. Left-Right Identifications and the Latin American Voter ELIZABETH J. ZECHMEISTER Navigating complex landscapes is difficult without signposts. In politics, the terms “left” and “right” frequently function as markers that guide indi- siduals through political and ideological terrain. Rather than communicate aprofuse number of policy stances to the partially tuned in average citizen, 1 politician can efficiently convey her or his stance in the political space by chiming to be “on the left” or “to the right” and the citizen can in turn se- lect the candidate closest to his or her own left-right stance. Thus, in their ideal form ideological labels facilitate issue-based vote choice. Yet, they can only do so to the extent that they denote policy content. In practice, the terms often act simply as rough proxies for party labels or, worse, are conceptually nebulous symbols that may or may not be connected to voter choice." _ Inthis chapter I document that on average Latin Americans’ left-right identifications correspond to their vote choices. However, this blanket statement masks important variation across countries: first, in the ten- dency for individuals to use left-right self-placements; second, in the ex- me Which the left-right semantics link to policy stances; and, third, in hh es of the connection between left-right placements and the vote. the terns 4ggreet Will demonstrate that, despite their value in theory, American a and “right” are eschewed by many members of the Latin comparative iH such thatnonresponse to left-right placement questionsis Frobuse idea gh in several countries. Moreover, there is scant evidence Sample o'ogical significance to left-right labels in Latin America. For ic Staneeg en Considering the classic economic dimension, economic pol- “gnificantly predict left-right identifications among the mass 10s Escaneado con CamScanner196 + 7 4 E LATIN AMERICAN y, OTER Public in little Fi j, More than h: ‘inally, in many countries any i the ighteen € regio, See For example, ezuela, the mass public i for self lace eka, o and willing individuals connect ther etergie right : pie structure vote dectiona ana to thi choic, hough Chileans are among the least willing ng s Olie right position. In other cases, such as Bran ie \e region to rf i. a 7 al f. oe of policy significance or vote relevance to eat ita The core of this chapter evaluates and sels ie variation in Latin America inthe exten o which the nas pa lefe-right identifications, connects these to policy santa, aff ad in the contempo: Is reflect ‘The first two of these considerations—rates of placement on the lef scale and policy significance—have received comparatively mete in previous research (see Zechmeister and Comal 01). Ths i dressing all three aspects of left-right identifications, this caper pl nore emphasis on examining the connection between leigh tions and the vote, and the contextual factors that shape this re In so doing, I present and apply a theoretical ee. a the importance of features 0! competition to the vote relevance Escaneado con CamScannerLeft-Right Identifications and the Latin American Voter * 197 palton 201 x), and yet the chapter also considers selec pat ‘cular to the Latin American context. One of these ver of the political left at the turn of the twenty-fi pone debated the extent to which this “pink tide” re conversion among the mass public or merely a simple Tetrospective calcu- Jus (under which poorly performing incumbent right parties were voted |p In pooled cross-sectional and case-specific temporal analyses, I find |) fe to no evidence that the pink tide left an imprint on the | fcance of the terms “le” and “right” in the region. (On the whole, the results in this chapter affirm that left-right identities in Latin America are often quite anemic and, yet, some substantive content ‘and vote relevance can be infused into the terms conditional on the nature of party competition. The tendency for that to happen does not hinge on the incumbent’ ideological stripes. Rather, through their general behay- jors and rhetoric in the process of competing for office, political elites can \ foster information environments that facilitate citizens’ roles as voters. . ‘The extent to which features of party competition increase the political significance of the left-right semantics varies significantly across countries : in Latin America, and thus, while this chapter provides mostly a snap- © shot of the current period, the patterns detected should lead us to expect change over time. The nature of that change will hinge significantly on how party competition evolves in Latin America and, specifically, whether party systems continue to fragment and prioritize vote-buying and other non-programmatic means of attracting votes or, conversely, stabilize and increase in the extent to which parties distinguish themselves on the basis of relevant policy stances. ted factors that are is the ascendance to irst century. Scholars flected an ideological political sig- § Theoretical Perspectives on the Left-Right Semantics and Politics Leftright labels are highly valued by scholars of democratic polities be- | Gus, in theory, they create “economies” and “efficiencies” (Converse | 1964) for political communication and competition Fuchs and Kline” {mann 19905 Thomassen and Schmitt 1997; Hinich and Munger Tet (This important, because most citizens have ltee interest in absOn 0 {and understanding the complexities of moder politics est Comey. FF | al-1960;Delli Carpini and Keeter 1996; Kinder and Sears 1985) P8080 i tifcations carve out easy-to-discern political divisions: 2 : are | Provide an effective tool for decision-making: Ye cough soon i { iquitous in polities, few political concepts are mor de perc [ ' Left-right labels and identifications are frequent'y toa | Fan underlying ideological structure. Ideology, in this sense shat Escaneado con CamScanner198 «THE LATIN AMERICAN VOTER hierarchical and interrelated structuring of beliefs about yt, “capped by concepts of «high order of absteaction® Cant — 193). In other words, left-right labels can be “eymbolier al aoe dnderlying “operational” ideology composed of bundles onion Federico, and Napier 2009). From this perspective, the ;, 7 ‘chen individuals make use of left-right matkers yo pee sey onthe lef-right scale, when those left-right placements enne underlying set of issue positions, and when they connect theqe nt ® tions to their vote choice. ‘deni However, itis well accepted that the mass public ypc fy of this ideal standard. Conventional wisdom offered by scholan et politics states that levels of ideological rhetoric and structuration a 3 the mass public are quite low (Kinder and Sears 1985; see aso Campa et al. 1960; Converse 1964). Campbell etal. (1960) and Converse (1 note that there could be individual logics that account for some of disconnect they find between stances that the sophisticated observer would. expect to be correlated (see also Ellis and Stimson 2011). Nonetheley, most individuals’ beliefs are “jumbled,” many individuals simply do notuy. ideological markers when discussing politics, and, when they do, there often a disconnect between their left-right (or liberal-conservative) ment and the specific issue positions that they hold (Converse 1964, Jost, Federico, and Napier 2009). : Given that most individuals have limited time and interest toi in politics, it makes sense that use of ideological markers can be | than ideal, with respect to both placement and vote decisions, and thatt markers lack robust substantive (policy) significance among the mass lic? In fact, an uncomfortable fact for those who would “left” and “right” anchor issue divides is that ideological relatively weak in policy content and are often more meani tives to party names (Butler and Stokes 1969; Inglehart and 1976; see also, among others, Dalton, Farrell, and McAllister Heath, and Lalljee 1996; Fleury and Lewis-Beck 1993; Knutsen 1 coby 2002; Zechmeister 2006b). 7” ‘Yet, the existence of some degree of substantively meaningful lefts semantics in the political system indicates that effective ideological mar 1S are available for citizens who require assistance in finding elites —_ oo stances roughly approximate their own. And, so, rather than a 1 = Benerally low levels of ideological struccuration in the mass pt! a Sain more from applying a comparative lens though which : iti ig ‘ ross mass ae levels of political significance of lefe-right terms #" Escaneado con CamScannerLeft-Right Identifications and the Latin American Voter + 199 stisindeed a fat that the political significance ofthe left-right seman- "aries across counties, across ime and even across subgroups ofa spon (AONE MARY SEE, Cy Aspelund, Lindeman, and Verkasalo | PoP Campbell etal 1960; Dalton, Farrell, and Mealister 201; de Vries | 2008 erdia, and Lancee 2013: Inglehart and Klingemann 1976; Kitschel ve Hellemans 1990; Nathan and Shi x996; Evans and Whitefield 1998; Dalton 2008; ‘Weber and Federico 2013). But according to what factors? Classic research has focused most attention on three factors: consistency ‘ver time, polarization, and fragmentation. in the first place, because consistency over time is critical to the estab- ishment of a well-developed ideological structure (Hinich and Munger 1996), scholars have proposed that the comparative youth of democracies jp asia and Latin America makes for less robust ideological semantics (e.g. Jou 2011; ee also Gonzélez and Queirolo 2009). Compared to Western Europe, for example, modern Latin American politics have been character ized by much less stability (Roberts and Wibbels 1999), which should work against the development of high levels of ideological structuration (Zech- meister and Corral 2013). Thus, in general, we should expect compara- tively low levels of left-right use and policy significance in Latin America. In addition, basic features of the party system matter to the degree that they clarify political alternatives. Polarization can make the end points of the political spectrum easier to discern, and may increase the use of ideological labels by elites, leading to more effective use of left-right identifications by the mass public (Ames and Smith 2or0; Dalton 2008, 2011; Inglehart and Klingemann 1976; Huber 1989; Singer forthcoming; Zechmeister and Corral 2013). In fact, the relevance of polarization to the political significance of left-right identifications is one of the most well- established findings in research on this topic. Fragmentation, or the number of parties in the system, may also affect ideological structuration, though this effect differs across cases. A posi- tive relationship between the effective number of parties and substantively meaningful left-right ideological labels is found in research on Europe, so ai thesght fai Poa 1976; Huber 1989; Knutsen 1997 wer, multi-country analyses by Dalton (2008, 2011) find no relationship between fragmentation and the extent to which left-right stances are correlated with vote choice. In Latin America, Zechmeister | | | aot and Corral (2013) find that the effective number of parties (and, as well, electoral volatility) is negatively correlated with individuals’ ability or will- ingness to place on the left-right scale. The authors argue that this re- lationship makes sense, given that party system fragmentation is highly Escaneado con CamScanner200 + THE LATIN AMERICAN VOTER related to party system flux in the Latin American pep: coming) applies this logic to the question of the ee TBiOM. Singer right identifications and voter choice in Latin America no” Peg negative relationship. TCA and fines Other factors can also influence the nature and «i, fen identifications. At the country level, for example, Dalton (ont ety lights the relevance of economic and political develop (2006, sory 2008, and Freire and Kivistik 2013). Building and pili (see aka ge research, I draw attention to two additional factors, The fren Poa rectly out of the micro-logic undergirding studies of the rena ® teen features of the party system and left-right identification: pee O& is that how elites bundle political ideas and package them sat ad “ei labels influences the degree of ideological structuration in the m (owns 1957; Hinich and Munger 1996). If this is the case, then aa stance of party competition should matter for ideological structraton eg the significance of the left-right semantics. In a study of ideological among citizens in Mexico and Argentina, Zechmeister (2006) ati the more robust and politically relevant significance of the left-right sem tics in Mexico at the time to Mexican elites’ comparatively greater tender to use the terms in political manifestos and speeches.’ In a study of left-right attitudes, Harbers, de Vries, and Steenbergen (2013) evidence that the level of programmatic party structuration (see Kitscheleet al, 2010) in Latin America matters, such that more variation in responsesis found in the least structured system, Ecuador, and less variation is found the most structured system, Chile, with Mexico falling in between. The second relates to tendencies to place oneself on the left-right sc Whether an individual self-identifies on the left-right scale can be du non-rival factors that are cognitive (recognition of tional (willingness to engage). A focus on the forme found in scholarship on the left-right, and thus scl look to individual predictors such as education and/or P tion (see Dalton 2011; Zechmeister and Corral 2013). have received comparatively less attention (but see Scholz In the Latin American region, where dissatisfaction with quo has manifested itself in low levels of expressed part more active demonstrations of discontent, we should exps or willingness, to report a left-right identification, t be portant. Therefore, this chapter includes an assessment of of political interest and perceived system efficacy inp the left-right question in Latin America. Escaneado con CamScannerLeft-Right Identifications and the Latin American Voter + 201 els of Left-Right Identification varying Lev | shows the distribution of left-right placements for eighteen Figate oe re Countries in response to the 2012 AmericasBarometer atin Amer atin American Public Opinion Project (LAPOP). The oy ao created by taking responses to the left-right self-placement ovis ar ro a 1-10 scale, and coding “lefe” those who respond 1, questions ee those who respond 4, 5, 6, or 7; and, “right” those who 1083S or 102 Those who indicate they do nt know or othervise nd 5 BO re question are noted as “NR” (this nonresponse option donot a geno ea aloud) The countries are arrayed inorder bythe was ynresponses. pues fnonttieg Nicaragua, Uruguay, the Dominican Republic, eee, Guatemal, Panam, and Brazil—more thin one ou of very vent pividalsslf-identifies on the let” In nine countries—the Domini-__ i y Mexico, viblie, El Salvador, Panama, Colombia, Paraguay, ' aid ie Venezuela, and Brazil—more than one out of every five peo car fs on the right. As others have noted, the public in Latin Ame teams toward the right (Arnold and Samuels 20115 Dodek 2011; Seligson soo7) The AmericasBarometer 2072 reflects this tendency, but only to a very slight degree.° om tom momo Aa OH Figure 8.1 Distribution of Left-Right Placements. Source: Americas: Barometer 2012 by LAPOP (vqg/v50)- Escaneado con CamScanner202 + THE LATIN AMERICAN voter Also noteworthy in figure 8.1 are i the left-right identification cused ae levels o average nonresponse is 18.6%.” As a polhcet ne, finds mean nonresponse rates between 12.1 Contrast, Mair ¢, rope and berween 16.5 and 22.8% in Eastern Beno, Wena are considerably lower for selected Western Eure" France, where they range from 2 to 10% (Dalton, Rane suche zor; Lewis-Beck, Nadeau, and Bélanger 2012) In vee Medligg, response to the left-right question in the Latin Ameri Et those found in Eastern Europe.’ rican region min Regional nonresponse rates mask the tremendous variati countries and, in some cases, over time (see Dogek 2011), Asper ge om seven countries have levels of response to the left-right seal ore viste, of are even lower in same cates then, the average nonreaga in Western Europe: Uruguay (0-3%), Panama (06%), El Salar gt event (32.1%), the Dominican Repoblie (123%), Nicaragua sai and Mexieo (13-59) Ina nomber of countries, however, anemone far exceed the average even in Eastern Europe: specifically, Costa Rican Paraguay (38.1% and 33.1%, respectively).” : Previous research on Latin America has found party system po at tion to be positively related to placement on the left-right sa reverse for fragmentation and electoral volatility Zechmester and tor). Yet, failure to self-place on the ideological scale may be cther factors as well. One such set of factors particularly reev cemporary Latin American politics contains those related 10 po satisfaction and disengagement. hhas been ripe with signs of political discontent actions (e-g., demonstrations) and dissociation (€» Ship) (ora classic framework: on voice and exit responses st 1 1970). Many have placed the blame on failures to engast: include, amt spond to the mass public (Se, xg, Agiero and Stark 1998 Hae Mainwaring 2006; Mainwaring, Bejarano, and Leongémez. 2006) perceived system fllures might be registered by individu a simply turn some citizens off from politics. ‘As such, I consider a ence of perceptions of system responsiveness and politic inverse ingness to place oneselfon the left-right scale iat based | Figure 8.2 presents results from a logistic regression m0 i | a pooled dataset for the region. The data are from the 2077 | Barometer. The dichotomous dependent variable, | is modeled as a function of basic controls (education, f non tion oa gender, # Escaneado con CamScannerLeft-Right Identifications and the Latin. ‘American Voter + 203 re 307 est F= 56.301 Ne aiese ey esta ers acto (00) mae Woah atest port pe ‘age Cohort mh Female bet += __ 150 =t00 -0150 0.00 oo 1} 98% Confidence interval (Design-Etfect Based) Figure 8.2 Individual Level Predictors of Non-Response to the Left- Right Question. Source: AmericasBarometer 2012 by LAPOP (vs6). Note: Analysis based on logistic regression with all independent variables rescaled from 0 to r. See text for discussion of substantive significance of these results. County fixed effects included but not shown. rural residence, wealth,!” and country fixed effects) and two measures re- lated to motivation: external efficacy and political interest. The external efficacy measure is based on a statement asking respondents the extent to which they agree or disagree that “Those who govern this country are in- terested in what people like you think.” Political interest is measured with 8 question asking “How much interest do you have in politics?” Responses are coded so that higher values indicate greater efficacy and interest on these 7-point and 4-point scales, respectively." The analysis reveals that, With the exception of the age variable, the independent variables are sig- nificant at p < 0.05, two-tailed.” ‘Women and those in rural areas are less ly to respond to ight question; those who are wealthier and ~ edu ly to respond. mote educated are mc __ As hypothesized, perceptions of system responsiveness and political interest predict left-right response. Individuals who report higher levels of extemal efficacy and higher political interest are more likely to place themselves on the left-right scale. The three variables with the largest Escaneado con CamScanner204 ¢ THE LATIN AMERICAN VOTER substantive effects are education, political inte - ; re Logistic regression results are not directly — omerest, and interpretable, bias ‘ the results of the analysis to cal : ares f culate a set of predi rere, 1 f) left-right nonresponse based on moving from the maces Pibiiig® a) mum value on each of the independent variables im t0 the, H s in turn, whi i] + while other variables constant at their means. I ; nin to the highest value on education, pole ne tom I increases the likelihood of placing oneself on the leery el and 7 percentage points, respectively." ight scale by This result is important because it affirms that high : to the left-right ideological question do not merely signal ete tion, Rather, for meny Latin Americans fire ty provides the Iefecight scale ia an act of political disassociation. Whil dn of general relevance to scholarship on left-right identifications in explaining the very high levels of nonresponse tothe lefcright in counties such as Costa Rica and Chile (se figure 8.1), both of have comparatively extensive democratic histories Levels of confdene and satisfaction with Costa Rican institutions and democracy have deh in recent years (Alfaro-Redondo and Seligson 2012), and discontent democratic ambivalence have also been brewing in Chile (Siavelis 0 Carlin 2011). In the 2012 AmericasBarometer dataset mean levels of ex temal effcacy in Costa Rica are lower than in any other county ind region, Chile ranks in the bottom four on politcal interest long Conta Rica) and in the lower half of countries on external efficacy. In costa Rica and Chile and for others acros the mary, for many in C lack of response to the left-right sale is fueled by a mativaton 1 dimers the political system. oe Weak Correspondence between Issues and Left-Right Identificat left-right semantics a8 tant to consider the & at is, in ¥ Many members of the public do reference | igate political space. It therefore is impor which left-right semantics reflect policy divides. Thi i tries is there substantively rich policy-relevant content ©0T mantis and where instead are these terms ideologically ne To the extent that ideological labels reflect programmatic we should be able to detect greater corresponde right stances within the mass public. This corresP public will tend to only weakly echo back ideological Pie elites. This is as it should be: possession of a highly detail kages Escaneado con CamScannerLeft-Right Identifications and the Latin American Voter + 205 petit semantics would undermine their utility as short-cuts that re- Z conomies” for political communication and decision-making. ate a ine the strength of the connection. between. policy stances and. ‘nt semantics, with.a focus.on five different types of predictors. cof these, the economic dimension, captures the classic meaning of sight terms (Downs 1957; Bobbio 1996) and the principal dimen- fe pa ele fail i of party competition 1n Latin America (Kitschelt et al. 2010; see also Moreno 19993. Wieschomeier and Doyle 2012; Baker and Greene this vol- ume). The ‘Gcond)the democracy (versus authoritarianism) dimension, has vin significance in Latin American politics overtime and across coun vate (Moreno 1999) and is kept relevant by debates and incidents in some srontres related to the centralization of power in the executive branch at the expense of checks and balances, rewictons op media, and so on. sede ictox, asks about support for tough Karices against criminals, in Grder to ta into the increasing relevance of issues of crime and violence in the region, fom) moral issues have been historically important in Latin America, given the history of church-state relations, and specific issues re- lated to gay marriage and abortion rights recently have been debated by politicians, in the media, and by the public (Boidi and Corral 2013; Lodola and Corral 2010). Finally, a fifth dimension—attitudes toward the United States—is included becausé*of fts theorized relevance to defining the new left in Latin America (Arnold and Samuels 2011; Remmer 2012)."* ‘Table 8.1 is based on a series of regression analyses, one per country, in which left-right self-placement (1-10, with higher values indicating right- lleaning tendencies) is predicted by individuals’ stances on seven policy asures: Government Ownership of Key Industries; Welfare Factor; Belief De- oracy is the Best; Belief Democracy is Preferred; Tough on Crime; Religion is Important; and Trust in the U.S. The fall table of results, with question ‘wordings for these variables, is available in chapter appendix table A8.r. ‘Table 8.x reports the r-squared statistic, which indicates how much varia~ tion in the dependent variable (left-right placement) the issue measures plain in that country, and lists which independent variables (issues) are (fiiican predictors, ifany. As the table shows, chefe is substantial varia- gre counties in the policy significance ofthe left-right semantics, a finding consistent with prior work by Zechmeister (2006b) and Zechmeis- /ter and Corral (2011, 2013). a Peinomic attitudes predict left-right placement in only ten of the a cases, and the signs on those measures are negative as expected lef at preferences for a greater role of the state are associated with leaning identifications), except in the case of Honduras. Interestingly, Escaneado con CamScanner206 + THE LATIN AMERICAN VoTeR TABLE 8.1 with Issues Country Honduras 0.16 Gov't Ownership (4), We Democracy is Prefer 0.15 Gov't Ownership Tony ie ‘Trost in the US.) Venezuela 0.14 Gov't Or i Chile 0.14 Gev' Uruguay S$.) Tine tg Argent Lan O12 ; = Religions Inpoe, anama 1 (), De i Bi Boli 0106 Trustinthe US.) 7 tO Trstinthe tis. g q Mexico 0.06 Welfare (), Democracy is Best (), Demoe Costa Rica 0.06 Gov't Ownership () *, itmocracy is Prefer Domi ° Cae eR DameeTE Bes 6) Ripon eo aie 008 Democracy is Best), Tough on Crime) Nicaragua 0.05 Trust in the U.S. () Paraguay 0.04 Welfare (), Trust in the US. () Colombia 0.03 Religion is Important (2) Guatemala 0.03 Religion is Important () Pern 0.03 Welfare (), De is cies a 103 Welfare (9), Democracy i Best) Religions porn) ElSalvador 0.02 None Ecuador 0.02: None ‘Source: AmericasBarometer 2012 by LAPOP (49/50) ‘Note: Table summarizes OLS analyses (adjusted for complex sample design) R-squared from models that predict left-right placement (original linear “Ix” measure) wit seen in variables: Gov't Ownership of Key Industries; Welfare Factor; Belief Democracy isthe Bs Preferred; Tough on Crime; Religion i Important; and Trust in the US, tai using a generous p < 0.10, two-tailed, cut-off. Full regression analysis results and ques available in the chapter appendix table A8.1. in Honduras, those with a greater belief that the government key industries and those who prefer a stronger welfare $3 the right.” Attitudes toward the United States are s and in each case the coefficient is positively ted States is associated with p nine cases that greater trust in the Unit right of the left-right scale. predictors in six of the cases. The as a significant predictor in seven ¢ coefficients’ signs are positive (greater with right-leaning identifications). Finally, significant in just three cases and positive 19 ith right-leaning jdentifications)- belief that r ases and, with the exe importance to Fe ime Ily, the tough on Cm aly, se cases (rougter are associated wi Escaneado con CamScannerLeft-Right Identifications and the Latin American Voter * 207 as expected, the issue-based content of the left-right semantics varies ‘i fet) across countries. ‘We can infer from the analysis that scholars of the regio? who give in to the temptation to use left-right identifications as proxies for economic divides would only be correct roughly half the ime. Further, based on the models assessed here, in a few cases (Brazil, El ttfxdr and Ecuador) there is no discernible issue basis to the left-right semantics at all. Insum, Contextual Factors and the Relevance ofLeft-Right Identities to Vote Choice ‘though ome individuals select not to place themselves on the left-right dimension and though left-right semantics may be only weakly associated ai policy divides, there is nonetheless a general tendency for left-right identifications to predict voter choice in Latin America. But to what ex- tentis this relationship present in all countries and what predicts cross-na- fora variation in the connection between left-right placements and voter Choice? This isa worthwhile exercise, given doubts regarding the extent to ‘which the Latin American public conceives of most parties as sitting clearly oneither the left or the right of the ideological scale and whether, in fact, tnembers of the mass public actually connect left-right placements to their vote choice (see discussion in Gonziilez and Queirolo 2009)..The extent to which eftright identifications piedict the vote gives us one measure of the political relevance of these.terms in the Latin American region. ‘As a first turn at assessing the strength of a connection between left- right placements and individuals’ vote decisions across countries in the Latin American region, I ran a series of basic OLS regression, analyses. The dependent Je, vote choice, is a measure that afrays the principal _ candidates in residential election from left to right along a 20- point scale; the measure is described in detail in the part IT introduction, The key independent variable is a measure of left-right placement on a rorpoint scale (linearly transformed to a 0 to 1 variable. The models also incorporate the individual-level variables included in earlier analysis (Gee figure 8.2)." The core results are presented in table 8.2, with the countries Presented in alphabetical order. In thirteen countries—Bolivia, Chile, Colombia, the Dominican Re- public, Guatemala, Honduras, Mexico, Nicaragua, Paraguay, Peru, El Salvador, Uruguay, and Venezuela—left-right self-plagement is a signif ve predictor of the left-right vote(In five countries>Argentina, and Panama—the left-right self-placement Escaneado con CamScannerE [ 4 i ‘Scare: ArmericasBaromcer 2013 by LAPOP age New: So10, two-tailed Ourpet s cocfSceass and sanderd| 2 OLS regressing snaivses for each couacry and. = the Seal rom indlode controls for age, gender, edocatice, west. coral GESRES Escaneado con CamScannerLeft-Right Identifications and the Latin American Voter. + 209 of the vote."* Not only is there varia- fi eftright sel placement as a predic- if the vores but there is also variation in the substantive relationship tof he Jeft-right identifications a voter a across countries. For + Chile, moving across the range of the ic i rt dant amples ae (rom fa lft to far sigh resuka i agaktt ofa ne fein Me choice dependent variable, which ranges from 1.85 as ta ‘ t eountry. Conversely, ee same maximum effect is only 1.2 anitson the dependent a ae ich ranges from 5.36 to 15.72 (for more details, se¢ chapter appendix table 8.2). On average, left-right placement jg related to voter choice in Latin America (see the last row of data). Yet, iferight self-placements are minimally or not at all connected to voter choice in many Latin American countries. What predicts cross-national variation in the connection between left- sightplacements and voter choice? Based on earlier discussion and extend- ing out of extant scholarship, I focus here on four core factors. Pélarizatic taprresthe extent to which parties take distinct stances and should earify snd thus increase the politcal significance of left-right terms. The Effeo toéNumber of Parties, in the Latin American case, provides a proxy for ggstem flux Zechmeister and Corral 2013); greater fragmentation should be asociated with less clarity and less political relevance to the left-right semantics. Programmatieness ‘should increase the political significance of left-right identifications, while higher levels of Clientelism (by diminishing the substantive nature of party competition) should do the reverse.” To assess the relevance of these contextual factors for voters in Latin America, I ran a series of multilevel regression analyses with cross-level_ interactions. In each case, I predict left-right vote choice with the same | individual-level variables reported in figure 8.2. The contextual variables | are entered as second-level variables; given degrees of freedom issues as- | sociated with a dataset containing only eighteen countries, T assess the | second-level variables separately, in a series of four analyses.” The results \ reveal that Polarization anid Programmuaticness. are associated with a stronger connection between left-right and.the vote, while Fragmentation and Clien=— telism are associated with a weaker connection, In short, the nature of both \¢ party system and elite competition matters. ' 8 Based on the regression analyses, figure 8.3 depicts the significant me ig, effect of each of the contextual variables on the extent to whic Tight identifications predict the vote choice." In figure 8.3A, we see ra there is no significant effect of left-right identities on VOR choice at ty low levels of polarization; as countries become more polarized, the ¢ Escaneado con CamScanner3A. Effect of 5 Predicted Effect on Voter Choice Oo AB 2 25 3 35 4 45 5 & Level of Polarization, DALP (Range: 0.103 to 0.599) 3B. Effect of L-R Ideological Identification, by Effective Number of Parties 6 4 2 O Predicted Effect on Voter Choice 2 8 7 2.01 0 104) 4 6 4 5 : ° umber of Parties (Rand Effective Ni Escaneado con CamScanner30. Effect of L-R Ideological Identification, by Programmatic Index Predicted Effect on Voter Choice 7 om 4 06 oe 1 42 4 16 18 2 22 Programmatic Index, DALP (Range: 0.04 to 0.26) 24 26 430. Effect of L-P Ideological Identification, by Level of Clientelism 2 2 0 [ [ of " 12 13 14 15 16 7 18 Clientelism, DALP (Range: 11.6 to 17.9) Figure 8.3A-D Predicted Associations between Left-right Identifica- tion and Voter Choice. Source: AmericasBarometer 2012 by LAPOP (wso) and DALP. Note: Figures based on four separate hierarchical regression analyses using individual level data from the 2012 ‘AmericasBarometer and contextual party system data; graphs show predicted average marginal effects for left-right identifications on voter choice at varying levels on the contextual (second-level) variables. See chapter appendix Table A8.2 for Tegression output. Escaneado con CamScannerTHE LATIN AMERICAN VOTER fect of left-right placement on the vote increases. This resy jus research on this contextual variable (e.g, te It 2010; Dalton 2008, 20115 Inglehart and Klingemann ata Siva Singer forthcoming). The ee of polarization it estnaned wn significant just past the value of 0.20 0n the polarization m eae ‘America, five countries score at OF below 0.21: Argentina, the 0, Ina ; Republic, Honduras, Panama, and Venezuela. Venezuela is an ominicay here, as it registers Jow polarization but a high connection betwen right placements on the vote; it could be that the contentious yes lef. polarized in other ways that generate left-right relevance, or Pepe factors are shaping that connection. In figure 8.3B, we sec that left-right identifications are significant nredictors of voter choice in systems that have low to moderate level of fragmentation, put not in systems that are highly fragmented, This finding is consistent with previous work on fragmentation and left-right identifications in Latin ‘America (Zechmeister and Corral 2013; Singer forthcoming). The evidence is thus quite strong that the relevance of a high number of parties for voter choice varies across regions. The specific i estimates in the figure @ ve driven by an outlier, Brazil, which has an eff ere number of parties a the high range of the scale, 10.4. Yet, if this case ;¢ removed, the interaction emains significant and negative, and in this case the prediction is that a significant relationship exists between left- right placements and the vote for those systems with approximately four pe fewer parties. In Latin America, five systems have scores greater than 4.0 on this measure: Cost Rica, Guatemala, Colombia, Argentina, and, of course, Brazil. 4 In figure 8.3C, we see that left-right identifications have no relation- ship to voter choice in systems with very low levels of programmatic Pay structuration, but they do have positive relationship in more & systems. The graph estimates that he effect of left-right placement’ comes significant at or above the 0.10 level of programmaticness The same five countries that score in the lowest ranks on polarization ls0S0° low (in this case, below 0.10) on programmaticness: ‘Argentina, the DO minican Republic, Hondas, Panama, and Venezuela. & the inverse relationship for clientelism: systems characteria levels of lientelistic effort are predicted to have n0 fi ship berween left-right identifications and the vote choice: 8 tios between these variables increases as clientelistic graph predicts that the effect of left-right placements Escaneado con CamScannerr- Left-Right Identifications and the Latin American Voter + 213 below a score of 17. Venezuela and Argentina have scores that ground and 7 i on this measure, while Panama and the Dominican Republic nd to 17 OF sot gt higher Scores hat number of these analyses, Venezuela stands out as an excep- larization, low programmaticness, and high clientelistic ef- Id predict lower connection between left-right placements rhe vote than we find in practic. One possible explanation for this is eas comparatively long history of democratic paty polities, while nother is that Hugo Chavez's Bolivarian revolution (and related politics and rhetoric) infused left-right significance into the system. While iden- ting which ofthese explanations is more apt is outside the scope of this hapten we cam assess che general question of whether leftist populist lead- ers who came into power at the turn of the twenty-first century triggered and/or reflect @ left-right ideological shift. The following section turns to that question. a tion: its low Ps fortscores woul Assessing the Potential Influence of Latin America’s Pink Tide ‘The above section focused on a set of factors found across party systems around the globe. In this last empirical section, I assess the potential rel- stance of Latin America’s pink tide to the political significance of left- right identities in contemporary Latin America, Latin America “eft turn” th the election of Hugo Chavez in Venezuela and began in the late 19908 wit continued with the subsequent election of left-leaning politicians across many Latin American countries. Given that the nature of party competi- tion matters (as identified in the previous section), then the pink tide may have influenced the political significance of the terms “left” and “right” in the region, Yet, extant scholarship offers two contrasting views of the pink tide and, thus, ewo different expectations regarding whether it would have influenced left-right identifications. On the one hand, some scholars have attributed the left’ electoral success to standard retrospective accountability voting, suggesting voters tossed ont right-leaning incumbents due to performance dissatisfaction rather than ideological conversion (Arnold and Samuels 2011; Levitsky and Roberts 20114; Queirolo 20133 but see Baker and Greene 2011), and have found little evidence of ideological coherence among this new left (Hiarbers, dé" Viies aid Steenbergen 2013; Wiesehomeier and Doyle 2012). On the other hand, the pink tide brought legitimacy to economic Programs defined for the mass public as belonging to the “left.” Take, Escaneado con CamScanner214 + THE LATIN AMERICAN VOTER for example, a speech by Ecuadoran President Rafael C, of 2011 in which he distinguished between the previous in fa eae that he stated revolved around capitalism and neolibes re detriment of workers and his new “gobierno de izquienda ae n dace), which he agved has improved the quality soci working class.” Thus, while many of these victories might ae for « been due to performance voting, it is conceivable that the age hate the lef, its programs, and the thetorical links offered by polit dation, pushed the public toward the left and strengthened ideological clans have tion in at least some countries. Structurg. To assess the potential impact of the pink tide on the political rel of left-right identifications across the Latin American region, I eee q oreaming to the question of the relationship between left-right ee fications and the vote. I develop a new moderating variable, which dis. saree between countries with presidents located on the left and 3 of system responsiveness: This finding is particularly relevant to the Latin ~ American context, where it helps to explain some cases in which we find comparatively long histories of democratic party competition combined we8IH Z'RV TTYL Escaneado con CamScannerpr Left-Right Identifications and the Latin American Vc foter + 223 notes ter pss ouside the scope of th : = is i ceived politica symbol, whether ee eee ese HY 3002). edlor noo partial pola cal whi ow eels fiterest me : se is eel eve OTe Mabe EE teamee saresfedgeable about politics who are those who are more interest ed do ra, Brod, and Telock 199%; fe more likely to take use of ch 3. There is pervasive variation in use of and fe poe Near co): : ee eee ees afc ovard the terms le and ah yesh lefeand ight tend obesh nce, referencing Brazil, Power ote ore desirable than anything right pain though left Ba gong the Mesa publi there as been eee cere vo erm “ee” (Zechmeister 20060). greater affect forthe term “right” ier reseten ba jg card there is a 1-10 scale that i , to change the subject . vegand 10 means Fight. nee lrg Te uber one ea ee one on ihe we speak of plea Ienigs me kof those eh the lef and others with the ri Lochs ade Sor people sympathize mor Nef and ‘ight have for you, and es i ae to the meaning that the a ii yourelfon this see” yon orn ee ee aes Analyses are adjusted for the complex sample desi ! scasBorometer studies. Replication code fo ple designs employed in the Ame reduc append or anaes in this chapers avaiable in & The mean erg espns rte re cored the qusion 5:5 th 18.1% identifying 0 swhle (atong those ho OF Dand 20.2% identifying on the right (place ying on be left (placements of 1 7-Countries are considered uni acements of 8, 9, OF 10). : Conus we ‘units of analy : cen oego nmr nals and tere ‘weighted equally in . The nonresponse rates in s in Ameri bat minted Ss ne Latin mein coun similar to those whi sow cos erica National Elesion Study (ANES), ae ee thought eer aa since 1972 ut which sr bene ey gy nd about i” option. Nonresponse averaged leant oni a sai snd 204 the ANES 1948-2004 cumulative ah 2 available at hep:/ ionstudi sc ri sn tpi//wwotelectionstuies.org significant varat i i sen ope poa be vee countries experienced changes ie pest tween aero and 2012. Honduras displays @ sede whe Tonresponse in 2012 (23.4%, a8 shown in figure 8.1) ue 01a nites 109%), Panama oe raga show far less non- a woe "3 respectively) than they did in 2010 (21.5% and xan alt is measure bas ttn orci ee on faexor analysing responses v2 batery of tm omer questions for moreinfomaton 9 its construction, see sears of education age cohort is linear measure in sight” across writ Escaneado con CamScanner+ THE LATIN AMERICAN VOTER 224 whch voting-age respondents are grouped a8 follows: 25 oF under ag. 56-65; 66 or older. : O85 6 | 11, The complete table of results, with country fixed effect, ig ays: Big the author and in the online appendix as table OA8.r, Abe og 12. Upon closer examination, som evidence ofa curvilinear relation j found between age cohors and left-right nonresponse, such that eth any the youngest cohort (25 and under) and those inthe older cohors (gg nein are marginally more likely to decline to answer the question compared rn) the middle cohorts, This is interesting in parc because it parallels fait tin reported by Carlin and Love (this volume). BS for age 13. Analysis accounts for the complex sample design and weights each 4 toan equ ube of respondents. Predicted probabilities calculated wat oY (heep://gking.harvard.edu/clarify). In separate analyses, I ran the basie ne ca") model for ench country separately. For every country, the signs on the cnet for political interest and external efficacy are negative; however, a : tailed, the coefficients for political interest are not significant in the and Costa Rica and the coefficients for external efficacy are not si Dominican Republic, Uruguay, and Peru, 14, These models should be taken as conservative. By not including issue are idiosyncratic to particular country contexts, for example, the civil conflict peace process in Colombia, I may underestimate the substantive significance , left-right semantics. I thank Sandra Botero, Rodrigo Castro, and the rest Public Opinion and Elections Working Group Workshop at Notre Dame U: sity for their suggestion that I clarify this point. 15, The positive coefficients on the economic measures for Honduras even in a reduced model in which the Gov't Ownersbip measure and the I measure, respectively, predict left-right identification, and are similar to a reported in Zechmeister and Corral (zorr). 16. The result for El Salvador is intriguing given that it may rep stantial decrease in the policy content of left-right semantics in that c Zechmeister and Corral 2011). 17. The results are consistent across models without controls, wit tion that the left-right variable becomes significant in the case of Pan list-wise deletion of those who do not respond to the left-right q these individuals at the mid-point of the left-right scale, and thenir variable (dichotomous measure coded t if the response g erwise) to account for their presence in the model. __ 18, Assessment of significance using a generous cut-off lev tailed. Among those identified for statistical significance, Pen threshold at p = 0.074. 19. Each of these measures is described in the introd Polarization, Programmationess, and Clientelism are based on dat cratic Accountability and Linkages Project (DALP) survey (see volume). The result presented here for Polarization is consistent if a measure, based on party polarization in Congress, is used (see Sing ing and the introduction of this volume). Likewise, the results for Clien® consistent with those based on a measure based on the 20ro AmericasBarom P< 0.10, tw. case of Panama ignificant in the Escaneado con CamScannerLeft-Right Identifications and the Latin American Voter + 22 5 of respondents indicating they were offered a material it ive ii of het and Altamrano this volume). amcor in tei anaes countries are weighted toan equal ub : oA rei nator of he dat Ts the ili cana FT cae viable from the autor ad took ers he jn St | variables are all included in the same model, the signs on the coefficient St esd ros neene; the eet for programma sy 6 eis gly cold with polation mesure hich a the programmaticness index). : santo gg-D presents the averge predicted marginal effects of left-right aon a aying levels on the contextual factors; the outputs based on ese conducted in Stata 12.0. si opinion com/nacional etl phpide12772, Last accessed July 16 2003, “ef” i ri ‘i 1 cae a “Tet” governments (pink tide countries) are Argentina, Bolivia secon Re, Ez, Sadar, Nias, Parag. Peru, Uruguay, and Venere. 7 es or Bla across the ears are allows: 198: 5.35 200: sat a 200 5.07 2006: 525 208: 5175 2010 525 202 5 Thus, 538 208 5 ight aie ovard he ee pro othe lesion of Coren, but ht vedo ge toward he center in the subsequent yer. In Ecuador mean i rows: 2001 46; 2004 53552006: 5.78 2008: 537; 2010 $435 208% 533. ‘Thus, in the case of Ecuador, there is a slight ‘tick toward the right in the year ‘har Correa was elected. ie Bolivia (ear: percent placing on the “lef,” meaning % 2 oF 3.00 the 1-10 seal) 1998 14.25 2000: 14.8; 2002! 17-55 2004: 15-8; 2006: 15.85 2008: 16.2; 2010: 2006: 9.8; 2008: 14.9; 2010: 13.35 ponent Of qpqi2012 1.7- Ecuador: 2001: 12.55 2004" 13.0 Be Tra: There is no clear pattern to nonresponse rates across time in either case creased over time in Bolivia. cxoept that, anything, nonresponse has in vy Considering the policy significance of left-right identifications, find no sridceo ofa post temporal increase in the policy significance of left-right iden- eitee oa ace rountres (see replication code, availabe from author, for tne on these anelyse), When looked at forthe region as a whole, [ also ses 20 dear influence ofthe pink tide on the policy significance of the left-right semantics (consider the lack of any pattern to the s-squared values reported in table 8.x and lefincumbency). 27. Many ofthese left leaders are part of a modern wave of populism. Kitschelt and Altanirano (this volume) find that left-populism is a stron predictor of cli- enn iefecvenes, is posible. 10 y then,,that.Latin~America’ of-populist victories and administrations have done little a 5 done, ltd Escaneado con CamScannerTABLE 8.2 Left: ‘Argentina Bolivia Brazil Chile Colombia Costa Rica Dominican Republic El Salvador Guatemala Ecuador Honduras Mexico Nicaragua a Paraguay Peru Uruguay Venezuela Latin America ‘Sou ment as a Predictor of Vote Choice N 879 1,155 1,116 612 630 806 927 644 431 966 698 781 863 778 692 973 1,034 768 14,753 AmericasBarometer 2012 by LAPOP (v ar r49/¥50). Note: $ 0.10, two-tailed. Output is coefficients and standard errors 7 R q 0.07 au 1 0.16 Guy Baage 0.01 0.60 | a2 ae ag a 0.02 oe 4 0.06 ‘a ol 0.06 ney 0.03 | 0.03 : oud 0.03 | 0.05 0.04 0.07 oat 036 | 0.08 4 for left-right pl on OLS regression analyses for each country (and, inthe final ov, for the pooled re 4 c rural (versus urban) residen 7 include controls antexien Replication code available in 1 voter choice. Independent vari ; results for age, gender, education, wealth, those and constants are not presente the online appendix and/or from author. Dependent v2" iables are scaled o to 1. cd here forthe sf ih Escaneado con CamScanner