Incorporating Marginal Citizens and Voters

Abstract

In this article, we study how social assistance shapes election results across Latin America. Case studies in several countries have found electoral effects, yet it remains unclear whether and how effects vary cross-nationally, and whether electoral effects are due to mobilization or persuasion. We theorize that programs mobilize non-voters and convert the opposition simultaneously, but that the effects vary based on country-level political and programmatic differences. Using 2012 AmericasBarometer data, we develop a unified cross-national model that confirms that public assistance makes recipients more likely to turn out and, once at the polls, to vote for the incumbent. Compulsory voting laws and program politicization magnify the electoral effects of social assistance, but effects do not vary by presidential ideology or program conditionalities. These findings are consistent with the perspective that Latin American voters are boundedly rational, retrospective agents whose behavioral choices depend on their resources and environmental context.

Keywords

Latin American politics political economy elections public opinion and voting behavior social welfare programs

Over the last two decades, Latin American governments have transformed their social policies. Where policies were once universal in principal, and in practice disproportionately targeted to urban middle classes and unionized workers, governments now seek to help their poorest and most vulnerable citizens through targeted means-tested aid (Barrientos & Santibáñez, 2009). Conditional cash transfer (CCT) programs such as Bolsa Família in Brazil or Oportunidades in Mexico provide highly targeted cash assistance for impoverished families on the condition that they keep their children in school and make use of public health services. Even in countries such as Venezuela, where assistance is not strictly conditioned on compliance with public health or education goals, social service budgets targeted at the very poor have grown rapidly. While controversy remains over these new programs’ exact contribution to Latin America’s modest declines in inequality, as well as its more pronounced declines in poverty, global development scholars broadly agree that the new programs have helped the poor by reducing reliance on child labor, increasing school enrollment and attendance rates, increasing public health service usage, and improving caloric intake (see Adato & Hoddinott, 2010).

Yet many observers note that these programs may boost not only human development prospects but also executives’ electoral bases. Indeed, by altering beneficiaries’ pocketbook calculations and directly linking recipients’ well-being with state actors and policies, social assistance has the obvious potential to alter voting behavior. Many scholars have criticized past assistance programs in Latin America as legalized, politically expedient vote-buying schemes (Bruhn, 1996; Brusco, Nazareno, & Stokes, 2004; Calvo & Murillo, 2004; Diaz-Cayeros, 2008; Graham & Kane, 1998; Penfold-Becerra, 2007; Rocha-Menocal, 2001; Schady, 2000). Although many new programs were designed to reduce or circumvent strategic manipulation (Fiszbein & Schady, 2009; Sugiyama & Hunter, 2013), informal practices may subvert formal institutions.

Case studies in Brazil, Colombia, Honduras, Mexico, and Uruguay provide compelling evidence that the new wave of CCT programs has helped incumbent executives maintain or build electoral support (e.g., Baez, Camacho, Conover, & Zárate, 2012; De La O, 2013; Licio, Rennó, & Castro, 2009; Linos, 2013; Manacorda, Miguel, & Vigorito, 2011; Zucco, 2013). Still, important questions remain unresolved. The first and most basic regards the extent to which the pro-incumbent effects found in these few countries, ones with politically important, large, and conditional social assistance programs, generalize more broadly across the region. Indeed, the case studies themselves differ in the magnitude and nature of reported effects. For instance, evidence from Honduras suggests that the country’s program has no significant effect on electoral behavior related to incumbent presidents but a highly significant one on mayoral voting (Linos, 2013), whereas evidence from Uruguay suggests that the program there increased presidential support by 11 to 13 percentage points (Manacorda, Miguel, & Vigorito, 2011). Other case studies find 7 and 9 percentage point increases in turnout and incumbent vote share, respectively, from Mexico’s Progresa program (De La O, 2013); effects on voter registration, turnout, and incumbent vote share of about 1.5 to 2.5 percentage points each, from Colombia’s Familias en Acción program (Baez et al., 2012); and a 0.12 to 0.18 percentage point increase in incumbent vote share for each percentage point increase in municipal Bolsa Família receipt in Brazil (Zucco, 2013). It can be hard to reconcile these cross-country differences due to varying measures, theoretical approaches, and methods. More worrisomely, this country-by-country approach may have led to the non-publication of relevant null results in countries where social assistance has been found not to have electoral effects.

This variation in outcomes is matched by great variation in political and institutional contexts. Social assistance programs are highly institutionalized in some settings and personalized extensions of the executive office in others. In some countries, they are implemented by leftist administrations publicly perceived as committed to helping the poor, and in others by centrist or center-right administrations. In some countries, programs entail strict conditions requiring compliance with public health and education goals; in other countries, conditionalities are unenforced or do not exist at all. Does institutionalization make program viability less dependent on electoral outcomes and therefore less subject to strategic appeals by the incumbent? Are beneficiaries more likely to reward leftist than rightist incumbents? And are they more or less likely to support incumbents when cash transfers bear behavioral requirements? Answering these questions requires a cross-national comparative framework.

Beyond the direction of vote choice, we also address the puzzle of whether and how programs pull non-voters into the electorate. Social assistance programs, we argue, can significantly decrease the costs and increase the benefits of going to the polls for marginalized citizens with few preexisting ties to the state or the political system. Indeed, a recent study attributes the pro-incumbent effects of Mexico’s Progresa program to mobilization of non-voters rather than conversion of the opposition (De La O, 2013). However, such mobilizing effects may be less pronounced in countries enforcing compulsory voting rules, which already drive turnout among many marginalized citizens targeted by social assistance programs. At the same time, we argue that newly incorporated voters under compulsory voting weight assistance benefits particularly strongly when deciding whom to support, as they lack other prior political ties.

In this article, we seek to develop a more comprehensive theory of social assistance and electoral outcomes. We argue that receiving assistance alters the boundedly rational utility calculations of both those who would otherwise stay home on election day and those who would otherwise vote for the opposition, calculations that depend strongly on the political, institutional, and policy context. Empirically, we propose a unified model of vote direction and turnout (Lacy & Burden, 1999), with the individual as the unit of analysis. We provide the first Latin America-wide test of the electoral effects of conditional and unconditional aid to the poor using the 2012 AmericasBarometer, the first public opinion survey to measure social assistance beneficiary status across the region. Our dependent variable, which we analyze with non-hierarchical and hierarchical multinomial logistic regression, is based on a question regarding how respondents would vote in a hypothetical election held during the week after the survey interview: abstain, vote for the incumbent, vote for an opposition party/candidate, or vote null/blank. Response options are thus comparable across the region, independent of domestic party systems and campaigns.

This analysis demonstrates that country context meaningfully shapes the electoral effects of social assistance in Latin America. In particular, assistance more strongly drives pro-incumbent voting under compulsory voting rules, and when presidents directly control the relevant agency. At the same time, the conditional aspect of CCT programs appears largely irrelevant for recipients’ utility calculations, as does the incumbent administration’s ideology. These findings suggest that social assistance recipients are boundedly rational, pocketbook voters, and that there are institutional arrangements that can rein in the tendency and ability of political elites to use social policy strategically.

Understanding the extent to which electoral effects are replicable across countries is important not only for scholars curious about the nature of the “social assistance electorate” but also for policy designers and the general public. If social assistance programs really do shore up a permanent pro-incumbent voting base, this is relevant information for public policy debates in countries in the process of implementing, expanding, or reforming assistance programs. And if aspects of program design can enhance or mitigate such effects, this is also important information. The significance of our approach is thus that it reveals that both preexisting institutional arrangements and elements of social programs themselves shape what policy makers, politicians, and publics can expect. At the same time, as we discuss in the conclusion, it appears that assistance programs present trade-offs between competing normative objectives, such as accountability and programmatic orientation.

Social Assistance, Electoral and Policy Institutions, and Electoral Outcomes

Why might social assistance programs affect presidential voting? Scholars have found that social policy shapes the democratic consequences (positive or negative) of economic hardship (Campbell, 2012). Both elite initiative and citizens’ behavioral responses to policy incentives may lead to a correlation between assistance receipt and electoral outcomes. From an elite perspective, incumbents seeking their own reelection or that of their allies typically claim credit for or politicize social assistance programs, especially as elections draw closer (Alt & Rose, 2007; Ames, 1987; Franzese, 2002; Schady, 2000; Tufte, 1978). Credit claiming can and does take many forms, from direct actions (e.g., press releases, interviews, or public announcements) to indirect ones (e.g., displaying pictures of the president in assistance intake offices, or mentioning the executive office of the president in ads). Assistance programs target a largely unorganized segment of the population whose interests can be met at a much lower marginal cost than those of middle- and upper-class citizens. Thus, vote-seeking but budget constrained politicians can more easily and often more visibly increase the number of people benefited by their administration or the per capita utility of benefits through assistance programs than through universal public goods (Taylor-Robinson, 2010). Targeted assistance policies are particularly effective tools for incumbents because their effects are more immediate than those of other economic policies and because they often employ unique brands and identifying features that are easily marketed (Akhmedov & Zhuravskaya, 2004). Thus, elite efforts can influence recipients’ attributions of blame or reward, shaping retrospective evaluations as well as perceptions of risk from an incumbent loss.

At the citizen level, we argue that social assistance benefits simultaneously reduce the costs and increase the benefits of turning out to vote, the latter in part by shaping the calculated utility of a vote for the incumbent. First, benefits reduce the costs of voting simply by providing new income that helps very low-income households deal with the immediate and opportunity costs of going to the polls, from bus fares to time away from work. In addition, many programs require (and help) recipients obtain identification documents necessary for both program registration and checking in at the polls on election day. Familiarization with state bureaucracy through program participation could also reduce the psychological or cognitive costs of turning out at the polls for marginalized families with very few prior positive experiences with the state. Thus, we expect that participating in social assistance programs may increase turnout across the board.

Second, social assistance can increase the calculated benefits of voting and of incumbent support. Most generally, benefits may increase psychological attachment to the state and national politics; this may be one of marginalized families’ few positive experiences with a national bureaucracy. Increased attachment (or decreased aversion) may spill over into higher turnout.

More specifically, and perhaps importantly, social assistance can affect the expected utility of voting for distinct candidates. Here, we draw on the psychological theories of bounded rationality (Kahneman & Tversky, 1979; Simon, 1985, 1995). Voters intend to make a rational choice between alternatives, but due to both individual cognitive limitations and environmental complexities they are unable to fully calculate the expected utility from all possible choices. Instead, they use heuristics to satisfice, or to come to a satisfactory rather than a utility maximizing decision (Lau & Redlawsk, 2001). Such choices are subject to a status quo (i.e., reference point) bias that favors the incumbent (Levy, 2003).

It is no easy task for voters to link their pocketbook economic status with national-level electoral choices even under the best of conditions (Gomez & Wilson, 2006; Powers, 2001). Indeed, the first hurdle to pocketbook voting based on public assistance is for citizens to recognize that they receive a benefit from the government (Mettler, 2011). Assuming that hurdle is met, assistance recipients have a strong and direct self-interest in maintaining benefits, not unlike lower-income recipients of Social Security in the United States (Campbell, 2003). The relative impact of even small benefits can be quite large for needy families. For instance, in Brazil, families that fit within the extreme poverty constraints of the Bolsa Família program can receive a cash benefit producing a minimum increase of 46.8% of the household’s formal declared income.¹ Once they discover that the state can directly meet their needs, assistance recipients gain a new stake in political contests (Graham, 1994). Thus, it is perhaps not surprising that assistance programs produce an instrumental behavioral change in recipients.

But social assistance receipt might trigger not only pocketbook but also issue voting (Schneider & Jacoby, 2003). Personally receiving assistance not only informs beneficiaries about public policy, but it increases the salience of that information. When asked in the survey interview context for political evaluations, recipients who recognize that they receive a government benefit have a strongly linked policy evaluation come immediately to the “top of the head” (Feldman, 1995; Zaller, 1992). While non-recipients may likewise approve of redistributive spending and have information about government efforts, that information will be less salient.

At the same time, benefits’ effects on the expected utilities of vote alternatives should depend on the political environment. Thus, we should observe that electoral effects vary across settings. We focus on and test the impact of four environmental conditions: assistance program independence from the executive, presidential ideology, enforcement of program conditions (if adopted), and compulsory voting laws. The country case-study approach used in the bulk of scholarship on the electoral effects of targeted assistance has thus far obscured the role of such country-level features.

First, we expect that assistance programs that are bureaucratically independent of the executive will have a weaker electoral effect. The mechanism may be retrospective; institutionalization constrains the executive’s capacity to claim credit for policy outcomes and may reduce citizen attribution of responsibility for program administration to the incumbent, even in the Latin American environment of strong presidentialism. Alternatively, the mechanism may be prospective. When an executive directly controls a program, citizens may perceive greater risk in an opposition victory, especially if the incumbent frames the election in those terms. A new party in power could plausibly bring sweeping changes, altering program structure or canceling a program altogether (Stokes, 2001). In contrast, when assistance is controlled by a permanent and independent bureaucracy, presidential discretion is limited, especially if the agency structure empowers political actors whose tenure or political or budgetary interests are not immediately dependent on the executive (Lewis, 2003). Such an organization is much more difficult to dismantle and doing so may entail steep political costs as well as legislative battles. The emergence of such an agency signals broader societal commitment to social welfare.

Second, following Latin America’s “leftward turn” of presidential politics in the mid-2000s, one might wonder whether leftist elites benefit disproportionately from social assistance. Consistent with theories of issue voting and issue ownership, the public may be more likely to perceive assistance programs as expressions of incumbent commitment to the poor when the incumbent is on the left (Bélanger & Meguid, 2008). Thus, for instance, Brazil’s center-left President Lula was widely perceived as a friend to the dispossessed, and the Bolsa Família program as an expression of his personal commitments. In addition, the prospective pocketbook risk associated with a potential opposition victory may often be more credible when the incumbent is on the left and the opposition on the right. Nonetheless, we can see no reason for a retrospective pocketbook electorate to take the incumbent’s ideological persuasion into account. Social assistance programs have been embraced by incumbent executives on both the right and left in Latin America; behavioral conditionalities have helped to make such programs ideologically agreeable on the center-right. Indeed, centrist and rightist incumbents (e.g., in Mexico and Colombia) have reaped substantial benefits from assistance (Baez et al., 2012; Sewall, 2008). Thus, we doubt whether leftists will manage to accrue extra advantage from the welfare dole.

Third, many assistance programs aim not only to reduce poverty but also to make society more inclusive and to help recipients become more able citizens, by stipulating public health and human development responsibilities as part of the assistance package contract. While in all but one of the countries we examine there is at least one major social program that on paper stipulates some behavioral requirement (the exception is Venezuela), not all countries effectively enforce conditions. Thus, conditionality has become an important variable in the region’s social assistance regimes. Various studies examine the electoral effects of CCT programs, yet existing scholarship does not fully theorize whether or how conditions themselves matter. The extent to which conditionalities amplify or reduce the impact of social assistance under a framework of bounded rationality ultimately depends on beneficiaries’ attitudes toward the conditions. If conditionalities decrease the expected utility of benefits, they will depress pro-incumbent effects; they will boost incumbent support if they increase the value of social assistance (for instance, perhaps leading to greater citizenship and inclusion, as described by Hunter & Sugiyama, 2009, 2011; Sugiyama & Hunter, 2013).² In addition, unenforced conditionalities may reflect a lack of state capacity, which may correlate with poor provision of program benefits. Thus, it is an open empirical question whether enforcing program conditionalities helps or hurts incumbents facing a boundedly rational public.

Fourth, enforced compulsory voting laws may somewhat reduce the turnout effects of social assistance, as the electorate incorporated under compulsory voting overlaps with those mobilized by social assistance (Panagopoulos, 2008). At the same time, new voters incorporated through either compulsory voting or social assistance programs often lack preexisting ideological or partisan ties, social networks, or other political information that would have induced them to vote in the absence of favorable public policies. Such voters may be especially likely to weigh benefits heavily in their boundedly rational utility calculations.

Data and Measurement

We test these hypotheses using individual-level data from the 2012 AmericasBarometer survey.³ Respondents from 24 countries in Latin America and the Caribbean answered the question: “Do you or someone in your household receive monthly assistance in the form of money or products from the government?”⁴ A response of “no” was coded as “0” and affirmative responses as “1.”⁵ This forms our key independent variable. The dependent variable, presidential vote, reflects what respondents said they would do “if the next presidential/general elections were being held this week”: not vote; vote for the incumbent candidate or party; vote for the opposition; or vote, but leave their ballots blank.⁶ We first model responses in 23 of the 24 countries using a multinomial logit model with country fixed effects to test whether prior findings in specific countries hold across the region as a whole, while accounting for the interdependent nature of turnout and candidate choice.⁷ We then assess the relative impacts within subgroups of countries defined by indicator variables for enforced compulsory voting, program conditionality, ideologically leftist president, and assistance program independence relative to the executive office (see Table 1).⁸ Finally, we test the statistical significance of the differential effects by moving to a hierarchical multinomial logit analysis. Due to data limitations with the country-level variables and to maintain consistency across our models, we only analyze responses from citizens in 19 of the 24 countries in the hierarchical models.⁹

Table 1.

Country-Level Variable Coding and Estimated Proportion of Population Receiving Government Assistance for Countries in the 2012 AmericasBarometer Sample.

Country	Compulsory voting (enforced)	Program conditions (enforced)	Presidential control of program	Leftist president	Proportion receiving assistance
Argentina	Yes	Yes	No	Yes	0.15
Belize	No	—	—	—	0.09
Bolivia	Yes	No	No	Yes	0.55
Brazil	Yes	Yes	No	Yes	0.19
Chile	No	Yes	No	No	0.15
Colombia	No	Yes	Yes	No	0.16
Costa Rica	Yes	Yes	No	Yes	0.15
Dominican Republic	Yes	Yes	Yes	No	0.23
Ecuador	Yes	No	No	Yes	0.17
El Salvador	No	Yes	Yes	Yes	0.11
Guatemala	No	Yes	No	No	0.08
Guyana	No	—	—	Yes	0.20
Haiti	No	—	—	No	0.16
Honduras	Yes	No	Yes	No	0.05
Jamaica	No	Yes	No	Yes	0.20
Mexico	No	Yes	No	No	0.16
Nicaragua	No	—	—	Yes	0.08
Panama	No	No	No	No	0.09
Paraguay	Yes	No	Yes	Yes	0.08
Peru	Yes	No	Yes	Yes	0.07
Suriname	No	—	—	No	0.22
Trinidad and Tobago	No	No	No	Yes	0.19
Uruguay	Yes	Yes	No	Yes	0.17
Venezuela	No	No	Yes	Yes	0.08

Source. Leader ideology is from the World Bank Database of Political Institutions. Data on program conditionality and presidential control are both collected by Cecchini and Madariaga (2011). Data on compulsory voting is from International IDEA’s Voter Turnout Database (http://www.idea.int/vt/viewdata.cfm). Coding for conditionality and presidential control in Venezuela is based on Penfold-Becerra (2007). The proportion receiving assistance is based on responses to 2012 AmericasBarometer survey question CCT1NEW (an indicator variable for “Received public assistance”).

Countries with no data (—) on any variable are excluded from hierarchical models. IDEA = Institute for Democracy and Electoral Assistance.

In addition, we account for several alternative individual-level explanations. A variable measuring effectiveness of the current administration is an index (rescaled to run from 0 to 1) and is based on a series of questions asking respondents to indicate on a 7-point scale to what extent they would say that the current government fights poverty, corruption, and unemployment; protects democratic principles; and improves security.¹⁰ Those who perceive the government as more effective should be more likely to support the incumbent, all else equal.

Partisanship is yet another important factor. Preexisting partisan ties to the incumbent might confound the analysis, particularly if co-partisans of the incumbent are more likely to be targeted for or to seek out social assistance benefits. Conversely, anti-partisans might be less likely to seek out or receive benefits. We control for this possibility by including a categorical measure of partisan identification in our models.¹¹

The complete model controls for several other potentially confounding factors.¹² Support for state intervention in the economy could drive participation in social assistance and support for the incumbent. Therefore, we include an index of support for an interventionist state (again rescaled to run from 0 to 1) to ensure that any relationship found between assistance and the presidential vote is not due to ideological perceptions of the role of the state in society.¹³

Another key concern is that participation in social assistance programs is non-random by design: Only the poorest citizens are eligible. Thus, any effect of receiving cash transfers might simply reflect an underlying tendency among impoverished families. Thus, we include a relative measure of household wealth, measured in quintiles within each country (running from 1 to 5).¹⁴ To determine the impact of social assistance independently of broader economic perceptions and events, we include perceptions of the current national and personal economy (on 5-point scales and rescaled from 0 to 1) and indicators for personal or household unemployment and household income loss, both in the last 2 years.¹⁵ Any decrease in economic well-being is expected to lead to a lower likelihood of voting for the incumbent. We also include two indicators of crime and corruption victimization; respondents victimized by crime and corruption in the previous year may show lower incumbent support.¹⁶

Furthermore, we account for variables that may relate to political participation. We control for the respondent’s levels of internal efficacy, general system support, support for democracy, and community participation, each coded to run from 0 to 1. Finally, we also control for the respondent’s age, level of education, gender, and the size of place of residence. Many programs target female heads of household; this necessitates a control for gender of the survey respondent to ensure that the effects observed are not due to gender. Furthermore, due to suggestions that social assistance programs are biased in favor of rural areas (“Brazil’s Bolsa Família,” 2010), the model controls for the size of the respondent’s area of residence.

Results: Social Assistance, Electoral and Policy Institutions, and Electoral Outcomes

Individual-level results from a multinomial logit regression analysis appear in Table 2.¹⁷ The model includes country-level fixed effects, which are not displayed for ease of presentation. Using a 23-country sample, the results from Table 2 partially confirm De La O’s (2013) findings based on the Mexican case: social assistance largely shapes electoral outcomes by mobilizing voters. On average, assistance recipients are less likely to abstain than non-recipients, and recipients who say they would turn out are more likely to choose to support the incumbent.¹⁸ The fact that social assistance has a robust impact on turnout and vote choice despite our stringent controls strengthens our confidence in our results.

Table 2.

Multinomial Logit Model: Determinants of Vote Choice and Turnout (With Incumbent Vote as Baseline), 2012 AmericasBarometer.

	Would abstain		Would vote for opposition		Would vote blank/null
	Coefficient	SE	Coefficient	SE	Coefficient	SE
Received public assistance	−0.268***	0.065	−0.144*	0.058	−0.174*	0.077
Government partisan	−2.601***	0.074	−2.524***	0.070	−2.489***	0.114
Opposition partisan	−0.700***	0.083	1.314***	0.063	−0.367***	0.106
Perception of national economy	−0.672***	0.114	−0.966***	0.103	−1.089***	0.140
Perception of personal economy	−0.095	0.125	0.023	0.113	−0.125	0.159
Wealth quintile	−0.021	0.018	0.035*	0.015	−0.003	0.022
Unemployment	0.000	0.087	0.051	0.082	0.153	0.108
Decline in income	0.124*	0.053	0.092^†	0.049	0.171*	0.068
Bribe victim	−0.090	0.109	−0.073	0.097	0.146	0.117
Crime victim	−0.055	0.060	−0.020	0.053	−0.017	0.066
Effectiveness of current administration	−2.148***	0.121	−2.802***	0.111	−2.308***	0.150
Support for state economic intervention	−0.421***	0.121	−0.022	0.119	−0.196	0.148
Educational level	−0.041	0.035	0.168***	0.031	0.048	0.045
Woman	0.096*	0.041	0.036	0.038	0.074	0.053
Size of place of residence	0.018	0.018	0.003	0.017	0.015	0.022
Age	−0.041***	0.007	−0.006	0.006	−0.012	0.009
Age squared	0.000***	0.000	0.000	0.000	0.000	0.000
Internal efficacy	−0.251**	0.079	0.038	0.072	−0.181^†	0.103
System support	−0.868***	0.138	−0.067	0.119	−0.924***	0.161
Support for democracy	−0.381***	0.081	0.079	0.074	−0.157	0.102
Community participation	−0.716***	0.111	−0.134	0.095	−0.147	0.138
Constant	3.268***	0.253	1.993***	0.226	1.577***	0.309
Number of observations	25,789
F	65.25

Baseline response is voting for the incumbent. Country fixed effects are included in analysis but omitted from the table for ease of presentation.

Coefficients are statistically significant at ^†p < .10. *p < .05. **p < .01. ***p < .001.

The main findings from Table 2 are easier to comprehend using predicted probabilities (Figure 1). Holding all other variables at their within-country means, we estimate that the change from non-recipient to assistance recipient is associated with a 2.5 percentage point decrease in abstention, a 3.6 percentage point increase in voting for the incumbent, a 0.5 percentage point decrease in the opposition vote, and a 0.6 percentage point decrease in null voting.¹⁹ Thus, the net effect of assistance across the region is to mobilize voters, who then vote for the incumbent.

Figure 1.

Predicted probabilities of hypothetical electoral outcomes by beneficiary status, Latin America and the Caribbean, 2012 AmericasBarometer.

These results are consistent with the argument that social assistance plays an important part in the evaluation of the costs of going to the polls as well as recipients’ candidate choices. Although the changes in the region-wide model are relatively small, they are enough to guarantee incumbents a slight electoral advantage over opposition parties, on average, and across the region. Even if beneficiaries constitute a small proportion of the electorate, this small benefit for the incumbent could tip a close election. Nonetheless, the results also demonstrate clearly that large effects found in some case studies are unlikely to be constant across the region.

Thus, the strength of the association may vary; we suggest that this variation is systematic and depends on the institutional features of elections and assistance policies in each country. In Figure 2, we develop a preliminary test of this argument by examining whether effects vary across different clusters of countries, as defined by indicator variables for enforced compulsory voting, program conditionality, ideologically leftist president, and assistance program independence relative to the executive office. The figure presents the marginal effect of social assistance receipt on each potential outcome (abstention, null/blank voting, voting for the incumbent, or voting for the opposition), by country type.²⁰ Thus, positive values suggest that recipients are more likely than non-recipients to choose the given option under the country-level condition listed at the left of the figure; conversely, negative values suggest that recipients are less likely than non-recipients to choose that option.

Figure 2.

Changes in predicted probabilities of hypothetical electoral outcomes between beneficiaries and non-beneficiaries, by country type.

Differences in marginal effects’ statistical significance across the different groups of countries presented in Figure 2 must be read tentatively, as we have not yet formally tested these differences in the context of a hierarchical model. Nonetheless, the results strongly suggest that political, institutional, and programmatic contexts alter how we understand the electoral effect of social assistance. In countries with presidential control over assistance, the predicted probability of voting for the incumbent is 10.7 percentage points higher among assistance recipients than among non-recipients, compared with a mere 0.9 percentage point boost in incumbent voting when an institutionalized agency oversees the program; the confidence intervals of the marginal effects strongly suggest that these differences are statistically significant. And while social assistance leads to similarly sized increases in turnout (i.e., drops in abstention) in the two types of countries, only in ones with presidential control over the program is assistance receipt associated with a statistically significant (5 percentage point) drop in support for the opposition. These results are consistent with our approach of bounded rationality, in that recipients more strongly favor the incumbent when the incumbent’s role in benefits provision is clearer. At the same time, it remains to be determined whether the mechanism is retrospective or prospective; recipients might simply be more likely to recognize a job well done under these circumstances, or they might reason that their benefits are in greater danger of discontinuation.

By contrast, Figure 2 indicates that the ideological orientation of the president may be relatively unimportant. While the marginal effects of social assistance on abstention and incumbent voting are statistically significant at p < .05 in countries with leftist presidents but only at p < .10 in ones with centrist/rightist presidents, these effects are very similar in magnitude and statistically indistinguishable. Thus, rightist incumbents can benefit equally with leftist ones from social assistance. This hints that the electoral gains reaped from social assistance are not a result of ideologically driven or issue voting among assistance recipients for whom social assistance programs are especially salient. Rather, they once again suggest that recipients reward incumbents for programs’ direct impacts on the family pocketbook.

Figure 2 also reveals how enforcement of both social assistance conditionalities and compulsory voting rules shapes (or not) the utility calculations of boundedly rational voters who receive social assistance. First, the marginal effects of social assistance on abstention and incumbent support are somewhat larger in magnitude in countries with enforced conditionalities, yet they are not statistically distinguishable from those found in countries with no or unenforced conditionalities. This non-finding may mean that across the region, benefit recipients simply do not inherently value the public health and citizenship-related conditions that accompany CCT programs. Alternatively, it is consistent with a pattern in which beneficiaries very much value program conditions in some countries and dislike them in others.

At the same time, though, Figure 2 shows that enforced compulsory voting rules strongly shape the boundedly rational utility calculations of social assistance recipients considering what to do on election day. First, social assistance has a somewhat (though not significantly) larger effect on turnout in voluntary voting countries (3.0 percentage points vs. 1.5 in compulsory systems). This suggests that the approximately 11% of the electorate for whom the expected utility of going to the polls turns positive under compulsory voting rules may overlap to some extent with the group mobilized by social assistance. At the same time, even in compulsory voting countries, some citizens may remain who do not respond to compulsory voting rules alone, but who do respond when voting costs are reduced through social assistance programs. In turn, the large electorate mobilized by compulsory voting weighs assistance benefits particularly strongly in their vote decisions, perhaps because they lack other strong community, partisan, and ideological ties to the political system. While the marginal effect of social assistance on candidate choice is statistically insignificant in voluntary voting countries, under compulsory voting rules social assistance is associated with a highly statistically significant 6.0 percentage point advantage for the incumbent and a 4.2 percentage point deficit for the opposition.

To further assess the statistical significance of the differences in effects across various groups of countries, in Table 3 we present the key results of four hierarchical multinomial logit models, each testing a country-level moderator of the relationship between social assistance and electoral behavior across 19 countries.²¹ First, though, we examine the direct association between vote choice and the country-level variables. In line with common sense and previous studies, compulsory systems are associated with lower abstention (11 percentage points in our data, p < .05 with 24 country cases) and higher blank voting (6 percentage points, p = .06). In addition, interestingly, countries with leftist presidents may have a higher turnout (9 percentage points, p = .09). Apart from these results, none of the other associations between country-level variables and the vote is statistically significant at the country level.

Table 3.

Hierarchical Multinomial Logistic Regression Models: Country-Level Determinants of the Relationship between Social Assistance and the Vote.

	Abstain		Vote for opposition		Blank vote
	Coefficient	SE	Coefficient	SE	Coefficient	SE
Model 1
Received public assistance	−0.171*	0.078	−0.039	0.065	0.020	0.083
Presidential control	0.296	0.564	0.213	0.322	0.186	0.459
Beneficiary × Presidential control	−0.452***	0.151	−0.475***	0.127	−0.548**	0.172
Random effects
SD	1.179		0.690		0.950
Variance component	1.391		0.448		0.903
Chi-square	2,031.549		999.627		878.741
Model 2
Received public assistance	−0.282*	0.112	−0.121	0.099	−0.381*	0.150
Leftist president	−0.895	0.522	−0.371	0.312	−0.277	0.456
Beneficiary × Leftist president	−0.024	0.139	−0.065	0.119	0.342*	0.171
Random effects
SD	1.093		0.650		0.947
Variance component	1.196		0.422		0.897
Chi-square	1,893.045***		991.436***		941.757***
Model 3
Received public assistance	−0.279*	0.109	−0.151^†	0.087	0.079	0.105
Conditionality enforced	−0.314	0.548	−0.436	0.300	−0.263	0.442
Beneficiary × Conditionality	−0.022	0.137	−0.006	0.113	−0.408**	0.147
Random effects
SD	1.174		0.637		0.936
Variance component	1.378		0.406		0.876
Chi-square	1,991.911***		890.334***		773.129***
Model 4
Received public assistance	−0.156	0.100	0.126	0.092	−0.142	0.142
Compulsory voting	−0.833	0.511	−0.169	0.315	0.212	0.448
Beneficiary × Compulsory voting	−0.209	0.135	−0.462***	0.116	−0.003	0.165
Random effects
SD	1.094		0.670		0.950
Variance component	1.196		0.459		0.903
Chi-square	1,920.018***		1,033.154***		876.755***

Random effects are estimated at the country level. Baseline response is voting for the incumbent. Controls for partisanship, perception of the personal/national economy, wealth, unemployment, income decline, victimization by bribery and corruption, ratings of the current administration, education, location of residence, age, efficacy, system support, support for democracy, and community participation are included in the analysis but omitted from the results (see full cross-level interacted model in the online appendix). All models include 21,468 Level 1 observations and 19 Level 2 units.

Coefficients are statistically significant at ^†p < .10. *p < .05. **p < .01. ***p < .001.

Results from the hierarchical models broadly confirm the country-by-country analysis; predicted probabilities (see the online appendix) are similar in magnitude and nearly identical in statistical significance.²² First, executive ideology and enforcement of program conditions significantly condition social assistance’s impact on voter behavior only for blank voting. Thus, conditionalities may effectively incentivize some types of social participation, but they appear not to shape electoral behavior.

Estimating the joint effects of social assistance and its interaction terms, we find that social assistance decreases the propensity to vote blank when the president is a centrist or a rightist, and when conditions are enforced.²³ However, presidential program control strongly and significantly conditions each of the three relationships. When the executive office administers the country’s major assistance program, recipients are less likely than non-recipients to abstain, to vote for the opposition, and to vote blank, compared with voting for the incumbent. By contrast, in countries where assistance is administered by an institutionalized agency, benefit receipt has a small association with turnout, but not with support for any particular candidate. Compulsory voting also strongly shapes the relationship between assistance and voting for the opposition; only in compulsory voting countries are recipients less likely to vote for the opposition.

Discussion and Conclusion

These results provide region-wide evidence that social assistance programs incorporate and anchor marginalized citizens in their countries’ electoral processes. While there is not, as yet, systematic evidence that programs help beneficiaries become more cognizant citizens, they do help these beneficiaries overcome obstacles to voting and give them a tangible economic stake in political outcomes. Not only does added household income help citizens bear the immediate and opportunity costs of going to the polls, but assistance programs also shape the expected benefits of voting—both expressive ones and the utility attached to an incumbent victory. These dual (though on average modest) impacts on turnout and vote choice make assistance an important electoral tool for incumbents. Still, the value of this policy tool depends to a great degree on institutional features. Compulsory voting and program control by the executive office, in particular, alter the effects of social assistance benefits in favor of the incumbent or their party. Conversely, in countries with voluntary voting and agency program control, benefits may have small effects on turnout, yet they appear not to shape citizens’ choices once in the voting booth.

With these results, we begin to make out a portrait of the “social assistance voting base” across Latin America and the Caribbean. This is a boundedly rational electorate that is responsive to the perceived costs and benefits of voting. Moreover, the fact that ideologically rightist presidents benefit equally with leftist ones, and that effects occur in countries with both conditional and unconditional assistance programs, suggests that recipients are swayed not so much by rhetorical issue ownership or by broader concerns about citizenship, as by concrete, implemented policies that directly affect their individual pocketbook economics. Thus, our results have implications for the literature on policy responsiveness, indicating that social assistance recipients in Latin America respond electorally to this particular public policy in direct and particularistic ways and that electoral and policy institutions can constrain the ability of incumbents to derive direct electoral benefit from such programs.

But two puzzles remain, defining a long-term agenda for research. First, is the vote retrospective or prospective? Our results and the perspective of boundedly rational pocketbook voting are consistent with two alternative stories: that recipients simply reward incumbents who have helped them or that recipients support incumbents out of fear the opposition might discontinue benefits. This question has normative and practical consequences. If the social assistance electoral effect is due to fear of program discontinuation, beneficiaries might be highly responsive to incumbent manipulation in the short term, yet in the long-term, the effect would fade as recipients discovered that programs instituted under previous administrations continued to provide benefits under opposition rule. Without information on perceived threat of program discontinuation, it is not entirely possible to solve this puzzle. Nonetheless, we note that it has become common for opposition candidates of all ideological stripes to publicize their commitment to social assistance. This observation and the fact that we find electoral effects among incumbents on the right and left suggest to us that prospective threat or risk is not driving the social assistance electoral effect.

The retrospective economic voting assumption underlying most studies of electoral behavior posits the electorate as, in V. O. Key’s (1964) famous phrase, a “rational god of vengeance and reward” (p. 568). Recent work argues that the broader Latin American voting public is strongly retrospective and sociotropic (e.g., Nadeau, Lewis-Beck, & Bélanger, 2013; Singer & Carlin, 2013). Here, we uncover evidence to support the perspective that even Latin America’s poorest and most marginalized citizens hold their politicians responsible for economic outcomes, albeit pocketbook ones. In the hands of (also rational) elected leaders, this pocketbook rationality can be manipulated, especially in certain institutional settings. Elites can proactively and strategically adopt relatively cost-efficient programs that strongly reinforce their voting base, and they can manipulate institutional design to claim credit more effectively. Still, manipulation has limits; variables such as national economic evaluations and perceptions of government effectiveness also shape electoral outcomes. As important as they are in beneficiaries’ utility calculations, social assistance provides no sure guarantees of incumbent electoral success.

The second remaining puzzle is related to the first: In what way does presidential control over assistance programs make it possible for executives to claim credit more effectively? When the executive office directly controls the program, the president may have greater capacity to strategically allocate benefits in relation to the electoral calendar, program marketing materials may be more likely to carry personalistic statements clearly identifying the president with the recipients’ benefits, or program administration may be less reliant on lower-level political actors who muddy attributions of responsibility to the president. Whatever the specific mechanism, the pattern we report may eventually break down because there may be a political equilibrium point beyond which further expansion of and investment in social assistance become politically untenable or unjustifiable. Thus, the findings presented here may reflect effects only observable at this particular point in time in the development of social assistance programs in Latin America. Beneficiaries may also eventually incur social costs, such as becoming the target of negative stereotypes. Evidence of such patterns has yet to emerge from the Latin American assistance programs, but it may only be a matter of time.

More broadly, our results have implications for scholars interested in the ability of the poor to hold elected officials accountable for policy decisions. Our results suggest that targeted social policies entail a trade-off between two normatively desirable political objectives. Boundedly rational voters who receive social assistance are more likely to link pocketbook concerns to voting behavior when it is easier to attribute responsibility to elected officials; this is most likely to happen when presidents control programs more closely. Yet, such control also makes it possible for executives to abuse programs for their own or their party’s electoral benefit if they choose. Thus, unless one can rely on the virtue of elected leaders, it seems that these targeted programs place programmatic politics and accountability-based politics at odds with each other. Ultimately, we suggest that policy makers will need to take into account complex electoral and representational as well as human development impacts when they design or reform social assistance programs.

Footnotes

Acknowledgements

We thank the editors and anonymous reviewers of Comparative Political Studies, as well as Mariana Medina, Dave Peterson, Mitch Seligson, Michelle Taylor-Robinson, Robert Urbatsch, Lee Walker, Liz Zechmeister, and participants in a Vanderbilt University workshop for very helpful comments. We thank the Latin American Public Opinion Project (LAPOP) and its major supporters (the U.S. Agency for International Development, the United Nations Development Program, the Inter-American Development Bank, and Vanderbilt University) for making the data available. The data can be found at .

Authors’ Note

This is a revised version of a paper presented at the Annual Meeting of the American Political Science Association, Seattle, WA, September 2, 2011, and the Annual Meeting of the Southern Political Science Association, January 9, 2014. A preliminary version of the paper also appeared as No. 66 in the AmericasBarometer Insights series on October 10, 2011. Data Replication Materials are available at

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

Matthew L. Layton is a PhD Candidate in the Department of Political Science at Vanderbilt University. His research interests include the study of comparative political behavior, social policy, and the political effects of government conditional cash transfer (CCT) policies in Latin America.

Amy Erica Smith is assistant professor of Political Science at Iowa State University. Her research focuses on comparative political behavior and public opinion, with particular interests in social influences and in religion and politics, and in democratization and civil society in Latin America and Brazil.

References

Adato

Hoddinott

(Eds.). (2010). Conditional cash transfers in Latin America. Baltimore, MD: Johns Hopkins University Press.

Akhmedov

Zhuravskaya

(2004). “Opportunistic political cycles: Test in a young democracy setting.” The Quarterly Journal of Economics, 119, 1301-1338.

Alt

J. E.

Rose

S. S.

(2007). Context-conditional political budget cycles. In Boix

Stokes

S. C.

(Eds.), The Oxford handbook of comparative politics (pp. 845-867). New York, NY: Oxford University Press.

Ames

(1987). Political survival: Politicians and public policy in Latin America. Berkeley: University of California Press.

Baez

J. E.

Camacho

Conover

Zárate

R. D.

(2012). Conditional cash transfers, political participation, and voting behavior (SSRN Scholarly Paper No. ID 2156795). Rochester, NY: Social Science Research Network. Retrieved from http://papers.ssrn.com/abstract=2156795

Barrientos

Santibáñez

(2009). New forms of social assistance and the evolution of social protection in Latin America. Journal of Latin American Studies, 41, 1-26.

Bélanger

É.

Meguid

B. M.

(2008). Issue salience, issue ownership, and issue-based vote choice. Electoral Studies, 27, 477-491.

Brazil’s Bolsa Família: How to get children out of jobs and into school. (2010, July 31). The Economist. Retrieved from http://www.economist.com/node/16690887

Bruhn

(1996). Social spending and political support: The “lessons” of the National Solidarity Program in Mexico. Comparative Politics, 28, 151-177.

10.

Brusco

Nazareno

Stokes

S. C.

(2004). Vote buying in Argentina. Latin American Research Review, 39, 66-88.

11.

Calvo

Murillo

M. V.

(2004). Who delivers? Partisan clients in the Argentine electoral market. American Journal of Political Science, 48, 742-757.

12.

Campbell

A. L.

(2003). How policies make citizens: Senior political activism and the American welfare state. Princeton, NJ: Princeton University Press.

13.

Campbell

A. L.

(2012). Policy makes mass politics. Annual Review of Political Science, 15, 333-351. doi:10.1146/annurev-polisci-012610-135202

14.

Cecchini

Madariaga

(2011). Conditional cash transfer programmes: The recent experience in Latin America and the Caribbean. Santiago, Chile: United Nations.

15.

Córdova

(2009, January 7). Methodological note: Measure relative wealth using household asset indicators. Nashville, TN: Latin American Public Opinion Project.

16.

De La O

A. L.

(2013). Do conditional cash transfers affect electoral behavior? Evidence from a randomized experiment in Mexico. American Journal of Political Science, 57, 1-14. doi:10.1111/j.1540-5907.2012.00617.x

17.

Diaz-Cayeros

(2008). Electoral risk and redistributive politics in Mexico and the United States. Studies in Comparative International Development, 43, 129-150. doi:10.1007/s12116-008-9020-1

18.

Feldman

(1995). Answering survey questions: The measurement and meaning of public opinion. In Lodge

McGraw

K. M.

(Eds.), Political judgment: Structure and process (pp. 249-270). Ann Arbor: University of Michigan Press.

19.

Fiszbein

Schady

(2009). Conditional cash transfers: Reducing present and future poverty. Washington, DC: The World Bank. Retrieved from http://siteresources.worldbank.org/INTCCT/Resources/5757608-1234228266004/PRR-CCT_web_noembargo.pdf

20.

Franzese

R. J.

(2002). Electoral and partisan cycles in economic policies and outcomes. Annual Review of Political Science, 5, 369-421. doi:10.1146/annurev.polisci.5.112801.080924

21.

Gomez

B. T.

Wilson

J. M.

(2006). Cognitive heterogeneity and economic voting: A comparative analysis of four democratic electorates. American Journal of Political Science, 50, 127-145.

22.

Graham

(1994). Safety nets, politics, and the poor: Transitions to market economies. Washington, DC: Brookings Institution.

23.

Graham

Kane

(1998). Opportunistic government or sustaining reform? Electoral trends and public-expenditure patterns in Peru, 1990-1995. Latin American Research Review, 33, 67-104.

24.

Hevia

F. J.

(2011). Relaciones sociedad-estado, participación ciudadana y clientelismo político en programas contra la pobreza. El caso de Bolsa Familia en Brasil [State-Society Relations, Citizen Participation, and Political Clientelism in Anti-Poverty Programs: The Case of Brazil’s Bolsa Família]. América Latina Hoy, 57, 205-238.

25.

Hicken

(2011). Clientelism. Annual Review of Political Science, 14, 289-310. doi:10.1146/annurev.polisci.031908.220508

26.

Hunter

Sugiyama

N. B.

(2009). Democracy and social policy in Brazil: Advancing basic needs, preserving privileged interests. Latin American Politics and Society, 51(2), 29-58. doi:10.1111/j.1548-2456.2009.00047.x

27.

Hunter

Sugiyama

N. B.

(2011, September). Documenting citizenship: Contemporary efforts toward social inclusion in Brazil. Paper presented in the 2011 APSA annual meeting, Seattle, WA.

28.

Kahneman

Tversky

(1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263-291.

29.

Key

V. O.

(1964). Politics, Parties, & Pressure Groups. New York, NY: Crowell.

30.

Lacy

Burden

B. C.

(1999). The vote-stealing and turnout effects of Ross Perot in the 1992 U.S. presidential election. American Journal of Political Science, 43, 233-255. doi:10.2307/2991792

31.

Lau

R. R.

Redlawsk

D. P.

(2001). Advantages and disadvantages of cognitive heuristics in political decision making. American Journal of Political Science, 45, 951-971.

32.

Levy

J. S.

(2003). Applications of prospect theory to political science. Synthese, 135, 215-241.

33.

Lewis

D. E.

(2003). Presidents and the politics of agency design: Political insulation in the United States government bureaucracy, 1946-1997. Stanford, CA: Stanford University Press.

34.

Licio

E. C.

Rennó

L. R.

Castro

H. C.

(2009). Bolsa Família e voto na eleição presidencial de 2006: Em busca do elo perdido [Bolsa Família and the Vote in the 2006 Presidential Election: In Search of the Missing Link]. Opinião Pública, 15, 31-54.

35.

Linos

(2013). Do conditional cash transfer programs shift votes? Evidence from the Honduran PRAF. Electoral Studies, 32, 864-874.

36.

Manacorda

Miguel

Vigorito

(2011). Government transfers and political support. American Economic Journal: Applied Economics, 3(3), 1-28.

37.

Mettler

(2011). The submergerd state: How invisible government policies undermine American democracy. Chicago, IL: University of Chicago Press.

38.

Moore

(2010). The political economy of social protection in Honduras and Nicaragua. In Adato

Hoddinott

(Eds.), Conditional cash transfers in Latin America (pp. 101-126). Baltimore, MD: Johns Hopkins University Press.

39.

Nadeau

Lewis-Beck

M. S.

Bélanger

. (2013). Economics and elections revisited. Comparative Political Studies, 46, 551-573. doi:10.1177/0010414012463877

40.

Panagopoulos

(2008). The calculus of voting in compulsory voting systems. Political Behavior, 30, 455-467. doi:10.1007/s11109-008-9058-9

41.

Penfold-Becerra

(2007). Clientelism and social funds: Evidence from Chávez’s Misiones. Latin American Politics and Society, 49(4), 63-84.

42.

Powers

N. R.

(2001). Grassroots expectations of democracy and economy: Argentina in comparative perspective. Pittsburgh, PA: University of Pittsburgh Press.

43.

Preacher

K. J.

Curran

P. J.

Bauer

D. J.

(2006). Computational tools for probing interactions in multiple linear regression, multilevel modeling, and latent curve analysis. Journal of Educational and Behavioral Statistics, 31, 437-448.

44.

Rocha-Menocal

(2001). Do old habits die hard? A statistical exploration of the politicisation of Progresa, Mexico’s latest federal poverty-alleviation programme, under the Zedillo administration. Journal of Latin American Studies, 33, 513-538.

45.

Schady

(2000). The political economy of expenditures by the Peruvian social fund (FONCODES), 1991-95. The American Political Science Review, 94, 289-304. doi:10.2307/2586013

46.

Schneider

S.K.

Jacoby

W.G.

2003. A culture of dependence? The relationship between public assistance and public opinion. British Journal of Political Science, 33(2), 213-231.

47.

Sewall

R. G.

(2008). Conditional cash transfer programs in Latin America. SAIS Review, 28, 175-187.

48.

Simon

H. A.

(1985). Human nature in politics: The dialogue of psychology with political science. The American Political Science Review, 79, 293-304.

49.

Simon

H. A.

(1995). Rationality in political behavior. Political Psychology, 16, 45-61.

50.

Singer

M. M.

Carlin

R. E.

(2013). Context counts: The election cycle, development, and the nature of economic voting. Journal of Politics, 75, 730-742.

51.

Stokes

S. C.

(2001). Mandates and democracy: Neoliberalism by surprise in Latin America. New York, NY: Cambridge University Press.

52.

Sugiyama

N. B.

Hunter

(2013). Whither clientelism? Good governance and Brazil’s Bolsa Família program. Comparative Politics, 46, 43-62. doi:10.5129/001041513807709365

53.

Taylor-Robinson

(2010). Do the poor count? Democratic institutions and accountability in a context of poverty. University Park: Pennsylvania State University Press.

54.

Tufte

E. R.

(1978). Political control of the economy. Princeton, NJ: Princeton University Press.

55.

Zaller

J. R.

(1992). The nature and origins of mass opinion. New York, NY: Cambridge University Press.

56.

Zucco

(2013). When payouts pay off: Conditional cash transfers and voting behavior in Brazil 2002–10. American Journal of Political Science, 57, 810-822. doi:10.1111/ajps.12026

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.12 MB