Kitschelt et al. 2010;Levitsky and Loxton 2013;Mainwaring and Scully 1995).
What sort of claims can we test? Until now, empirical studies using the ideational definition have focused on contemporary parties and movements, especially in Europe (Jagers and Walgrave 2007;Mudde 2007; Rooduijn and Pauwels 2011) but somewhat in Latin America (Armony and Armony 2005;Hawkins 2009). These studies are valuable because they show the presence of populist discourse across parties and movements in a variety of countries, thus validating the claim that populist ideas are a key unifying feature. By creating valid and reliable measures, they also demonstrate the scientific utility of the ideational approach. The empirical study of populism was traditionally guided by impressionistic categorizations of parties and movements that were not replicable; this has especially been true for studies using the earlier discursive approach.
Of course, by providing data on which actors can be considered as populists, we do not answer questions about the causes and consequences of populism. Our point is that, to address these questions, we need to start by identifying the cases that one should consider for undertaking more rigorous empirical studies on the causes and consequences of populism. Without these types of data there is the real possibility that scholars and pundits pick the cases that they prefer, but without empirically demonstrating that these cases should indeed be seen as examples of populism. Therefore, we believe that measuring the supply of populism (i.e., identifying agents who speak the populist language) is an important step also for those who are interested in studying populism empirically.
At the same time, a historical dataset offers additional opportunities to test the differences between the ideational approach and other competing ones. First, if populist discourse is the defining characteristic of populist parties and movements, we should see it not just across contemporary parties and movements but also among historical parties and periods traditionally identified as populist. This is especially important for a Latin Americanist audience, which is well aware of older cases that have helped define the study of populism. Second, although populist discourse should be prevalent at key moments, it should not be a constant feature of elite-level political discourse; otherwise, it is not a distinguishing attribute of populist parties and movements. Third, we should find this discourse among economically right-wing parties and movements. The electoral strength and frequency of these movements should be much lower in Latin America than in economically developed Western democracies with larger propertied classes, but the recent era of neoliberal reform provides an important window for their emergence. Finally, populism may coexist with doctrinaire Marxist ideology from the past, especially in the electorally successful movements highlighted by our data.
There are practical challenges to measuring elite-level populist ideas across time through textual analysis. To begin with, in most countries, the universe of political texts-even the universe of texts for any one chief executive-has not been collected. Comprehensive archives such as those for US presidents are the exception. Moreover, even where resources are now available for creating these archives, there is less published material the further back in time one goes. In some countries, except for a few well-known politicians, the available record consists solely of addresses within the legislature such as an annual report by the chief executive; speeches outside the halls of government have not been preserved.
These challenges were present in our own study. Our first task was to create a sample of speeches for each president. Because we could not count on having the same types of speeches for each term, we opted for random sampling rather than quota sampling and chose four speeches per year using our catalog and a random number generator. This approach tended to oversample ribbon-cutting and international speeches (the most common types but also the least likely to display populism) resulting in lower averages than in the study of contemporary presidents we have conducted elsewhere (Hawkins 2009). The main challenge with this step of the analysis was the lack of available speeches for some presidents. The first two or three decades for all of these countries had spotty coverage (sometimes with only one speech per year, the annual report to the legislature), and this gap reemerged for a few later presidents who were unlucky enough to be forgotten by historians. Because no other speeches were likely to be forthcoming, we felt it was better to press forward with the analysis while acknowledging the potential weaknesses of the data. 3 Our second and final task was to code the speeches. For this, we used a form of textual analysis that the educational psychology literature calls holistic grading (Sudweeks, Reeve, and Bradshaw 2004;White 1985). Rather than breaking the text down by sentence or paragraph, as in traditional human-coded content analysis, coders read the text in its entirety and compare it to a rubric and a set of anchor texts that define the points of a measurement scale. The technique is especially useful for measuring latent, diffuse attributes of a text, such as we find in populist discourse. In our analysis, we had two coders read each speech in its original language and assign a score using an interval-level scale of 0 = little or no populism, 1 = moderate or inconsistent populism, and 2 = strong and consistent populism. A score of 1 means that the two essential elements of populist discourse-a reified popular will and a conspiring elite-are clearly present in the text, albeit inconsistently or weakly; a 2 means that these two elements are used consistently and clearly, coupled with ancillary features such as a bellicose tone; and a 0 means that one or both of these elements are absent. Final scores for the chief executive are an average of each coder's scores for all speeches in each year. The rubric and anchor texts are available in the online appendix, but the technique is also described and tested in Hawkins (2009).
The resulting data showed high intercoder reliability, even though coders were unable to see each other's work until the end of each round. For the historical dataset in Argentina, there was at least 83 percent agreement, a Cohen's kappa of .42-.60, and a Krippendorff's alpha of .41-.67; for the historical dataset in Chile, there was at least 96 percent agreement, a kappa of .60-.79, and an alpha of .68-.86; and for the dataset in Peru, there was at least 84 percent agreement, a kappa of .71, and an alpha of .78. These are substantial levels of intercoder reliability (Krippendorff 2013;Landis and Koch 1977). As additional checks on reliability and validity, we had coders fill out a detailed response for each speech in which they provide illustrative quotes and a short explanation for their judgment; we draw on these qualitative measures in the discussion that follows. Figures 1-3 present the data for the three countries we examined: Argentina, Chile, and Peru. Assessing the validity of this data is difficult, since there are no parallel datasets from other definitions that we can use for comparison. Instead, we compare the results with a variety of scholarly depictions of populism across the region and in these three countries, while highlighting ways in which the results confirm the ideational approach and cast doubts over some interpretations that are common in the scholarly debate. Figure 1 presents data on presidential populist discourse in Argentina during the twentieth century. As is true in the other two countries, populist discourse appears selectively and is associated with only a few well-known leaders. The fact that Perón shows up as populist is not especially surprising; his turbulent first presidency (including the collaboration of his wife, Eva Duarte) inspired much of the early theorizing about populism in Latin America. It reaffirms the depiction of his movement as one that sought to incorporate blue-collar workers into political life, the catalyst for a "critical juncture" in the country's democratic politics and especially its party system (R. B. Collier and Collier 1991). Perhaps more impressive is the fact that populism almost entirely disappears from Perón's discourse after his brief return to politics in 1973 and 1974, averaging a flat 0.0 (the cleft in the peak on the graph). His Peronist predecessors, Héctor Cámpora and Raúl Lastiri, as well as his successor, Isabel Martínez de Perón, were more clearly populist. This result confirms the common understanding that Perón moderated his rhetoric before returning from exile.

Argentina
Another interesting finding is the earlier incidence of populism under Hipólito Yrigoyen's first term (1918)(1919)(1920)(1921)(1922). Data here are admittedly thin-we have only one speech per year for most of this termbut they show a populist streak that distinguishes Yrigoyen's rhetoric from those of previous presidents representing the old oligarchic parties. Although political scientists have tended to focus on the role that Peronism played in unlocking Argentine democracy to the working class, Yrigoyen's Unión Cívica Radical (UCR) played a similar role of opening democracy to the middle class in the earlier decades of the twentieth century. As Rock (1975) notes in his depiction of the rise of the UCR, the party took an intransigent stance until 1912, demanding reforms that would end electoral corruption and empower the middle class before it would participate in elections. Crucially, Yrigoyen's ideological pronouncements and the party's printed platforms were populist in tone. Although vague about policy specifics, they claimed to be acting on behalf of "the people" and framed the demand for electoral reform in moral terms that condemned the existing regime as "a pile of decaying rubbish" (Rock 1975, 54).
One last aspect worth noting is the absence of populist discourse in the case of Carlos Menem. This finding contradicts the position of scholars such as Gibson (1997) and Weyland (1996), who have argued that Menem should be seen as a prototypical example of neopopulism in 1990s Latin America. We failed to find clear references in his speeches to the Manichaean distinction between the elite and the people. Menem's speeches in office often strike a highly redemptive tone, but they do so while assuming a very pluralist stance, as in this quote from his inaugural address at the Congress in 1989: "We are all responsible for and partners in the failure of Argentina. And together, only together, will we become instruments of a profound change. . . . From this moment on begins the reunification of all Argentines. The country of ' all against all' is over; the country of ' all together' begins." That said, Menem's speeches during his first presidential campaign in 1989 reveal a more populist streak. For example, in one of his campaign speeches, analyzed for the 2006 study (Hawkins 2009), he talks at length about the existence of a corrupt oligarchy constituted by the financiers and the party in government that has been acting against the will the people. In this speech (at the Luna Park Arena in Buenos Aires in March 1989) he proclaims: "Our people will not tolerate another situation like this. We don't want more perks and corruption. . . . In the city porteña [Buenos Aires] there are many men who . . . are always there with the bankbook and pen, checking the interest rate and the price of the dollar. . . . The Argentina that we want is an Argentina where the biblical command 'in the sweat of thy face shalt thou eat thy bread' becomes reality." We think these data confirm other scholarly accounts of Menem as someone who engaged in a major policy switch once in office, from opposing market-oriented reforms to enacting them wholeheartedly (e.g., Levitsky 2003;Stokes 2001). It should not surprise us that this policy switch was accompanied by a rhetorical shift away from populism. Scholars who claim that Menem's presidency was populist may have been reacting to his rhetoric during the election. Figure 2 presents the use of the populist discourse by Chilean presidents during the twentieth century. Several aspects are worth emphasizing. To begin with, although Chile is normally seen as a paradigmatic example of a country with a strong party system (Mainwaring and Scully 1995;Kitschelt et al. 2010), populism has not been absent. There are five clear moments in which presidents employed the populist Alessandri is often referred to by historians as the key early populist in Chile (S. Collier and Sater 2004;Drake 1999). The speeches we examined confirmed this depiction, as in the following quote from a public speech in the city of Talcahuano in December 1923: "This country has been governed by a ruling class and only a minority of its citizens have taken part in its government. . . . My coming to the Moneda [the presidential palace] means the end of one regime and the beginning of another: the democratic government by the people and for the people."

Chile
Much the same could be said for Ibáñez, who is labeled a populist in multiple accounts (Drake 1999;Fernández Abara 2007). Allende is a more surprising case; scholars generally do not categorize his presidency as populist (cf. Drake 1978). Nevertheless, Allende's speeches are rich with references to conspiracies by an oligarchy and its international allies and the eventual triumph of the people, which he argues that his government represents for the first time in Chile's history. The following extract from a speech in April 1972 is symptomatic: "To make democracy and liberty effective, we have destroyed the oligarchy's centers of power. . . . They never believed that we would destroy their privileges. They never believed that we would do away with their monopolistic plans. It's because they were used to deceiving the people, creating programs that they never believed in, mobilizing the masses with smoke and mirrors. Today, the people is the government." This finding not only confirms Laclau's earlier arguments about the ability to combine populism with Marxism but also sheds light on the reasons for Allende's dramatic rise and fall. His combination of socialist ideology and populist discourse enabled the mobilization of different social sectors and the construction of an alliance between rival political forces such as the Communist and Socialist Parties. As Angell (1998) has noted, at least until the end of the Cold War the history of the Latin American left reveals that the more it was able to combine Marxism with populism, the better were its chances of electoral success. However, this combination was difficult, because Marxism in Latin America often became equated with Soviet communism and a Leninist model of political organization that made it hard for populist discourse to be inclusive. Indeed, even at the height of Allende's popularity, his coalition struggled to achieve an absolute majority of votes. 4 This empirical finding raises the possibility of other leaders who have been able to combine populist and socialist discourses to develop broader electoral alliances. The obvious contemporary case that comes 4 As far as we know, Peter Winn (1986) is one of the few authors, who conceives of populism and socialism as different ideologies that were combined in a unique way by Allende's Popular Unity experiment.

Figure 2:
Populist presidential discourse in Chile, 1900Chile, -2000 to mind is Hugo Chávez in Venezuela, who in effect has also been categorized as populist according to the methodology used in this contribution (Brading 2012;Cannon 2009;Laclau 2006

Peru
Data for Peru are found in Figure 3. Our data suggest that Peruvian presidents expressed very little populist discourse until the latter part of the twentieth century. Except for a noticeable upturn during the first years of Manuel Odría's presidency (1948)(1949)(1950)(1951)(1952)(1953)(1954)(1955)(1956); the score here is heavily influenced by a single speech from his second year in office), there is no high, sustained populism until the military government of Juan Velasco Alvarado (1968)(1969)(1970)(1971)(1972)(1973)(1974)(1975). This does not mean that there were no populist movements earlier.
Historians typically describe Victor Raúl Haya de la Torre and the Alianza Popular Revolucionaria Americana (APRA) in the late 1920s and early 1930s as a populist movement (Stein 1980), and our own, unsystematic examination of a few of his early speeches (not included in the graph) suggests that Haya de la Torre had a strongly populist discourse during these years. 5 But no populist movement or leader was allowed to gain power until later in the century. This contradicts Collier and Collier's (1991) depiction of Peru's critical populist moment as having taken place around the election of 1945, when APRA supported the winning candidate, José Bustamante (himself not a populist by our measure). Whatever the change in the country's party system between the 1930s and the 1940s, the lower classes were not really allowed to compete in national politics on their own terms until at least two decades later. The left-populist government of Velasco Alvarado deserves some of the credit for this incorporation.
In any case, the strongly populist leaders that we do find are generally those described in the scholarly literature. Velasco Alvarado is often referred to as a populist or "inclusionary" leader who stands out from other military leaders in Latin America during the twentieth century (Stepan 1978;Dietz 1980). Indeed, he was an important influence on Hugo Chávez, who visited Peru as a young military cadet in the early 1970s and was profoundly impressed by the revolutionary government (Harnecker 2002). Our analysis shows the strength and consistency of Velasco Alvarado's populism, which only tapers off in the last year of his rule.
The other confirming instances are Alan García (1985García ( -1990 and Alberto Fujimori (1990Fujimori ( -2001. While these cases will not surprise most readers of this journal, it is satisfying to see that two leaders with very 5 Specifically, we looked at a speech from the 1931 campaign ("Discurso ante el proceso electoral"), one from 1945 ("El reencuentro") and another from 1946 ("A la clase obrera"). different economic policies (economists often mention García as a populist but almost never Fujimori) in fact had discourses with fundamentally similar elements. It is even more interesting to see the sharp break in Fujimori's discourse between his first and second terms, when populist rhetoric entirely disappears. Observers may be inclined to see this break as a sign of his disingenuousness and the increasingly authoritarian nature of his regime, but it may be that his initial discourse was radical because Fujimori lacked powerful allies; eventually he was able to make an alliance with sectors of the establishment to defeat his key enemies, namely the old political parties and the insurgency.

Analysis (of Other Definitions)
In line with the predictions of the ideational approach, populist discourse has a noteworthy presence across much of the past century among leaders and periods generally regarded as populist. It is found in various ideological combinations, including (more rarely) right-wing and socialist. In addition, it can be found even in countries with highly institutionalized party systems. And it is a selective, rather than constant feature of elite-level democratic discourse. In this section, we are interested in discussing the implications of our findings for the three other common conceptual approaches to Latin America populism. Even though Roberts (1995) and Weyland (1996) convincingly argue that populism should not be defined as a specific set of economic and/or social policies, there are still a few scholars who advance an economic conceptualization of populism. The classic work here is Dornbusch and Edwards's (1991) edited volume, in which they characterize Latin American populism as an "approach to economics that emphasizes growth and income redistribution and deemphasizes the risks of inflation and deficit finance, external constraints, and the reaction of economic agents to aggressive nonmarket policies" (1991,9). More recently, Edwards (2010) reaffirmed his argument by claiming that contemporary leftist presidents in Latin America are implementing "populist policies" that will certainly lead to a cycle from euphoria to regret. In addition, Acemoglu, Egorov, and Sonin (2013, 772) argue that populism is "the implementation of policies receiving support from a significant fraction of the population, but ultimately hurting the economic interests of this majority." The chief deficiency we see in economic definitions is that they limit populism to leftist or inclusionary forms (which are more common in poor countries) and effectively exclude right-wing populists (which are more common in rich countries). The clearest example of right-wing populism in our historical data is Fujimori. After winning office with a more traditional populist message of opposition to neoliberalism, Fujimori imposed a series of market-oriented reforms that brought inflation in check and resulted in a decade of stellar economic growth and low inflation, although wages failed to grow for many of the poor and income distribution worsened. The economics literature hardly ever refers to Fujimori as a populist, and the few references that do exist are careful to suggest that he was populist only during his campaign (Durand 2002;Martinelli 2002). Nevertheless, other social scientists widely perceived him as a populist through much of his time in office (Kay 1996;Roberts 1995;Weyland 1996). Our own analysis finds that Fujimori spoke the language of populism throughout his first term, often to the very business elites who supported his economic policies.
Furthermore, our findings confirm the position of other scholars who argue that structuralist definitions defended by authors such as Cardoso and Faletto (1979), di Tella (1965), Germani (1978), and Oxhorn (1998) are overly narrow. Structuralist definitions regard populism as a historically specific type of political regime that was the expression of an alliance between the urban proletariat, factions of the middle class, and sectors of the bourgeoisie with the purpose of promoting an interventionist state. By seeing populism as a type of regime, those who adopt this definition overlook the fact that populist discourse can be employed by political actors and constituencies who are in opposition, and who may never have the opportunity to fully structure a heterogeneous coalition around actually implemented policies. Even though our empirical analysis deals only with leaders who have been in power, the ideational approach permits the study of populist actors and parties that are outside of government. In fact, almost all European forces that adhere to the populist worldview obtain a relatively small amount of votes and, in consequence, are not able to conquer the executive branch (Mudde 2007). We see contemporary examples of minority populist parties in countries such as Ecuador (Álvaro Noboa's PRIAN), Bolivia (Max Fernández's Unión Cívica Solidaridad and Carlos Palenque's Conciencia de Patria), and Chile (Roxana Miranda's Partido Igualdad). While unsuccessful at winning the presidency, many of these shaped the policy agenda and created opportunities for populist parties that followed.
What is perhaps more problematic for the structuralist definitions is their excessive emphasis on the multiclass nature of populism. While there is little doubt that populist chief executives in Latin America have been capable of constructing an alliance with different socioeconomic sectors, it is problematic to assume that the development of multiclass coalitions is a defining attribute. Populist radical right parties in Europe employ the populist set of ideas, but their voters are not very diverse in terms of class; most of them have a working-class background (Mudde 2007;Bornschier and Kriesi 2013). In our own analysis we see something similar in the case of Irigoyen's UCR. This party was closely linked to Argentina's emerging middle class, with almost none of the appeal to the working class that emerged later under Perón.
Finally, our findings argue against the political-strategic definition. The political-strategic definition is not necessarily opposed to a consideration of ideas and rhetoric (Barr 2009;Levitsky and Loxton 2013), but it tends to see the discourse of populists as less important predictors of their behavior. It privileges organizational features, especially charismatic leadership, outsider status, and top-down and uninstitutionalized organization (Weyland 2001). What this definition fails to notice is that populism is first and foremost a moral set of ideas that is shared by different constituencies, who have emotional and rational motives for adhering to the Manichaean worldview inherent in populism. Otherwise stated, those who supported Alessandri in Chile in the 1920s or Perón in Argentina in the 1940s did not vote for them simply because they were enchanted by a charismatic leader. These voters were pleased to see that someone was finally taking into account their own ideas about how politics should function. In this respect, it is worth noting that the word "populism" comes from two movements of the late nineteenth century characterized by their leaderless nature: the US people's party and the Russian Narodnik (Taggart 2000). These cases reveal that manifestations of populism do not necessarily lead to the rise of strong and charismatic leaders, while the literature on populist radical right parties in contemporary Europe has demonstrated that populist forces can advance within strong party organizations (e.g., Art 2011;Mudde 2007).
Furthermore, political-strategic definitions put too much emphasis on the outsider status of populist leaders. The first thing we notice in our historical data is that several of the populist presidents were clearly not political outsiders in the objective sense of this term, that of a leader who has never held high elective or appointive office. There are at least two clear populist insiders in our dataset: Alessandri, who was already a member of Congress and a leader of the Liberal Party before winning the presidency in 1920, and García, who was a leader of APRA and had served in the Peruvian congress before first winning the presidency in 1985. Only Velasco Alvarado and Fujimori were complete political outsiders in the strictest sense. Yet even if we relax this strict objective criterion and consider only elective office, there is the problem of incumbency. All of the populists in our historical dataset were presumably experienced politicians when they ran again for office later on, as did all of them except Velasco Alvarado. Even Velasco Alvarado was in office for almost seven years. At what point do outsiders become insiders?
In the end, we think that objective outsider status is irrelevant to the definition of populism. Whatever their actual political biography, all populists frame themselves as outsiders (Mudde and Rovira Kaltwasser 2017: 68-78). Indeed, all of the populists in our dataset claimed a kind of unity with the popular will and proclaimed themselves the defenders of that will against some conspiring class of elites, whether or not they were objectively political insiders at some point in their careers. Consider, for example, Alessandri. Although a political insider because of his connections to the Liberal Party and his years as a member of Congress, he frequently denounced powerful enemies that he claimed were conspiring against the will of the people. 6

Conclusion: The Lessons of Alternative Approaches
By discussing the empirical measure of populist discourse for samples of past and contemporary chief executives in Latin America, we have tried to convince skeptics that defining populism in ideational terms is analytically and methodologically convenient. Measures of populism drawn from a systematic reading of political speeches not only do an impressive job of capturing scholarly wisdom, but they help reveal the shortcomings of alternative definitions. By focusing on the ideas that underlie populist parties and movements, we put ourselves in a better position to identify contingent behaviors and explain them.
However, our point is not that the alternative definitions championed by Latin Americanists lack theoretical insight or connections to discourse. We think that alternative definitions provide important insights for future research into the causes and consequences of populism. The empirical results that we presented and discussed here help us to identify populist presidents across the twentieth century in Argentina, Chile, and Peru, but they do not tell us why these leaders came to power or why their administrations had a positive or negative impact on democracy. As scholars using the ideational approach move on to these questions of causality, they will need to draw from the insights of these alternative approaches.
For example, political-strategic definitions in Latin America point out the connection between populist discourse and organization. They remind us that while populism can manifest itself through institutionalized parties, the most electorally successful instances of populism in Latin America are broader movements that coalesce around charismatic leaders. While we do not see charismatic leadership or movement organization as definitional, we do think that charisma is a necessary explanation for the electoral success of populist movements. Populist discourse can generate great enthusiasm-every citizen must become an activist with a voice in the movement-but it creates problems of coordination that are not easily resolved with professionalized, hierarchical organizations. As the mystic embodiment of the voice of the people (Laclau 2005), a charismatic leader provides an unmediated connection between the people and the government while simultaneously providing direction and control of the movement.
Economic definitions also make important contributions to the comparative study of populism. While we think that economists err in associating populism only with short-sighted redistributive policies, we think they highlight the issue of blame attribution. The most radical instances of populism often manifest a pattern of dogmatic, unscientific thinking in which policies are often made based on the mobilization of emotions with the aim of highlighting the fraudulent behavior of the establishment. Hofstadter (1965) famously described this "paranoid" side of populism. Part of this "paranoia" reflects an anti-elitist disregard for professionalized training and a celebration of popular know-how and "common sense," but another part represents an inclination to the development of conspiracy theories. This raises serious questions about the quality of deliberation and the civic properties of voters, let alone their economic judgment, and it suggests that we should be giving more attention to the psychology of populist participation.
Finally, structuralist definitions point us to the deeper, historical roots of populism's emergence and success. Causal theorizing was a prominent feature of early studies by Latin Americanists. Yet in recent years scholars have paid surprisingly little attention to the causes of populism, both in the sense of macro-conditions that make populist movements likely to emerge and succeed, and in terms of microconditions that make some individuals more open to populist rhetoric. Thinking about populist ideas redirects our attention to these larger causal questions. While we disagree with the structuralists' tendency to connect populism to late modernization, we agree that there is something about the low quality of democratic governance in developing countries that makes populist movements attractive (de la Torre 2010; Del Tronco 2013; Hawkins 2010; Slater 2013). Populist discourse is, after all, a claim that democratic norms of citizenship and equality before the law have been systematically violated, and these are longterm problems in much of Latin America and the developing world. Hence, one could argue that different political cultures and historical legacies play an important role in obstructing or facilitating the emergence of populist discourses.
Incorporating all of these insights into a comprehensive theory of populism's causes and consequences is a challenge for proponents of the ideational approach and those of alternative definitions. It requires looking beyond the discourse of presidents to that of other elite-level actors such as opposition politicians and civil society activists. And it requires moving beyond elite-level data to the thinking of ordinary citizens, viewable through survey data, ethnographic research, and experiments (Akkerman, Mudde, and Zaslove 2014;Fernandes 2010;Hameleers, Bos, and de Vreese 2016;Hawkins, Riding, and Mudde 2012). Putting it another way, it requires looking at both the supply of and the demand for populism, as well as how these interact.
Ultimately, we think Latin America-and Latin Americanists-still have a great deal to offer to the study of populism. The work of scholarly pioneers from the region continues to inspire us. And populism remains a prominent feature of politics in the region, far more than in Western Europe or portions of North America, where populist actors have become salient in the last few years. Yet the scholarly landscape has shifted, and contributions to populism studies can best be made through a cross-regional, comparative engagement with scholars and cases from outside. Only in this way can we distinguish local features from global patterns.