“Take a photo of [your] vote for me because if we lose, you will not be able to take the $10,000 [pesos] anymore because I am not going to use my resources anymore for social and political aid” (Clarín 2021). This message, which the Peronist mayor of the Argentine town of Santa Clara de Buena Vista sent in November 2021 to a potential supporter, contextualizes an entrenched practice in Latin American politics: vote buying (Corral and Martínez Rosón Reference Corral, Martínez Rosón and Alcántara2020; Johnson Reference Johnson2020).
Vote buying can be defined as the act by politicians to offer money or other quantifiable goods for votes at election time (Nichter Reference Nichter2014; Schaffer and Baker Reference Schaffer and Baker2015). As a political phenomenon, vote buying has manifested itself worldwide, but it is particularly pervasive in societies where poverty and income inequalities are endemic and long-lasting (Jensen and Justesen Reference Jensen and Justesen2014; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013).
Although vote buying occurs only at election time, it can analytically be classified as a type of clientelism, since it reinforces traditional clientelistic practices, which are employed in between electoral cycles (Kitschelt Reference Kitschelt2000). Indeed, vote buying may have pernicious effects on political and socioeconomic development at multiple levels (Stokes Reference Stokes2007). First, it violates the basic democratic principle of allowing people to exercise their right to vote and express their choices autonomously (Mares Reference Mares2015). Second, it undermines the fairness of the electoral process because those who have the financial means to buy votes can potentially defeat their competitors and, consequently, delegitimize election outcomes.
Third, vote buying may allow the election of incompetent or corrupt leaders who use public office to aggrandize their personal interest, that of their cronies, and on occasions, that of organized crime (Gambetta Reference Gambetta1993). Conversely, it diminishes the funds available for the universal provision of public goods (Baland and Robinson Reference Baland and Robinson2007). Fourth, in cases where party machines can monitor voters, it sets in motion what Stokes (Reference Stokes2005) defines as “perverse accountability,” where citizens, not politicians, are held to account for their actions. Indeed, as Schaffer (Reference Schaffer2007, 10) points out, “purchased delegation” may justify politicians’ assumption that they have an “unconstrained” mandate to act as they see fit. Furthermore, where such a practice is entrenched, it may lead to political alienation and distrust (Carreras and Irepoglu Reference Carreras and Irepoglu2013; Carlin and Mosley Reference Carlin and Moseley2015). Therefore, vote buying violates the basic tenets of democratic government and corrodes its very foundations. Consequently, understanding its causes and limiting its effects has important implications in both theory and practice.
To understand the causes of vote buying, we need to explain what type of voters clientelistic parties target. In the current literature, two factors are systematically emphasized as the key drivers of vote buying (Mares and Young Reference Mares and Young2016; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013): poverty and partisanship. The best-established finding to date is arguably that poor voters are the key targets of vote buying during election campaigns (Jensen and Justesen Reference Jensen and Justesen2014; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013).
An unresolved puzzle that emerges from the literature is who among the poor become clients and why. Indeed, we have little knowledge about why this is so. In terms of partisanship, a large amount of the scholarly debate has centered on whether parties target “core” or “swing” voters—or both (Nichter Reference Nichter2008; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013), and how effective parties are in monitoring voters (Stokes Reference Stokes2005; Szwarcberg Reference Szwarcberg2015), as well as their brokers (Larreguy et al. Reference Larreguy, Marshall and Querubin2016; Novaes Reference Novaes2018). To date, however, no consensus has emerged on these issues, in part because most analyses are based on single-country or small-N studies. Moreover, the current literature does not offer a clear picture of what role, if any, partisanship plays in the link between poverty and vote buying. That is, few studies have considered how partisanship may moderate the link between poverty and parties’ attempt to mobilize support through vote buying.
This article addresses this gap in the literature and makes two contributions to our understanding of vote buying. First, it presents new, cross-nationally based evidence that suggests that party machines employ mixed vote-buying strategies to maximize their electoral chances (compare Gans-Morse et al. Reference Gans-Morse, Mazzuca and Nichter2014). Parties use vote buying to maintain the support of their core voters (partisans), but also to target those independent voters who are most responsive to material incentives: poor swing voters. To begin with, the fundamental problem for parties pursuing electoral mobilization using vote buying is that—under the secret ballot—it is difficult to monitor voter compliance with commitments to vote as promised. Therefore, parties must direct the resources used on vote buying toward voters they believe are more likely to comply. The safest bet for parties is to use preelection goods to mobilize core supporters during elections—what Nichter (Reference Nichter2008, Reference Nichter2014) calls turnout buying. Targeting partisan voters means that parties need not worry about monitoring compliance—because partisans rarely vote for other parties—but simply need to ensure that they turn out to vote (Nichter Reference Nichter2008).
In this case, the main requirement for receiving preelection goods and benefits is whether the voter is a partisan or not, meaning that income should matter less for parties’ vote- (or turnout-) buying efforts. However, to win elections, parties also need the support of at least some swing voters—particularly as this is often the numerically largest group in the electorate. To counteract the monitoring problem, parties need to target those swing voters they believe are more likely to comply with the vote-buying transaction. Party machines may therefore be more likely to target poor voters who have no strong partisan attachments—that is, poor swing voters. Swing voters, by definition, have weak or no ties to a particular party. If swing voters are also poor, they may be easier to sway using material incentives. If so, the effect of poverty on vote buying should be moderated by partisanship.
This article’s second contribution is empirical: it uses data based on individual-level surveys from 22 countries in Latin America to examine how poverty is related to vote buying among groups of swing voters and partisans. To the best of our knowledge, this constitutes the most comprehensive comparative and cross-national analysis of the relationship between poverty and vote buying in the Latin American context—and the potential role that partisanship plays for this relationship. The results show that both poverty and partisanship matter for vote buying. The results also provide some suggestive evidence that poverty matters mainly for swing voters, although these results are conditional on the measure of poverty that we employ.
The remainder of paper is organized as follows. The following section surveys the existing literature on poverty, partisanship, and vote buying, with a particular focus on Latin America. The subsequent section develops the argument about how poverty and partisanship may jointly affect vote-buying strategies. The data and econometric model are described, and the empirical findings are discussed. The concluding section sums up the main findings and the article’s contribution to the ongoing debate on the relationship between poverty and vote buying.
Related Literature
At the heart of the study of vote buying is the issue of what types of voters clientelistic parties target and how their machines monitor whether voters honor their part of the deal. To explain these issues, much of the existing literature points to either poverty or partisanship as the key driver of vote buying and selling (Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013; Mares and Young Reference Mares and Young2016).
The role of poverty was emphasized in the seminal contribution by Scott (Reference Scott1969), who—on the basis of studies of postcolonial societies—argued that clientelism thrives in conditions of poverty and that political machines rely on political support from the masses of the poor. Indeed, poverty is today widely believed to constitute the root cause of vote buying in new democracies around the world. The nexus between vote buying and poverty is well established in Africa (Jensen and Justesen Reference Jensen and Justesen2014; Vicente and Wantchekon Reference Vicente and Wantchekon2009; Bratton Reference Bratton2008). In Latin America, too, accounts of how party machines mobilize the votes of poor citizens by dispensing particularistic benefits abound. This is documented in the case of Argentina (Auyero Reference Auyero2001; Calvo and Murillo Reference Calvo and Victoria Murillo2004; Dinatale Reference Dinatale2004; Stokes Reference Stokes2005; Nichter Reference Nichter2008; Weitz-Shapiro Reference Weitz-Shapiro2012; Murillo et al. Reference Murillo, Oliveros and Zarazaga2021), Brazil (Zucco Reference Zucco2013), Colombia (Rueda Reference Rueda2017), Guatemala (González-Ocantos et al. Reference González-Ocantos, Chad Kiewiet de Jonge, Nickerson and Osorio2020); Mexico (Fox Reference Fox1994; Magaloni Reference Magaloni2006; Díaz-Cayeros et al. 2015; Castro Cornejo and Beltrán Reference Castro Cornejo and Beltrán2020), Paraguay (Finan and Schechter Reference Finan and Schechter2012), Peru (Schady Reference Schady2000; Gans-Morse et al. Reference Gans-Morse, Mazzuca and Nichter2014), and Venezuela (Kornblith Reference Kornblith2002; Penfold-Becerra Reference Penfold-Becerra2008).
Although the lion’s share of the literature on vote buying in Latin America consists of single-country studies, in recent years interest has increased in tackling the question from a comparative approach (Faughnan and Zechmeister Reference Faughnan and Zechmeister2011; González-Ocantos et al. Reference González-Ocantos, Kiewiet de Jonge and Nickerson2014; Schaffer and Baker Reference Schaffer and Baker2015; Calvo and Murillo Reference Calvo and Victoria Murillo2019; González-Ocantos and Oliveros Reference González-Ocantos and Oliveros2019). More important, even though it is almost universally accepted that poverty is a significant correlate of vote buying and electoral clientelism (Jensen and Justesen Reference Jensen and Justesen2014; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013), the current literature provides divergent explanations about why not all poor people are targets of vote buying and what conditions may strengthen or weaken the effect of poverty on vote buying.
One way to address these questions is to look into the literature on partisanship. Explanations of vote buying focusing on partisanship typically take their cue from the literature on distributional politics, which distinguishes two broad strategies that party machines use to mobilize voters: parties target either core (Cox and McCubbins Reference Cox and McCubbins1986) or swing voters (Lindbeck and Weibull Reference Lindbeck and Weibull1987; Dixit and Londregan Reference Dixit and Londregan1996). Cox and McCubbins (Reference Cox and McCubbins1986, 378–79) emphasize that parties tend to reward core voters because they are “well-known quantities” with a safer return to investment, whereas swing voters are “risker investments,” meaning that money spent on swing voters may be wasted in political terms. However, a common view in political science is that parties target swing voters (Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013, 31). This argument emphasizes that parties target swing voters because they are decisive and easier to sway, due to their lack of ideological and partisan commitment. Therefore, swing voters should be particularly responsive to electoral incentives and redistributive benefits (Dixit and Londregan Reference Dixit and Londregan1996; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013; González-Ocantos and Oliveros Reference González-Ocantos and Oliveros2019).
However, the link between poverty and vote buying is increasingly being challenged (González-Ocantos et al. Reference González-Ocantos, Chad Kiewiet de Jonge, Osorio and Nickerson2012). For instance, the “conditional party loyalty” model by Díaz Cayeros et al. (Reference Díaz-Cayeros, Magaloni and Estévez2016) shows that Mexican politicians invest in a much more sophisticated mix of both private and public goods tailored to different constituencies than previously assumed. Similarly, in their analysis of Argentina and Chile, Calvo and Murillo (Reference Calvo and Victoria Murillo2019) contend that party machines deliver a multiplicity of clientelistic and nonclientelistic policies according to the needs of specific constituencies, which are not limited to the poor.
In Latin America, evidence on vote buying and electoral clientelism targeted at party loyalists has been found in Mexico (De la O Reference De la O2013) and Argentina (Auyero Reference Auyero2001; Calvo and Murillo Reference Calvo and Victoria Murillo2004; Nichter Reference Nichter2008; Zarazaga Reference Zarazaga2014). However, other studies bring evidence to the swing voter model in the cases of Argentina (Stokes Reference Stokes2005), Peru (Schady Reference Schady2000), Honduras (Linos Reference Linos2013), and Mexico (Hiskey Reference Hiskey1999). Moreover, while some scholars contend that party machines in Argentina and Chile focus on politically and civically engaged voters (Faughnan and Zechmeister Reference Faughnan and Zechmeister2011), a more recent analysis of Argentina argues that party brokers avoid democratically minded citizens and concentrate on those who are either ambivalent or opposed to democracy (Carlin and Moseley Reference Carlin and Moseley2015).
Given these seemingly inconsistent results, a small number of studies theorize that clientelistic parties may simultaneously appease both loyalists and swing voters through different means and in varying degrees (Díaz-Cayeros et al. Reference Díaz-Cayeros, Magaloni, Estévez, Kitschelt and Steven2006; Albertus Reference Albertus2013). An example of this argument is the theory of electoral clientelism by Stokes et al. (Reference Stokes, Dunning, Nazareno and Brusco2013). Their model suggests that while political leaders tend to dispense resources to target competitive (swing) electoral districts, party brokers in those districts prefer to target loyal voters and spend less on swing voters.
A further twist to this debate centers on Stokes’s Reference Stokes2005 thesis that in Argentina the Peronist Party targets “weakly opposed” voters and also can monitor the candidates for whom people actually cast their ballot. However, her “monitoring” thesis runs counter to several studies examining clientelism in Argentina (Auyero Reference Auyero2001; Calvo and Murillo Reference Calvo and Victoria Murillo2004; Zarazaga Reference Zarazaga2014), while a comparative analysis of Latin America contends that monitoring strategies are unnecessary because citizens, particularly low-income ones, doubt the secrecy of the ballot in the first place (Kiewiet de Jonge and Nickerson Reference Kiewiet De Jonge and Nickerson2014). Nichter (Reference Nichter2008) also challenges Stokes’s thesis by arguing that the Peronist electoral machine focuses on soliciting the turnout of “unmobilized” supporters, rather than swing voters.Footnote 1 Following up on this study, Gans-Morse et al. (Reference Gans-Morse, Mazzuca and Nichter2014) have revisited the multipronged thesis by proposing a formal model according to which machines pursue four distinct clientelist strategies: vote buying, turnout buying, abstention buying, and double persuasion.
Poverty, Partisanship, and Vote Buying
Political parties in general do not necessarily pursue single-minded strategies focusing on either core supporters or swing voters (Albertus Reference Albertus2013: Carlin and Moseley Reference Carlin and Moseley2015; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013; Gans-Morse et al. Reference Gans-Morse, Mazzuca and Nichter2014). Instead, parties may hedge their bets to maximize chances of success by targeting both groups. Given this premise, it is possible that clientelistic parties use vote buying to mobilize electoral support among poor voters conditional on their partisanship. Parties’ vote-buying campaigns therefore may be guided by concerns to mobilize large groups of voters based on economic need (poverty), combined with an assessment of the strength and direction of those voter groups’ commitment to a particular party (partisanship).
Poverty as a Source of Vote Buying
We can think of poverty-induced vote buying as the result of two types of mechanisms, which constitute the foundation of the market for votes (Aidt and Jensen Reference Aidt and Sandhold Jensen2016; Jensen and Justesen Reference Jensen and Justesen2014). In this market, voters sell the commodity—votes—to political parties, whose purchase of votes is driven by concerns to maximize chances of getting elected to office. The vote-selling mechanism is fairly straightforward and follows the rationale of previous works by Kitschelt (Reference Kitschelt2000) and Keefer and Vlaicu (Reference Keefer and Vlaicu2008), at times referred to as the “discount rate” logic. It postulates that in a political system in which clientelistic politics is widespread, politicians lack the credibility to gain votes through programmatic campaign promises because voters do not believe that they will deliver once elected. Therefore, poor people take a minimum winning approach by opting to sell their vote before the election. This guarantees an immediate return of tangible benefits, as opposed to casting a ballot for programmatic policies that are unlikely to materialize in the future.
The vote-buying mechanism works by making it attractive for politicians to buy votes when the pool of poor citizens who are eligible to vote is large, because this lowers the per capita cost of exchanging money for votes. This relationship can be further reinforced if elections are contested and voters have little or no knowledge of what candidates stand for, or are unsure about the secrecy of their ballots and consequently may have doubts about the impact of their vote. In a scenario of this kind, preelectoral vote buying provides politicians with a more effective way of mobilizing large groups of poor voters, compared to campaigns based on promises of programmatic redistribution after the election.
While the arguments linking poverty to vote buying are both compelling and intuitively appealing, recent empirical work has challenged that link. For instance, findings from Nicaragua by González-Ocantos et al. (Reference González-Ocantos, Chad Kiewiet de Jonge, Osorio and Nickerson2012) show that there is no relationship between poverty and vote buying. This suggests that poverty may not matter after all for clientelism. But it also opens up the question of whether clientelist parties are more prone to target some groups of poor people with vote-buying campaigns.
The Role of Partisanship
To understand which groups of poor voters become targets of vote-buying campaigns, we explore whether party machines target voters on the basis of poverty (Jensen and Justesen Reference Jensen and Justesen2014; Scott Reference Scott1969), partisanship (Nichter Reference Nichter2008; Stokes Reference Stokes2005), or a mix of the two.
Targeting partisan supporters in order to mobilize turnout is a rational strategy when voter compliance is difficult to monitor, which is the case when the secret ballot is effective (Nichter Reference Nichter2008). Partisanship typically induces strong feelings of commitment to a particular party and shapes the way voters perceive the actions of political parties (Tilley and Hobolt Reference Tilley and Hobolt2011; Bartells Reference Bartells2002). Even for poor voters, partisan attachments may create a strong (dis)inclination to vote for a particular party. Voters’ feelings of partisanship may therefore affect the reservation price of a vote—that is, the price at which a voter is willing to sell their vote (Gans-More et al. Reference Gans-Morse, Mazzuca and Nichter2014). This means that if parties target strongly opposed voters, the cost of buying votes becomes prohibitive. In addition, strongly opposed voters are more likely to renege on their promises to vote as instructed once they are in the secrecy of the voting booth.
Even for poor voters, vote bribes are unlikely to be successful if those voters are strongly opposed partisans. Similarly, strong partisan supporters are inclined to vote for their favored party regardless of their income level. However, parties may still allocate significant resources to the distribution of gifts or money to partisan supporters with the aim of mobilizing turnout and maintaining the loyalty of core voters (Nichter 2014; Reference Nichter2008). Indeed, both party leaders and brokers have incentives to keep their rank-and-file voters loyal and content, since, in clientelistic systems, people have expectations that past experiences in accessing benefits will be fulfilled again in the future (Calvo and Murillo Reference Calvo and Victoria Murillo2013). Ignoring such expectations may put the party at risk if loyal supporters choose to abstain from voting on election day. Indeed, brokers in the field are interested in building networks of followers over time, not just during electoral contexts. Therefore, rewarding supporters is part of this calculation, even if the cost may be high (Auyero Reference Auyero2001; Szwarcberg Reference Szwarcberg2015).
However, to diminish uncertainty, parties must also broaden their electoral base, particularly in districts where elections are close and attracting swing voters is imperative (Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013). In this case, the higher the income of swing voters, the more expensive it will be to attract them. Therefore, as the price of a vote increases, party machines will be able to target fewer swing voters (Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013). Clientelistic parties may therefore choose to target voters who are poor and also have no strong partisan attachment—that is, poor swing voters.
From the perspective of clientelistic parties, poor swing voters possess the combination of being economically deprived—giving them incentives to sell their votes and making them cheap to buy—and lacking ingrained feelings of commitment to a particular party, making them flexible with regard to their party preferences. This group of poor swing voters may therefore be both more willing to exchange their vote for money or material benefits and less attached to a particular political party.
Note that we are not arguing that parties do not mainly target partisans through, for example, turnout buying (Nichter Reference Nichter2008). Indeed, it is very plausible that income and poverty may matter separately. What we explore is the additional possibility that poverty matters more for vote buying among swing voters and less for partisan supporters.
Data and Methods
The Latin American Public Opinion Project (LAPOP, https://www.vanderbilt.edu/lapop/) constitutes the most comprehensive source of cross-country survey data on vote buying in Latin America. We used the 2014 round of the LAPOP survey, which contains data from standardized questionnaires covering 26 countries in Latin America and the Caribbean, with data on vote buying available for 22 of those countries. Descriptive statistics for all variables used in the analyses are available in table A in the online appendix.
We combined data from these 22 country surveys into a dataset of individualinterviews (i) across countries (j). Interviews were conducted face to face using tablets, with questionnaires available in all the major languages of the region. Sampling was based on a stratified multistage procedure, in which stratification was based on region, urban or rural area, and municipality size. The sampling procedure generated a representative sample of the voting-age population in each country. The standard sample size was about 1,500 respondents, but in some countries (e.g., Bolivia) it increased to more than twice thatnumber. Therefore we employed cross-country survey weights designed to reflect a standard sample size of n = 1,500 for each country.
Dependent Variable: Vote Buying
To measure voters’ experience with being targets of vote-buying campaigns by political parties during elections, we used the following question: “Thinking about the last presidential elections of [YEAR], did someone offer you something, like a favor, gift, or any other benefit in return for your vote or support?” Respondents could answer yes or no. Accordingly, the dependent variable is binary. with 1 (yes) denoting those respondents who had been offered favors, gifts, or benefits in return for their vote, and 0 (no) indicating that respondents had not received such offers. This question allowed us to capture what we were interested in—the extent to which parties targeted particular groups with vote-buying offers in return for electoral support—for a comprehensive cross-section of countries in Latin America. The 2014 LAPOP data on vote buying are available for the 22 countries shown in figure 1.
Figure 1 shows the country-level percentage of the population that answered yes to being offered favors, gifts, or benefits in return for their vote. Overall, 8.2 percent of the population in the 22 countries acknowledged being approached with vote-buying offers by political parties during elections. However, figure 1 clearly demonstrates that there are huge differences across countries—ranging from almost no vote buying in Chile to a fairly widespread use of vote buying in Mexico and the Dominican Republic.
While the vote-buying question is useful for our purposes, caveats still apply. First, we cannot distinguish between different uses of electoral bribes to mobilize voters, such as turnout buying—paying voters to turn out on election day—or paying voters to change their vote choice in order to support a particular party (Nichter Reference Nichter2014; Nichter et al. Reference Nichter2014). Nor are we able to measure whether voters have been targeted by the incumbent party or by opposition parties. That is, we can measure only whether voters have been offered material benefits or favors in return for their votes.
Second, the vote-buying question directly asks respondents about their experiences with being offered electoral bribes by parties or party agents. Since vote buying and selling is illegal in most countries, direct vote-buying questions are often thought to give rise to social desirability bias in the sense that respondents have incentives to underreport their actual experiences (González-Ocantos et al. Reference González-Ocantos, Chad Kiewiet de Jonge, Osorio and Nickerson2012;). If so, the numbers in figure 1 may underestimate the true levels of vote buying in Latin America. However, the wording of the vote-buying question in the LAPOP data goes some way toward alleviating concerns that social desirability is a major issue. In particular, the questions ask respondents if they have been offered favors, gifts, or benefits by parties—not whether respondents have asked for benefits themselves. Asking the question in this way is an attempt to measure the extent to which parties use targeted vote-buying campaigns during elections—rather than trying to measure vote-selling efforts by voters.
Similarly, LAPOP’s vote-buying question does not ask respondents whether they accepted the offer, but simply whether someone approached them offering them a targeted good. That is, the question does not ask whether respondents have taken electoral bribes and sold their votes; nor does it ask whether respondents who engage in vote bargains comply with their commitments to vote as promised. In this way, the question places “responsibility” for the vote-buying act on parties rather than on voters, which should reduce tendencies toward social desirability bias.
Explanatory Variables: Poverty and Partisanship
The two key explanatory variables are Poverty and Partisanship. To measure poverty, we used two variables. The first is a measure of self-reported income that asks respondents to place the total monthly income of their household within an income interval. For each country, we transformed the original 17-category income variable into a 10-category percentile scale, with a minimum of 0 (the 10th percentile of the variable) and a maximum of 9 (the 90th percentile). This variable provides a measure of respondents’ self-reported location in the national income distribution, ranking from low to high. However, self-reported survey-based income measures are subject to reporting problems. For instance, people may be unwilling to reveal their income; they may not recollect their income; or they may under- or overestimate their income. We therefore also used a measure of people’s material living standards; that is, the extent to which they possessed things like a refrigerator, washing machine, indoor plumbing, and a bathroom in their household. Based on items in the survey, we created an index of people’s living standard.Footnote 2 The index was scaled from 0 to 1, with low values denoting low material living standards (poverty) and high values denoting high material living standards.Footnote 3
To measure partisanship, we constructed a binary indicator that simply distinguished partisan voters—people who identify with a particular party—from independent or swing voters—people who do not identify with a particular party. To this end, we used the question, “Do you currently identify with a political party?” Respondents who answered yes were coded as partisan supporters (1); respondents who answered no were coded as independent or swing voters (0).
Control Variables
Our regressions include a number of (respondent-level) control variables. While it is inherently hard to make causal inferences using nonexperimental (observational) data of the kind we employ, the inclusion of a substantial set of controls should, at minimum, help guard against spurious correlations. We generally refrained from including posttreatment variables (e.g., political interest) that could be considered “bad controls” and might depress the effect of the key explanatory variables of interest ().
First, we controlled for whether respondentswere recipients of conditional cash transfers (CCTs). This correlates with respondent poverty. Receiving CCTs could also signal to political parties that respondents are in need, and in that way help clientelist parties target groups of poor people. Second, we controlled for a series of standard socioeconomic variables. We included control variables for education (in years), gender (female = 1), age (in years), and respondent’s residence (urban or rural).Footnote 4 We accounted for country-level difference by including a full set of country fixed effects. This eliminates omitted-variable bias at the country level and focuses attention on within-country variation in the data. Accordingly, we ran the analyses using logit regressions with country fixed effects, in which the dependent variable takes on the value 1 if individual i in country j has received a vote-buying offer from a political party (and 0 otherwise).
Results
The results from the fixed effects regression model are shown in table 1. (Appendix table B shows results that reproduce table 1 but without the control for conditional cash transfer receipt, which, by definition, is correlated with poverty and may thereby lower the coefficient of the income and living conditions variables.)
*** p<0.01, ** p<0.05, * p<0.1.
Notes: Dependent variable is a binary indicator of vote buying. All results were obtained using logistic regression with country fixed effects. Coefficients are logged(odds). The data are weighted to reflect a standard sample size of n = 1,500. A chi2-test for the difference between the “income, percentile” coefficients in model 5 (swing voters) and model 7 (partisan voters) gives a value of 2.29 (0 = 0.13); a chi2-test for the difference between the “living standards” coefficients in model 6 (swing voters) and model 8 (partisan voters) gives a value of <0.00 (p = 0.99). Z-statistics in parentheses.
Models 1 and 2 include only the income and living conditions variables, along with partisanship and the CCT control. Swing voters constitute the reference category. The coefficient—expressed in logged odds—for the income variable shows that higher income decreases the chances of experiencing vote buying; the results for living conditions are similar (although less precisely estimated). Overall, this suggests that poorer people are more likely to be targets of vote buying by political parties during elections. This is consistent with the existing literature showing that political machines tend to focus vote-buying campaigns on the poor (Stokes Reference Stokes2005; Jensen and Justesen Reference Jensen and Justesen2014).
The result for the partisanship variable shows that partisans are more exposed to vote buying, indicating that parties tend to target supporters rather than swing voters. This finding is consistent with Nichter’s Reference Nichter2008 argument that parties use electoral bribes to mobilize core voters, but it contradicts the view that swing voters are the main targets of preelection redistribution (cf. Stokes Reference Stokes2005). Interestingly, models 1 and 2 suggest a strong relationship between receiving CCTs and vote buying. Even though the allocation of—and selection into—CCTs need not be characterized by clientelist transactions (Frey Reference Frey2019; Zucco Reference Zucco2013), this suggests that receiving CCTs increases the chance of being targeted by vote-buying campaigns during elections.
Models 3 and 4 include the full set of individual-level controls. Doing so only increases the coefficient sizes and significant levels for the two poverty measures, but otherwise does not substantially change the results. Models 5–8 move on to explore whether the relationship between poverty and vote buying changes for swing and partisan voters, respectively. This is done by reproducing models 3–4 for subsets of swing voters (models 5–6) and partisans (models 7–8).
To begin with, a cursory look at the regression coefficients suggests some interesting findings. In models 5–6, the coefficients for both the income and living conditions variables are negative and statistically significant, suggesting a negative relationship between poverty and vote buying for swing voters. That is, among swing voters, poorer people have a higher likelihood of being targeted by clientelist parties’ vote-buying campaigns during elections.
Models 7 and 8 show similar results for the subgroup of partisans. Here the results for the poverty measures are somewhat different. For the income measure in particular, the effect is weaker for partisans (model 7) than for swing voters (model 5). This would seem to suggest that income matters mainly for swing voters, whereas partisans are targeted more broadly across the income distribution. However, the results for the living standard measure do not fully align with this interpretation. The coefficients for living standards are almost identical for swing voters (model 6) and partisans (model 8). (Results for a formal test of differences in coefficient sizes are reported in the note to table 1.)
However, the association between living standards and vote buying is estimated less precisely among partisans, and accordingly has a lower level of statistical significance (although in table B in the appendix, the coefficients are somewhat larger and estimated more precisely). This provides somewhat mixed evidence for the conjecture that poverty matters differently for swing voters and partisans. On the one hand, there seems to be a clear negative association between poverty and vote buying for swing voters. On the other hand, the evidence is mixed for partisans, and we therefore cannot rule out that a similar pattern applies for people who align with a political party.
Overall, then, three results appear quite clear from table 1. First, poverty matters: there is a clear negative relation between poverty and vote buying, supporting findings from previous literature (Jensen and Justesen Reference Jensen and Justesen2014; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013). Second, partisans are, overall, more likely to receive clientelist offers from parties during elections. Third, people who are part of conditional cash transfer programs are also more likely to be approached with vote-buying offers by clientelist parties—possibly because CCT recipients are already dependent on state welfare and are easier to identify and approach for political parties during election campaigns.Footnote 5 While our results provide some suggestive evidence that poverty may matter more for swing voters than for partisans, the evidence presented here is mixed, meaning that we cannot make firm inferences about this relationship based on the data used here.
Conclusions
In the past two decades, there has been a resurgence of scholarly interest in the relationship between clientelism and vote buying worldwide. In Latin America, several country and small-N analyses have theorized that parties pursue mixed strategies to lure poor voters, but there is very little evidence at the cross-national level. This study has filled this important gap in the literature by bringing evidence to this thesis through a survey of 22 countries in the region. Much of the current debate focuses on poverty and partisanship as having independent effects on vote buying. This study adds to this literature while also exploring whether poverty and partisanship might jointly be related to vote buying.
The basic premise is that parties target loyalists as well as swing voters through vote-buying practices to maximize their chances of electoral success. Under the constraints of secret ballot voting, parties try to maintain the loyalty of their core voters by distributing preelection goods while at the same time pursuing those swing voters whom they regard as being most likely to exchange their votes for tangible benefits. This might suggest that the most responsive, uncommitted swing voters are the poorest.
Our results suggest that poverty and partisanship do indeed matter across a broad group of countries in Latin America. Indeed, to the best of our knowledge, these results constitute the most comprehensive cross-country analysis of the relationship among poverty, partisanship, and vote buying across countries in Latin America to date. We also find some evidence that poverty has a stronger relationship with vote buying among swing voters, but the relationship is not altogether robust and is contingent on the measure of poverty being used.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/lap.2022.66
Supporting Information
Additional supporting materials may be found with the online version of this article at the publisher’s website: Appendix.