Ever since Simón Bolívar envisioned a Latin American confederation, as famously described in his Carta de Jamaica (Reference Bolívar1815), the economic, political, and societal unification of Latin America has been a recurrent theme in public debates on the continent. Over the past few decades, countries in South and Central America have again come closer to realizing such a union, as they entered deeper into a growing network of agreements and consultations, created multinational infrastructure projects and energy consortia, formed free trade areas and customs unions, and finally engaged in political integration. With the founding of the Union of South American Nations (UNASUR) in 2008, the unification of Latin America has reached a new stage. Twelve Latin American states joined forces in a union that covers 17 million square kilometers, almost 400 million inhabitants who generate a GDP of 1,000 billion US dollars per year (Reference Flemes and WestermannFlemes and Westermann 2009). Article 2 of UNASUR’s Constitutive Treaty states that the “objective of the South American Union of Nations is to build, in a participatory and consensual manner, an integration and union among its peoples in the cultural, social, economic and political fields” (UNASUR 2008). Today, Latin America might thus be closer to the “confederation” Bolívar dreamt of than ever before.
At the same time, however, it is clear that regional integration in Latin America still has a long way to go. Many of its organizations are as yet in their infancy and the recently founded Parlamento del Mercosur (PARLASUR) is devoid of power (Reference DabèneDabène 2009, 153; Reference Hurrelmann and SchneiderHurrelmann and Schneider 2015, 8). Some scholars have even stated that regional integration in Latin America has failed (Reference MalamudMalamud 2013). Others have been more optimistic (Reference Bulmer-ThomasBulmer-Thomas 2001, 368; Reference Riggirozzi and GrugelRiggirozzi and Grugel 2015, 781), but one thing is clear: if regional integration is to continue to deepen in the future, it will become more and more important to consider public opinion, which has the power to sway integration (Reference Gabel and PalmerGabel and Palmer 1995; Reference GabelGabel 1998). In Europe, for instance, the negative outcomes of referenda on the Treaty on European Union in Denmark in 1992 and on a European constitution in France and the Netherlands in 2005 led to significant decelerations of the integration process (Reference Gabel and WhittenGabel and Whitten 1997), showing how central grassroots support becomes in the course of regional integration. Tellingly, the term democratic deficit, which was originally only used with reference to the European Union (Reference BoyceBoyce 1993), has recently been transferred to the context of Latin American integration (Reference EstradesEstrades 2006; Reference SerbinSerbin 2011). Latin American integration projects now “have become the object of explicit legitimation debates” (Reference Hurrelmann and SchneiderHurrelmann and Schneider 2015, 251).
Yet, whereas in Europe the development of public opinion on regional integration over time has been studied for decades and can be traced back without major gaps until the 1950s (Reference KriesbergKriesberg 1959; Reference InglehartInglehart 1977; Reference Anderson and KaltenthalerAnderson and Kaltenthaler 1996; Reference Eichenberg and DaltonEichenberg and Dalton 2007), public opinion toward regional integration in Latin America has until now only been studied sporadically, at single time points, and mostly with respect to single countries. What is lacking is an encompassing analysis of long-term trends. The first goal of this article is to contribute to filling this gap. Combining data from eight Latinobarometer surveys from 1997 to 2010, we provide the first longitudinal analysis of public opinion toward Latin American economic and political integration among citizens of seventeen South and Central American countries.
The second goal of this study is to investigate the social structure behind this support for Latin American integration, as well as its potential transformation over time. Apart from examining major societal cleavages such as gender and education, we are particularly interested in analyzing whether there is a connection between the swing to the left that occurred in Latin American politics during the same time period (also known as the “pink tide”), in which a considerable number of left and center-left presidents became democratically elected in the region (Reference CastañedaCastañeda 2006; Reference RodriguezRodriguez 2014; Reference KraulKraul 2016), and support for regional integration. Traditionally, the political center and right have been more supportive of economic and political integration than the left (Reference Inglehart, Rabier and ReifInglehart, Rabier, and Reif 1987; Reference GabelGabel 1998; Reference Milner and JudkinsMilner and Judkins 2004). But what if the integration projects are increasingly driven by leftist governments as in Latin America during this period? Have left-leaning Latin Americans become more supportive of integration over the years as a consequence? By looking into these questions, our study provides new insights not only into the changing structure of public opinion on Latin American integration, but also into the relation between political orientation and support for regionalism in general.
State of Research and Conceptualization
Research on public support for regional economic and political integration in Latin America is still scarce. Longitudinal analyses do not exist as yet and most studies are limited to specific countries or societal groups. Davis, Gabel, and Coleman (Reference Davis, Gabel and Coleman1998) examined public support for Central American integration in Costa Rica and El Salvador using local 1994 surveys. Estrades (Reference Estrades2006) studied support for Mercosur among Uruguayans using 2004 International Social Survey Programme data. Díaz-Domínguez (Reference Díaz-Domínguez2010) looked at the effect of religion on attitudes toward free trade in Latin America, based on the 2008 AmericasBarometer. Nitsch Bressan (Reference Nitsch Bressan2010) researched the attitudes of national elites in Bolivia, Brazil, Chile, and Venezuela toward the Andean Community, ALBA (Bolivarian Alliance for the Peoples of Our America), Mercosur, and other entities using NUPRI (Núcleo de Pesquisa em Relações Internacionais) data. Similarly, Alcántara Sáez (Reference Alcántara Sáez2000) studied support for economic integration in South America among political elites. In a smaller, self-conducted research project, Stiller Titchener (Reference Stiller Titchener2010) examined Bolivians’ attitudes toward cooperation with neighboring countries and knowledge of regional integration efforts.
The first spatially encompassing study was conducted by Seligson (Reference Seligson1999), who examined support for economic regional integration in seventeen Latin American countries using Latinobarometer (1996) data and identified a range of micro-level predictors. His article, however, is now outdated and neglects potentially relevant country-level determinants. Furthermore, it does not discuss support for political integration, which is problematic since recent integration efforts in Latin America have been described as “post-trade regionalism” (Reference RiggirozziRiggirozzi 2010, 2) that is “firmly political … in character” (Reference Riggirozzi and GrugelRiggirozzi and Grugel 2015, 781). The first attempt to look at support for both economic and political integration while taking micro- and macro-level determinants into account was conducted by Jara Ibarra (Reference Jara Ibarra2014), who applies multilevel models to 2009 Latinobarometer data. However, by looking at only one point in time, she can’t discern any longitudinal trends.
Before conducting our longitudinal analysis, we use the existing studies—augmented by insights from research on European integration where necessary—to identify potentially relevant cleavages within societies, macro-level influencers, expectable trends, and underlying theoretical mechanisms.
Intra-societal Cleavages
Based on the existing literature, eight individual-level factors seem to influence support rates for regional integration. The most relevant cleavage in the context of this study is political orientation, a major determinant of support for integration (Reference GabelGabel 1998). Milner and Judkins (Reference Milner and Judkins2004) showed that right-wing parties in developed countries are consistently more supportive of free trade than are left-wing parties. For support for European integration, Inglehart, Rabier, and Reif (Reference Inglehart, Rabier and Reif1987) revealed that right-wing voters were more supportive of the European Community than were left-wing voters, while Gabel (Reference Gabel1998) demonstrated that respondents who identified with a proletariat party were less supportive of integration than those identifying with a bourgeois party. Different attitudes toward protectionism, neoliberal market reform, hegemony, and international solidarity could underlie these political differences. For Latin America, Magaloni and Romero (Reference Magaloni and Romero2008) showed that voters of left-wing parties were less supportive of free trade in the 1990s. By contrast, Jara Ibarra (Reference Jara Ibarra2014) found that in 2009, left-wing Latin Americans were more supportive of both economic and political integration than the right. This divergence indicates that a shift from right- to left-wing support for integration may have taken place in Latin America.Footnote 1 As we discuss in more detail below, this shift may be the result of the swing to the left that occurred in Latin American politics from the late 1990s.
A second factor of interest is gender. Studies show that women are less likely than men to support European (Reference Glüpker-KesebirGlüpker-Kesebir 2015) and Latin American (Reference SeligsonSeligson 1999) integration. To determine mechanisms behind this gap, Glüpker-Kesebir (Reference Glüpker-Kesebir2015) tested two competing hypotheses, a socialization-based as well as a utilitarian, self-interest-based explanation. Her research supports the latter, which implies that women’s lower support rates are mainly due to their exclusion from core labor markets.
Education is another central social cleavage. Seligson (Reference Seligson1999) found that in 1995, education clearly stratified support for Latin American economic integration, with the least educated being least favorable. Similar findings have been reported for support for free trade (Reference BakerBaker 2003) and support for political integration (Reference Jara IbarraJara Ibarra 2014) in Latin America. A potential mechanism behind this pattern was proposed by Inglehart (Reference Inglehart1970a), who argued that “cognitive mobilization” that comes with higher education increases support for European integration. Fittingly, Hainmueller and Hiscox (Reference Hainmueller and Hiscox2006, 469) argued that the effects of education on support for free trade result from the “exposure to economic ideas and information among college-educated individuals.”
A closely related factor of relevance is occupational class. With respect to support for European integration, a large gap between more favorable elites and less favorable masses is widely acknowledged (Reference HoogheHooghe 2003). In Latin America, Arenas (Reference Arenas1997, 11) argued that “las elites están mucho más conectadas con lo que pasa en el globo que los grupos más desfavorecidos”; Guilhon Albuquerque (Reference Guilhon Albuquerque2009, 22) observed “uma inclinação das elites sul-americanas para um discurso fortemente integracionista”; and Zizumbo-Colunga and Seligson (Reference Zizumbo-Colunga, Seligson, Hurrelmann and Schneider2015, 117) stated that regional organizations in Latin America “were advocated by elites in a top-down process.” Seligson (Reference Seligson1999) showed empirically that the higher occupational classes in Latin America were more supportive of economic integration than were the lower ones (although this effect disappeared once other variables were controlled for). Two different mechanisms have been proposed to explain gaps in support rates between occupational classes: the job-based argument states, based on the Heckscher-Ohlin trade model, that trade liberalization hurts individuals with scarce factor assets while benefiting people with abundant factor assets. In developing parts of the world like Latin America, skilled labor is a scarce factor asset, while unskilled labor is abundant. Hence, this hypothesis predicts that support for free trade will be highest among unskilled laborers—which is however the opposite of what the empirical evidence suggests. The consumption-based argument, on the other hand, states that (a) economic integration increases the quality, variety, and affordability of products; and (b) decisions to support free trade are influenced by the chances to benefit from these improved opportunities for consumption. Accordingly, support for integration should be lower among those with humble purchasing power, that is, the lower occupational classes (Reference BakerBaker 2003, Reference Zizumbo-Colunga, Seligson, Hurrelmann and SchneiderZizumbo-Colunga and Seligson 2015). This argument is more in line with the prevailing empirical evidence.
Over and above the broad occupational class, the personal economic situation may be relevant. Empirically, support for economic integration (Reference SeligsonSeligson 1999) and free trade areas (Reference Zizumbo-Colunga, Seligson, Hurrelmann and SchneiderZizumbo-Colunga and Seligson 2015, 122) is higher among Latin Americans with better personal economic situation. Furthermore, studies show that the (perceived) economic performance of the country also influences attitudes toward integration positively in both Europe (Reference Eichenberg and DaltonEichenberg and Dalton 1993; Reference Anderson and KaltenthalerAnderson and Kaltenthaler 1996) and Latin America (Reference SeligsonSeligson 1999; Reference Zizumbo-Colunga, Seligson, Hurrelmann and SchneiderZizumbo-Colunga and Seligson 2015).
Another aspect of relevance is satisfaction with the functioning of democracy. Seligson (Reference Seligson1999) showed that people who were satisfied with how democracy worked in their country were more supportive of economic integration. While he did not specify any mechanisms behind this relation, we could easily argue that people who regard their government’s actions as democratically legitimatized also accept these governments’ decisions to integrate as legitimate. Furthermore, the argument can be linked to Inglehart’s political values theory (cf. Reference GabelGabel 1998). Inglehart (Reference Inglehart1970b) argued that when institutionalized competition between alternative groups of decision-makers exists (i.e., when democracy works), these groups will bid for societal support. Therefore, support for integration should be higher when people feel that democracy functions well.
A final factor is age, which has proven to be positively related to support for integration in Europe (Reference GabelGabel 1998). In the Latin American case, however, the evidence is rather mixed: Davis, Gabel, and Coleman (Reference Davis, Gabel and Coleman1998) report a positive effect of age on support for economic integration among El Salvadorians, but no effect among Costa Ricans. Neither Seligson (Reference Seligson1999) nor Jara Ibarra (Reference Jara Ibarra2014) found significant age effects for public opinion on Latin American integration.
Macro-level Cleavages
Macro-level factors have thus far been largely neglected by research on Latin American integration, but a look at the European case reveals that three such factors stand out as particularly promising. First, the dependence of the economy on foreign trade, that is, its connectedness to other countries via imports and exports, seems worth considering. The argument that people in countries whose economies are more dependent on trade with other countries are likely to be more open to integration has a long history and can be traced back to Kant (Reference Kant1903 [1795]) and Karl Deutsch’s transactionalist theory (Reference Deutsch, Burrell, Kann, Lee, Lichterman, Lindgren, Loewenheim and Van WagenenDeutsch et al. 1957).
Closely related to this factor is the influence of other countries via foreign direct investments (FDIs). However, while trade is often seen as benefiting both partners, or at least as creating mutual dependencies, FDIs are more likely based on a one-sided and hierarchical influence by a powerful hegemon. Accordingly, FDIs have been described as a central means of the United States to ensure its influence on “its” continent (Reference Kohl and FarthingKohl and Farthing 2006, 182), and “control” has even been specified as a defining element of FDIs (Financial Times, n.d.). It is important to see this factor in light of the shift that has taken place in Latin American regionalism. “Hemispheric cooperation” that includes the United States has long been the “other” form of regional integration on the continent, in competition with pan–Latin American regionalism (Reference HurrellHurrell 1992, 122). Recently, however, the “declining ability of the US to shape regional orders, institutions and discourses” (Reference RiggirozziRiggirozzi 2010, 2) has led observers to describe current Latin American integration efforts as “post-hegemonic” (Reference Riggirozzi and TussieRiggirozzi and Tussie 2012), or even “counter-hegemonic regionalism” (Reference MurhMurh 2010). FDIs do not seem to mesh with this new spirit and might therefore be negatively related to support for integration.
Finally, the question of whether membership in intergovernmental organizations (IGOs) that represent current integration efforts affects support rates for integration was already identified by Seligson (Reference Seligson1999) as an interesting direction for future research. For Europe, Anderson and Kaltenthaler (Reference Anderson and Kaltenthaler1996) showed that timing of entry and length of membership in the European Union influence support for European integration positively. The Latin American case is more complex, though, since there is not just one organization that states can join. Instead, Latin America “offers alternative pathways to region building” (Reference RiggirozziRiggirozzi 2010, 2), or, expressed negatively, constitutes a “long-lasting patchwork of segmented regionalisms” (Reference MalamudMalamud 2013, 4). Hence, membership could have different effects depending on the specific organization of which a country is member. For instance, a “singularity that distinguishes ALBA from the others” is its anticapitalist narrative, which “has significant potential to consolidate a social dimension in the integration process” (Reference RiggirozziRiggirozzi 2010, 13), and promises “a new kind of integration between peoples” (Reference DabèneDabène 2009, 156, emphasis added). Mercosur, on the other hand, has been quite successful in integrating the participating countries economically, and quickly became popular among entrepreneurs and the private sector in general (Reference DabèneDabène 2009, 164). Accordingly, citizens of Mercosur countries might be particularly fond of economic integration, whereas political integration might be more popular among ALBA member citizenries due to ALBA’s extensive social politics.
Predicting Longitudinal Trends
There seem to be reasonable arguments both for and against increasing public support for Latin American integration between 1997 and 2010. A point in favor of an upward trend is that simultaneously with the founding of several new institutions over the first decade of the new millennium (ALBA, UNASUR), poverty and inequality rates dropped and the middle classes grew in most of the region’s societies (Reference Ferreira, Messina, Rigolini, López-Calva, Lugo, Vakis and LóFerreira et al. 2012; Reference Riggirozzi and GrugelRiggirozzi and Grugel 2015, 788). Moreover, many of the projects (of ALBA for instance) were purposefully intended to consolidate public support (Reference Flemes and WestermannFlemes and Westermann 2009). Thus, people throughout the continent could now be more familiar with integration or may even have experienced its positive effects (perceived or real). However, as we know from the European case, a deepening and democratization of regional integration does not necessarily translate into increased support rates. To the contrary, while the European Parliament continuously gained more influence from the 1960s on, voter interest and turnout dropped steadily (Reference DabèneDabène 2009, 153; European Parliament 2014). Furthermore, some scholars today regard Latin American integration as failed (Reference MalamudMalamud 2013). Due to these antithetical arguments, it is difficult to predict the direction of public support for Latin American integration over time ex ante.
It is also hard to foresee how the two trend lines—support for economic integration and support for political integration—might relate to each other. Modern economic integration has a longer history than political regionalism in Latin America (as in other parts of the world): the customs unions Andean Community and Mercosur existed long before UNASUR was created. If we accept the assumptions that people are generally more skeptical toward the unknown and that sharing political power is often perceived as more severe a step than a mere increase in trade with neighboring countries, we may expect support rates for economic integration to be higher than those for political integration. Yet, given that what was unknown yesterday is not so unfamiliar today, a convergence between the two trends might occur. On the other hand, the paradoxical effect of decreasing support with deepening regionalism known from the European case may also apply to political integration. Therefore, public support for economic integration could remain consistently higher than support for political integration. Thus it is not clear whether this potential gap diminishes over time.
But what about social cleavages in public support for integration; will they have closed or widened? To answer this question, we have to consider again how the “pink tide” in Latin American politics affected regional integration projects. Concomitant with the swing to the left, Latin America saw a shift from neoliberal approaches to regional integration to alternative strategies that focus on poverty reduction and inclusive growth: “Latin American regionalism is in the process of ‘catching up’ with social concerns, moving away from the exclusive emphasis on economic drivers of integration toward a broader agenda driven by social goals” (Reference Riggirozzi, Hurrelmann and SchneiderRiggirozzi 2015, 229). This new form of appearance may have affected the political base of support for regionalism and, if successful, diminished gaps in support between social groups. Let’s have a closer look at political orientation, gender, and education.
Political Orientation
As we have seen above, the left has traditionally been less in favor of free trade and regional integration than the center and the right. Recent Latin American integration projects, however, have been driven by left-wing governments as part of the region’s general shift to the left (Reference SeligsonSeligson 2007). UNASUR, for instance, has “as its chief strength its close fit, or at least its capacity to engage, with the new architecture of centre-left democracy in South America” (Reference Riggirozzi and GrugelRiggirozzi and Grugel 2015, 787). This fit implies a focus on defending democracy and human rights, improving health care, supporting social inclusion, and addressing “the needs of the region’s poor and most vulnerable citizens” (Reference Riggirozzi and GrugelRiggirozzi and Grugel 2015, 782, 788). ALBA, in turn, was the brainchild of the radical, socialist left, spearheaded by Chavez’s Venezuela. It involves literacy and healthcare programs (Reference Flemes and WestermannFlemes and Westermann 2009) just as an anti-neoliberal agenda that puts emphasis on “greater trade, solidarity and exchange outside of the usual market-based strictures” (Reference KozloffKozloff 2008, 54). Hence, these regional integration projects were not only driven by left-wing governments, they also took forms of appearance that should appeal to the left. Not only the actors behind, but also the substance of Latin American regionalism changed. Since these organizations were founded within our time frame (in 2008 and 2004, respectively), we might accordingly see a shift toward increasing support by left-leaning respondents (accompanied by decreasing support from center- and right-leaning interviewees). The general “politicization” (Reference Hurrelmann and SchneiderHurrelmann and Schneider 2015, 251) or rather “re-politicization” (Reference RiggirozziRiggirozzi 2010, 2) of regional integration in Latin America might further intensify this swing to the left.
Gender
Although inequality and “machismo” remain central features of gender relations in many Latin American societies, the region has made enormous progress in recent years and has in some aspects of gender equality overtaken the United States and drawn level with Europe (Reference Lopez TorregrosaLopez Torregrosa 2012). These transformations could also have repercussions in support rates for integration. We therefore expect the gender gap in support for economic and political integration to decrease over time.
Education
Some of the continent’s current integration projects, ALBA in particular, have put emphasis on projects that aim at the least educated (Reference Flemes and WestermannFlemes and Westermann 2009, 4), which could have increased support rates among these social groups, whereas support among the more highly educated might be bound to a “natural” upper limit (ceiling effect). Therefore, it seems plausible that the gap between the less and the better educated in support for integration has also declined.
Table 1 summarizes the discussed factors, the anticipated direction of the corresponding effects, and potential underlying mechanisms.
Factors and expected effects on support for integration.
Note: + positive effect, – negative effect, +/– unclear effect. For interactions, + denotes closing gaps. GDP = Gross Domestic Product, FDI = Foreign Direct Investment, IGO = Intergovernmental organization, ALBA = Bolivarian Alliance for the Peoples of Our America.
Research Design
Data
The Latinobarometer is a cross-sectional annual survey conducted in up to eighteen Latin American countries (Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, Venezuela) and Spain since 1995. In order to increase the comparability between annual waves, we exclude the Dominican Republic, which is not contained in pre-2005 surveys, and Spain, where the integration questions were not asked (and which is European, after all), from the analysis. Based on the availability of the dependent variables, waves from 1997, 1998, 2001, 2002, 2005, 2008, 2009, and 2010 could be used for analyzing support for economic integration (106,590 respondents), and waves from 2002, 2008, 2009, and 2010 were available to study support for political integration (54,256 respondents).
Dependent Variables
The dependent variables are “support for Latin American economic integration” and “support for Latin American political integration.” Related questions appear irregularly in the Latinobarometers between 1995 and 2010. Appendix Table A1 provides an overview of the occurrences and exact phrasings, which vary slightly over the years.Footnote 2 We recoded the response categories into a binary variable in which the original categories [1] and [2] are merged into “in favor” [1] and all other categories are recoded as “not in favor” [0].
Independent Variables
At the micro level, we implemented eight independent variables. Political orientation was originally measured on an eleven-point scale. In line with common practice (Reference RodriguezRodriguez 2014, 66), we recoded it into a categorical variable with the characteristics “left” [0–3], “center” [4–6], and “right” [7–10].Footnote 3 Gender is binary in this dataset with 1 = female and 0 = male. Education measures the educational level of the respondent and consists of three merged categories: primary or illiterate (“primary or less”), secondary or incomplete secondary (“secondary”), and tertiary or incomplete tertiary education (“tertiary”). Occupational class consists of eleven categories: high-level executive, independent professions, mid-level executive, salaried professions, employee, business owner, independent worker, retired, doesn’t work, homemaker, and student.Footnote 4 Personal economic situation measures the perception of being able to cater for the family’s needs with the respondent’s and the family’s income. Participants self-rate on a four-point scale from “can save” to “great difficulties.” Perceived economic situation of the country is originally a five-point-scaled variable with categories ranging from “very good” to “very bad,” which we recoded into the three categories “good,” “neither,” and “bad,” since the extreme categories contained only few cases. We also recoded satisfaction with democracy into a dummy variable (1 = satisfied). Age in years is continuous and mean-centered.
At the macro level, we included five variables. Dependence on trade is measured via trade as a share of the gross domestic product (GDP) based on World Bank (2016) data. Information on FDIs as a share of the GDP was obtained from the United Nations Conference on Trade and Development (UNCTAD 2016). To measure membership in IGOs, we created binary variables for Mercosur and ALBA, in which 1 indicates membership of country i in year t, whereas 0 denotes no membership (associate membership also counts as 0). Furthermore, we implemented year as a continuous variable to account for time effects.
Methods
We draw on multilevel mixed-effects logistic regression specified for longitudinal cross-sectional survey data (Reference FairbrotherFairbrother 2014). The advantage of multilevel models is that they account for group membership of observations and thus include higher-level variances. They thus allow us to circumvent atomistic and ecological fallacies that would occur if findings were mistakenly transferred from one level to another (Reference HoxHox 2010). By including random effects on all relevant levels, we control for these biases. Hence, this study takes into consideration that individuals are nested in country-years, and the latter are again nested in countries (for this approach, see Reference Schmidt-Catran and FairbrotherSchmidt-Catran and Fairbrother 2016). While we would ideally have more than seventeen countries to include at the higher level (Reference StegmuellerStegmueller 2013), it is not uncommon in the social sciences to apply multilevel modelling to sets of countries smaller than twenty (Reference FairbrotherFairbrother 2014, 120). Maas and Hox (Reference Maas and Hox2005) showed that while small numbers of higher-level units tend to bias error terms, the bias is small, except for extremely small numbers.
In order to disentangle longitudinal effects (within effects) from cross-sectional effects (between effects) at the macro level, we define our multilevel mixed-effects logistic regression as a hybrid model (Reference Frondel and VanceFrondel and Vance 2010; Reference FairbrotherFairbrother 2014). First, by calculating means across all years for each country and macro-variable, we measure the extent to which changes in political and economic integration depend on differences between countries (between effect). Second, to measure how longitudinal changes in the macro variables affect attitudes toward integration, we subtract the year-specific values from the corresponding country mean (within effect). This procedure allows us to control for unobserved heterogeneity by estimating longitudinal fixed effectsFootnote 5 (or within effects) for countries over time, while simultaneously accounting for the cross-sectional nature of the Latinobarometer and observed differences between countries.
In order to measure how effects of specific individual-level variables have changed over time, we introduce interaction effects between the respective variables and year, which we then illustrate graphically using marginal effect plots at representative values for subpopulations.Footnote 6 We use the pwcompare command in Stata to test whether the estimates are significantly different from each other.
Results
The results are presented in three consecutive steps. We first look descriptively at the overall trends in public opinion on Latin American integration, then examine the predictors of support for integration, and finally scrutinize in more detail whether and how gender, educational, and political orientation cleavages in support rates changed over time.
General Trends
The development of support for Latin American integration between 1997 and 2010 is illustrated in Figure 1, first for each country individually and then for the region as a whole. Three things are notable. First, both forms of integration are generally seen favorably by the majority of Latin Americans: values range from 90.7 percent (1997 Colombia) to 48.6 percent (2005 Ecuador) for economic integration and from 84.6 (2002 Nicaragua) to 40.4 percent (2008 Honduras) for political integration with means of 72.6 percent and 63.1 percent respectively. In every country, the average support rate across time is greater than 60 percent for economic integration and greater than 50 percent for political integration.
Support for Latin American economic and political integration, 1997–2010.
Note: Survey weights (‘wt’) are applied.
Second, there is no universal trend toward more support for integration. To the contrary, in most countries (twelve out of seventeen), support for economic integration was lower in 2010 than in 1997 and support for political integration decreased in nearly all countries (fourteen out of seventeen) between 2002 and 2010. Overall, support rates dropped by 5.5 and 9.4 percentage points between the first and the last measured point in time for economic and political integration, respectively.
Third, in every country, support rates were on average higher for economic integration than for political integration. The overall mean difference is 9.4 percentage points. Over time, there are more signs for a divergence than for a convergence between the two trends: While support for political integration was in many countries on a par with support for economic integration in 2002, values tended to be much lower (relative to economic integration) in later years.
Determinants of Support for Integration
Table 2 contains four multilevel logistic regression models. The first model shows how micro-level variables influence support for economic integration. It reveals that all intrasocietal cleavages under study are highly significant and go in the expected directions (where specified): support for economic integration is more prevalent among younger people, men, the better educated,Footnote 7 and the upper occupational classes, as well as those with a better personal economic situation. The latter two effects suggest that the consumption-based rather than the job-based mechanism is at work. People who are satisfied with democracy and who see their country in a better economic situation are also more supportive of integration. Compared to the political center, people on both the left and the right appear to be less supportive of economic integration, but this relation changes between 1997 and 2010 as we will see below. In line with our descriptive findings (Figure 1), support for economic integration decreases over time.
Multilevel mixed-effects logistic regression.
Note: Standard errors in parentheses. Likelihood ratio tests of all models show Prob > χ² < .001. *p < .05, **p < .01, ***p < .001.
In the next model, macro-level variables are added, which discloses some unexpected effects: first, when trade as a share of the GDP grows within countries over time, support for economic integration decreases—a counterintuitive result that conflicts with transactionalist theory. A potential explanation is that prosperous nations that are well-integrated into the global market are reluctant to integrate with smaller countries with weaker economies (cf. the 2015 debate on whether staggering Greece should be forced to exit the Eurozone). Second, when FDIs as a share of the GDP rise within countries over time, support for economic integration increases. This finding is also unforeseen and might indicate that there are other mechanisms at work over and above the proposed one (FDIs as an indicator of hegemonic foreign influence). For instance, FDIs may preferably go to countries with well-performing, stable, and predictable political and legal systems (Reference MonteroMontero 2008), in which support for integration is also likely higher. Unfortunately, the limited degrees of freedom at the macro level don’t allow testing of such alternative explanations in detail. Differences between countries in trade and FDIs as shares of the GDP have no significant effect. Membership in Mercosur and ALBA is not significant either, but coefficients are positive. As these macro-level variables are added, all individual-level effects remain very stable.
The micro-level effects for support of political integration (third model) are very similar to the ones observed for support for economic integration.Footnote 8 The only exceptions are age and the difference between the political center and the right, which are no longer significant, as well as the political left, which is now significantly more (not less!) supportive of integration than the center (but again, see the trend discussion below). Once more, support rates decrease significantly over time. Concerning the macro-level variables (fourth model), increasing trade as a share of the GDP again has a negative impact on support for political integration. FDIs, however, have no significant impact on public opinion on political integration. Contrary to our expectation, citizens of Mercosur (not ALBA) are more likely to be in favor of political integration.
How Social Cleavages in Support for Integration Change over Time
To examine whether and how gender, education, and political orientation gaps in support have changed over time, we first conduct six additional multilevel logistic regression models which each expand the full models reported in Table 2 by adding interactions between year and one of the three variables in question (Table 3). Diverging patterns become apparent: the interaction between gender and year is nonsignificant in both the model predicting support for economic integration and the one predicting support for political integration, suggesting that the gender gap is stable over time. The interaction between education and year is also nonsignificant in the former case (economic integration) but significant in the latter (political integration). Finally, the interaction between political orientation and year is significant in both cases—at least for the center-left gap.
Multilevel mixed-effects logistic regression with interaction effects.
Note: Standard errors in parentheses. All models contain all variables from the full models of Table 2 (omitted due to spatial restrictions). *p < .05, **p < .01, ***p < .001.
To get a better understanding of what these effects mean, we create marginal effect plots based on the six models in Table 3 (Figure 2). In accordance with our previous observations, all lines point downward, indicating trends toward less support for integration. The gender gap is stable over time with regard to support for both economic (Figure 2A) and political integration (Figure 2B). The gap between educational levels (Figure 2C–D), which is particularly large, is also quite robust. Yet, while the education gap in support for economic integration even seems to widen slightly (though nonsignificant), the educational gap in support for political integration becomes somewhat narrower over time (significant). However, this convergence is negative and due to steeper drops in support rates among the better educated rather than a catch-up of the less educated. Despite this minor shift, the demographic divisions by gender and education appear very stable overall.
Marginal effects for gender, education, and political orientation.
Note: Plots are based on the corresponding multilevel models in Table 3. Marginal effects were calculated using representative values of subpopulations.
The same cannot be said for the effect of political orientation, where considerable shifts occurred (Figure 2E–F). Whereas the political left was less supportive of economic integration than the right (P > |z| = .000) in 1997, it was significantly more likely than the right (P > |z| = .022) to support economic integration in 2010. The gap between the center and the left declined sharply; yet the center, which was clearly more likely to support economic integration in 1997 (P > |z| = .000) remained slightly more supportive than the left in 2010 (P > |z| = .016).Footnote 9 The left is also the only social group in Figure 2 whose support rate for economic integration did not decrease significantly between 1997 and 2010 (P > |z| = .318). Furthermore, the left was significantly less likely than the center (P > |z| = .000) and as likely as the right (P > |z| = .684) to support political integration in 2002, but became clearly more likely than both the center (P > |z| = .000) and the right (P > |z| = .000) to do so in 2010. In other words, during the time frame under study, a significant political shift occurred as the left turned from stumbling block to stepping stone of regional integration in Latin America.
Summary and Discussion
This article has examined the development and social structure of public opinion on Latin American economic and political integration in seventeen South and Central American countries between 1997 and 2010. In summary, levels of public support for economic and political integration are generally high in Latin America, but dropped over time. Support for economic integration tends to be higher than support for political integration. Furthermore, both within and between societies, there are considerable cleavages in support for integration. Regionalism is seen more favorably, inter alia, by men, the young, the better educated, the upper occupational classes, and those satisfied with democracy and the economic situation of their country. Finally, while gender and educational cleavages in public support remained relatively stable over time, a significant shift occurred with regard to political orientation, as the left transformed itself from least to most supportive wing of the political spectrum.
These insights have several implications. For one thing, our main finding—of a shift toward the political left as the major supporter of integration—is remarkable as it contrasts with the general portrayal of the left as uncomfortable with the idea of regional integration, prone to protest against free trade agreements from NAFTA to TTIP, and critical toward institutions such as the European Union, which it is said to perceive as undemocratic and illegitimate. Such views were substantiated in the past by empirical evidence from Europe and other regions, including the Latin America of the 1990s. Our findings, however, show how quickly this supposedly universal pattern can reverse: when regionalism is increasingly driven by left-wing governments and takes corresponding forms of appearance (counterhegemonic, socialist), public support for regional integration may swing from the right and the center to the left. Thus, regionalism may take different forms and is not necessarily affiliated with a particular political force. Future research could build on this finding and analyze more closely which specific mechanisms led respondents with leftist orientations to develop greater support for regional integration. (For example, what role did parties, leaders, the media, or civil society play in shaping these trends? How do different degrees of politicization come into play?) Moreover, the recent ascent of the nationalist right in Europe and the United States raises the question of whether a reverse “swing to the right” in support for integration is also possible, for instance if the appearance of integration were to change from neoliberal to nationalist—instead of socialist (cf. Reference Minkus, Deutschmann and DelheyMinkus et al. 2018).
Related to this point, our findings may be of interest to the growing field of comparative regionalism (e.g., Reference LaursenLaursen 2010, Reference DeutschmannDeutschmann 2015). For us, drawing on studies from the European context proved useful for developing ideas about which variables to consider in the far less researched Latin American case. While some patterns appear to follow the same logic in both regions (e.g., the gender gap), others seem to be idiosyncratic and related to specific characteristics of the Latin American context (e.g., the swing to the left). To arrive at an even better understanding of the universality or historical uniqueness of specific structural features of support for regionalism, we hope that comparable polls from other world regions will become available in the future (e.g., via a standardization of questions throughout the Global Barometer survey family). Additionally, it may be useful to incorporate people’s knowledge and expectations about integration (cf. Reference Díez Medrano and BraunDíez Medrano and Braun 2012), or to analyze how public opinion on regional integration differs from general support for free trade and global integration, that is, how closed toward the outside world Latin American regionalism is. Similar questions have recently been addressed in the European context (Reference Delhey, Deutschmann, Graf and RichterDelhey et al. 2014). Furthermore, as new waves of the Latinobarometer become available, the time period covered will expand (provided that the relevant questions remain in the survey), increasing the disposable number of observations and thus options for further macro-level analyses.
Finally, the overall drops in support rates indicate that Latin American regionalism could soon be facing problems similar to those experienced in Europe: the further European integration proceeded and the more competencies were transferred to the European Parliament over the years, the more did interest among the general public and voter turnout decrease (Reference DabèneDabène 2009, 153; European Parliament 2014). While integration was still supported by the majority of the population in all Latin American countries in 2010, the downward development should be monitored closely in the future to diagnose further drops early on. Policy makers interested in increasing public support for Latin American integration could use our findings purposefully to address those societal groups identified as least consentient. Particularly high potentials for gains can be found among the least educated and the lowest occupational classes (i.e., farmers and individuals working in the informal sector). These are the people whose perceptions and needs need to be addressed if public support for regional integration is to remain high in Latin America. As Riggirozzi (Reference Riggirozzi, Hurrelmann and Schneider2015, 230) has warned, “Social mobilization is likely to arise when region building is perceived as an ‘elite compromise’, with regional policy and politics unduly favouring privileged parts of the citizenry.” Of course UNASUR and ALBA are, as we have seen, examples of a post-neoliberal regionalism that does in fact attempt to address the disadvantaged parts of society. However, the stability of the gender and educational gaps in public support for integration indicates that fostering support rates and diminishing societal cleavages is anything but an easy task. Simple and intuitively reasonable mechanisms (e.g., rising gender equality leads to harmonization of gender-specific support rates) do not readily lead to quick, measurable change. Apparently they may be delayed, disrupted, or even disabled by ingrained systemic stability or more complex causal chains. And yet, if Latin American regionalism is to be successful in the long run, we have to ensure that the lower social strata are not left behind in this process. The more concrete and effectively realized economic and political integration becomes in Latin America, the more challenging it will be to keep support rates high in all segments of society.
Additional File
The additional file for this article can be found as follows:
Table A1. Latinobarometer questions on support for economic and political integration. DOI: https://doi.org/10.25222/larr.250.s1
Acknowledgements
We would like to thank Colin Vance, Adalbert F. X. Wilhelm, and three anonymous reviewers at LARR for excellent advice. Previous versions of this paper were presented at the 112th Annual Meeting of the American Sociological Association in Montreal and the Interdisciplinary Workshop on Opinion Dynamics and Collective Decision at Jacobs University Bremen. We are grateful to the participants for their invaluable feedback.
Open access
