Lessons from Latin America and the Caribbean

ELIZABETH J. ZECHMEISTER

back to issue

THERE has been an explosion in the quantity and quality of surveys, election polling, and research into voter decision making in the Latin American region in recent years. Election polls and pollsters have gained experience, sophistication, and credibility across the region. In a number of countries, election and exit polls have entered into a (mostly) positive pact with democratic processes, unpacking otherwise unseen election boxes and adhering to rules and professional norms about how and when results can be released.1

Despite these advances, election forecasting is ‘incipient’ in the region, with pundits often relying on individual polls to make predictions rather than using more data-rich approaches that aggregate polls in sophisticated ways.2 At the same time, comparative survey projects in the region have multiplied in number, rigour, and use. Comparative election studies and, to an extent, panel election studies are increasing in the region, most notably via the global Comparative Study of Electoral System (CSES) project. Other comparative public opinion projects3 have likewise been expanding in scope and rigour, yet these – like the Americas-Barometer by the Latin American Public Opinion Project (LAPOP)4 – typically deliberately avoid entering the field to collect data in close proximity to elections.

What value do such comparative non-election polls hold for understanding voter behaviour in the region? The answer is: a lot. The first section of this essay unpacks this answer, demonstrating ways in which comparative survey projects are improving understandings of electoral behaviour and related attitudes in Latin America and the Caribbean. The second part of this essay takes note of some advances and some challenges to survey research in the region.

 

Comparative survey projects allow insight into the ways that features of the political system affect the considerations that citizens use in making decisions at the polls. This is the premise of the recently published volume, The Latin American Voter, which spotlights how the structure and content of the menu of political options influence voter behaviour.5 Similar to the conclusions offered by another recent product of the prospering field of scholarship on Latin American comparative electoral behaviour,6 it concludes that, in general, the average Latin American considers group identities, policy, and performance at the ballot box. Yet, at the same time, the tendency of a voter to select candidates whose descriptive profile – by class, gender, ethnicity, or region – matches her own is conditional on the successful activation of those identities by entrepreneurial politicians.

Simultaneously, the voter’s inclination to make decisions based on policy stances and substantive programmes is shaped by the degree to which party systems, via polarization, offer clearly distinct alternatives. Further, an individual’s predilection to cast a vote that punishes or rewards output is greater to the extent that a particular performance domain is salient, and politics structured in a way that enables the individual to assign responsibility to the incumbent. In this way the region (Latin America) is much like others around the world: casting a vote for or against an incumbent based on perceptions of how well the economy is faring is more likely to the degree that economic volatility makes that issue salient, and when a limited set of stable party options makes it easy to pin responsibility for the economy on the sitting executive.

 

Common questionnaires at the core of comparative survey projects facilitate systematic analysis of how contextual factors influence citizens’ orientations toward the political system. At the very least, scholars of Latin American elections can build multi-level models that assess the ways that country-level factors affect voter decision making. The Americas-Barometer by LAPOP provides even more leverage in the pursuit of answers to questions about how and when context matters. For a start, the survey project covers 34 countries within the Americas; in each case, the same core set of modules is asked of a nationally representative sample of voting age adults in face-to-face interviews according to highly standardized procedures. The significant variation across the region in terms of the nature of electoral systems (e.g., variation in electoral rules regarding voter turnout [compulsory in some countries, voluntary in others]; variation in party entry rules, term limits, ballot structure, and the composition of congress and the electoral formulae that produce it; and variation in the nature and age of political parties) makes the region particularly fertile for research into the influence of country-level features on voter choice.

 

In the second place, the stratified nature of the sample design in each country allows analysts to investigate the extent to which over one hundred varied sub-national contexts affect the dynamics of political support. From such an approach, for example, we learn that the public’s tendency to hold the executive accountable for corruption among public officials is conditional on regional economic context: individuals who live in more prosperous regions within countries are more tolerant of corruption than those who live in comparatively more depressed zones.7

Comparative survey projects also facilitate investigation into how much proximity to elections matters for the attitudes citizens hold and the decision-making processes in which they engage. Taking advantage of variability in how close or far away surveys were fielded prior to elections in countries around the world, Michelitch and Utych8 have combined the AmericasBarometer with data from the Afrobarometer, AsiaBarometer, and the European Social Surveys to document the fact that levels of partisanship increase as elections draw nearer. Panel studies shed light on similar dynamics with respect to instability in vote-relevant attitudes. For example, via the Mexico panel study, McCann and Lawson9 used surveys conducted at regular intervals around the 2000 election in Mexico to show that while partisanship is more stable than other attitudes that are potentially relevant to elections, such as left-right self-placement and policy stances,10 it nonetheless fluctuates at the individual level across the course of the campaign.

 

The very fact that partisan ties fluctuate to a fair degree as one moves towards, through, and away from elections in Latin America, not to mention other evidence of attitudinal instability, should elicit concerns about the ability of comparative survey projects to reveal accurate insights into the dynamics of voter decision making. Add to this the fact that support for the incumbent is typically over-reported in non-election surveys,11 and one certainly has reason to raise questions about the utility of comparative survey projects for the study of Latin American voter behaviour.

Yet, there is a strikingly high correlation between the reported votes of respondents to the AmericasBarometer and actual electoral outcomes.12 Furthermore, replications of vote decision models with data sets containing only those national surveys from comparative survey projects that were collected in close proximity to an election tend to return similar findings with respect to the determinants of voter choice in Latin America.13

Regardless of the timing of the survey, pooled analysis of comparative data sets reveals that the average Latin American voter is more likely to vote for parties from the political left to the degree that voter is male, poor, not very religious, indigenous, in favour of state intervention in the economy, an advocate for prevention oriented anti-crime policies, more approving of same-sex marriage, distrustful of the U.S., and likely to identify as a leftist. On average, Latin Americans are more likely to cast a vote against the incumbent to the degree that they perceive the national economy to have deteriorated, find corruption to be the norm among public officials, and feel unsafe in their own neighbourhoods.14

 

The fact that similar average tendencies appear across analyses of data sets that pool all available data or election-proximate data alone is an important reminder that one should not make the fallacy of assuming that because levels (that is, mean values on a variable) change that correlations among variables necessarily also shift. Rather, average values on election-relevant attitudes may fluctuate across time within and across countries in the Latin American region, but the relationships among these variables are typically better predicted by features of the political system and campaign dynamics than by proximity to an election.

 

Comparative survey projects also shed light on myriad other factors that influence citizens’ decisions about how much legitimacy to accord to elections and the political parties that compete in them, whether to turn out to vote or not, and for whom to cast a ballot. The comparative nature of such projects allows critical analytical leverage when it comes to interpreting the meaning of values returned by public opinion data collection efforts. For example, the 2014 AmericasBarometer found that one-in-four Brazilians was in some level of agreement with the notion that elections in that country can be trusted.15 Is 25% high, low, or middling? Is it a sign of fragility in the democratic system?

Since the fielding of the survey, political protests and crises have mounted in Brazil and the elected president has been removed from office to face an impeachment trial.16 With the benefit of hindsight, one can use the recent political turmoil in the country to look back and interpret 25% as meaningful. If all one had was a single data point from 2014, this would be the best, and yet deeply unsatisfying, available option.

Instead, the comparative survey project provides cross-national and cross-time perspective. The 2014 AmericasBarometer asked members of the public in twenty-four countries about their degree of trust in elections.17 Figure 1 displays results for the region on this question. The data show that, considering just the 18 Latin American countries, only one – Colombia – has a public that is less trusting of elections than Brazil. Bringing the Caribbean into the picture, trust in elections in Brazil in 2014 was on par with that found in Jamaica and higher than that found in Haiti, where just 14.8% of survey respondents expressed some degree of trust in elections. In contrast, far higher degrees of trust in elections are found elsewhere in the Southern Cone: in 2014 in Argentina, 52.4% of the public expressed some degree of trust in elections, in Chile 61.5% did, and in Uruguay an impressively high proportion of citizens, 76.5%, reported that they found elections in that country to be trustworthy.

FIGURE 1

Trust in Elections in Latin America and the Caribbean, 2014

 

Given this comparative context, the level of trust in elections expressed by Brazilians in 2014 can be declared to be decidedly low. The fact that the AmericasBarometer has been conducted regularly, with national samples and a standardized approach across time, also permits a temporal lens through which one can assess the degree of legitimacy the public gives the region’s electoral systems. Cross-time output from the AmericasBarometer’s surveys in Brazil was in fact another harbinger of political crisis. As Russo18 documents, general respect for Brazil’s political institutions had been sliding for years prior to the current turmoil in the country.19

In summary, despite being intentionally scheduled to not coincide with national elections, comparative survey projects have much to offer with respect to understanding the dynamics of voter decisions and dispositions. The availability of high quality survey data for the Latin America and Caribbean region permits multi-level analyses of the degree to which contextual factors predict voter behaviour and facilitates investigations into the factors that influence presidential approval, party support, turnout, vote choice, and vote intention. These data also facilitate prospective prognoses for political and electoral stability, or instability, within countries in the region.

 

In taking stock of the quality of survey research in the region, it is fair to say that it varies significantly, and yet the general trend has been toward higher quality and greater transparency and openness with respect to design and data. The availability of new technologies, and innovations based on technological advances, has further improved the region’s capacity to deliver high quality survey data to analysts and scholars of electoral behaviour. It is fair to say that LAPOP, via its AmericasBarometer, has been a standard-setter in its advancement of the scientific rigor and reliability of survey research in the region.

 

For example, one persistent challenge to quality survey research in Latin America and the Caribbean is the tendency for census data and maps to be unavailable or outdated. Without accurate information regarding the distribution of a population, it is impossible to draw a nationally representative sample. To address this challenge, LAPOP developed the LAPOP Survey Sample Optimizer (LASSO 2©), a software programme that reads rooftops from publicly available satellite images and allows LAPOP to use these data to generate or refine samples. LAPOP has also advanced the use of hand-held electronic devices (tablets and phones) for data collection in the field, and has developed and adopted mechanisms by which these devices can be used to ensure that interviews are conducted in the right place (via a geo-fence programme that alerts interviewers and the quality control team at LAPOP when an interview registers GPS coordinates outside of a pre-programmed sample location) and to ensure that all survey protocols are followed (e.g., voice recordings for quality control allow the quality control team to verify that the consent form and other questions are read aloud in their entirety; other modules are used to verify the identification of the interviewer to assure that only interviewers who completed project training are involved in the fieldwork).20

 

One challenge that remains a critical issue for survey researchers in the Latin American and Caribbean region is the dangerous criminal activity that is present and increasing in certain countries.21 Latin America and the Caribbean contain some of the most criminally violent places in the world. High levels of criminal and gang activity prevent the entry of survey teams into certain communities and, in the worst case scenarios, have resulted in interviewers being robbed, assaulted, and even killed. While Internet- and phone-based surveys would be safer, these technologies have not penetrated the region to the point where such approaches yield representative and/or economical results. For this and related reasons, face-to-face interviewing remains the mode of choice. Thus, looking to the future of election polls, comparative survey projects, and other public opinion research in Latin America and the Caribbean, it is impossible not to consider the rising challenges to survey based research that stem from exceptionally high levels of crime, violence, and insecurity.

 

Election polls and exit polls have been gaining in use, quality, and reputation across Latin America and, to a lesser degree, the Caribbean. At the same time, these provide only one of many data sources on electoral behaviour in the region. In tandem with increasing professionalism in the survey research industry in the region and the spread of expertise, technology, and commitments to scientific practices advanced by the World Association of Public Opinion Research (WAPOR) and others, comparative survey projects have blossomed in Latin America and the Caribbean in recent years. The data from these projects are frequently referenced by citizens, policymakers, the media, and scholars when discussing factors of relevance to electoral democracy in the region. Findings from the AmericasBarometer, for example, were referenced more than once per day on average in media in the region in 2015. The rigorous scientific approach used by the project and its non-partisan nature have made it a widely used resource for the study of voter choice within and across Latin American and Caribbean countries. This comparative project, among others, is advancing our understanding of who votes, for which options, and why in the Americas and, as well, facilitating the study of the extent to which, and when, the public views the electoral system as legitimate.

 

* Many thanks to Daniel Montalvo and Mollie Cohen for providing feedback on an earlier version of this essay; all errors and omissions are the author’s own.

Footnotes:

1. Even within countries, the degree to which exit polls have helped legitimize or call into question electoral processes has varied across recent national contests; see René Bautista, ‘An Examination of Sources of Error in Exit Polls: Nonresponse and Measurement Error.’ PhD Dissertation, University of Nebraska-Lincoln, 2005.

2. Kenneth Bunker and Stefan Bauchowitz, ‘Electoral Forecasting and Public Opinion Tracking in Latin America: An Application to Chile.’ Working Paper, London School of Economics and Political Science, 2015.

3. In addition to LAPOP’s AmericasBarometer and the CSES, comparative survey projects in the region include the Latino-barómetro, the World Values Survey, and CIDE’s Americas y el Mundo survey project. Many countries in the region are also included in comparative studies conducted by Pew.

4. The AmericasBarometer is a regular regional survey designed and implemented by the Latin American Public Opinion Project (LAPOP) in collaboration with partners across the region. Core funding is provided by USAID and Vanderbilt University, and other support has come from the Inter-American Development Bank, the United Nations Development Programme, the Tinker Foundation, the U.S. National Science Foundation, Brazil’s national science foundation (the CNPq), and various academic institutions across the Americas. The AmericasBarometer data are available free of charge, with no embargo or registration process, at the project’s website: www. vanderbilt.edu/lapop.

5. Ryan E. Carlin, Matthew M. Singer, and Elizabeth J. Zechmeister, The Latin American Voter: Pursuing Representation and Accountability in Challenging Contexts. University of Michigan Press, Ann Arbor, MI, 2015.

6. Richard Nadeau, Éric Bélanger, Michael S. Lewis-Beck, Mathieu Turgeon, and François Gélineau, Latin American Elections: Choice and Change. University of Michigan Press, Ann Arbor (forthcoming); Helcimara Telles and Alejandro Moreno (eds.), Comportamento Eleitoral e Comunicação Política na América Latina. Editora UFMG, Belo Horizonte, 2013.

7. Elizabeth J. Zechmeister and Daniel Zizumbo-Colunga, ‘The Varying Political Toll of Concerns about Corruption in Good versus Bad Economic Times’, Comparative Political Studies 46(10), 2013, pp. 1190-1218.

8. Kristin Michelitch and Stephen Utych, ‘Electoral Cycle Fluctuations in Partisanship: Global Evidence from 86 Countries.’ Working Paper, Vanderbilt University and Boise State University (n.d.).

9. James A. McCann and Chappell Lawson, ‘An Electorate Adrift? Public Opinion and the Quality of Democracy in Mexico’, Latin American Research Review 38(3), 2003, pp. 60-81.

10. The greater instability of left-right self-placements, owing to their tendency to not be anchored in substantive meaning to the same degree in Latin America as in advanced industrialized countries (see chapter by Zechmeister in Herbert Kitschelt, Kirk A. Hawkins, Juan Pablo Luna, Guillermo Rosas, and Elizabeth J. Zechmeister, Latin American Party Systems. Cambridge University Press, 2010), may be one reason why election proximity does not predict tendencies for individuals to place themselves on the left-right scale in Latin America (see Elizabeth J. Zechmeister and Margarita Corral, ‘Individual and Contextual Constraints on Ideological Labels in Latin America’, Comparative Political Studies 46(6), 2013, pp. 675-701).

11. Ryan E. Carlin, Matthew M. Singer, and Elizabeth J. Zechmeister, 2015, op cit., fn 5.

12. Ibid.

13. Ibid.

14. Ibid.

15. AmericasBarometer question B47a asks: To what extent do you trust elections in this country? Individuals answer on a 7-point scale, where 1=not at all and 7=a lot. For the purposes of this discussion, responses above the mid-point of 4 on the scale are considered to indicate some degree of trust.

16. http://www.nytimes.com/interactive/2016/world/americas/brazil-dilma-rousseff-impeachment.html

17. While the 2014/15 round of the Americas-Barometer covered 28 countries, the question was not asked in the Bahamas, Barbados, Suriname, or Trinidad & Tobago.

18. See analysis and discussion in this 2015 report by Gui Russo: http://www.vanderbilt. edu/lapop/insights/ITB025en.pdf

19. AmericasBarometer question B2 asks: To what extent do you respect the political institutions of (country)? With responses scaled to run from 0 to 100 degrees of trust, Russo (2015) shows that mean degrees of trust in Brazil’s political institutions fell from 57.5 degrees in 2010 to 53 degrees in 2012, and then to 48.2 degrees in 2014.

20. For a thorough discussion of LASSO and LAPOP’s use of smartphone and tablet technologies to advance the quality of fieldwork, see J. Daniel Montalvo, Mitchell A. Seligson, and Elizabeth J. Zechmeister, ‘Data Collection in Cross-national and International Surveys: Latin America and the Caribbean.’ Chapter in Tim Johnson, Brita Dorer, Ineke Stoop, Beth-Ellen Pennell (eds.), Advances in Comparative Survey Methods: Multicultural, Multinational and Multiregional (3MC) Contexts. Wiley, New York (forthcoming). All respondents to LAPOP’s surveys are provided with information about the study prior to its start, via an IRB-approved informed consent process; when a section of the interview will be recorded for quality control purposes, participants are told this prior to the start of the survey.

21. Rising insecurity in the region has left some neighborhoods impenetrable to survey researchers, either because they are too dangerous for interviewers to enter or because they are gated off as a precaution against crime.

top