Latin America experienced a stunning shift in social policy since the late 1990s. Whereas earlier forms of social policy focused on the protection of the formal sector and often failed to reach the poor, the introduction of means-tested programs has provided assistance to those in need (Díaz-Cayeros and Magaloni 2009). The conditional cash transfer (CCT) program is the primary example of this shift in social policy. In CCT programs, poor households receive a monetary transfer in exchange for fulfilling certain conditions, such as attending health checkups and health education workshops and having the children attend a certain percentage of school days (Fiszbein and Schady 2009; Cecchini and Madariaga 2011).
CCTs receive a large amount of praise for their effective targeting procedures and positive human development outcomes. In contrast to the political manipulation of beneficiary selection in many other antipoverty programs in Latin America, CCT benefits are generally recognized as being distributed according to objective criteria based on need (Fiszbein and Schady 2009; Díaz-Cayeros and Magaloni 2009; Fried 2012; Lindert, Skoufias, and Shapiro 2006). Certainly, there is variation in targeting effectiveness across CCT programs, and even across a single program over time, with some studies uncovering evidence of political interference in certain countries (De La O 2015; Rocha Menocal 2001). These negative findings are exceptions, however, as most evidence indicates that CCTs are well targeted according to need and lie outside the reach of clientelism. In addition, human development scholars find that CCTs have a positive effect on several child health outcomes and also lead to reductions in poverty and inequality (see Fiszbein and Schady 2009).
More recently, researchers have started to examine the consequences of CCTs for electoral behavior in Latin America. This growing literature produces mixed findings. Some studies find that CCTs have a positive effect on incumbent support and/or voter turnout (Hunter and Power 2008; Zucco 2013; Baez et al. 2012; Díaz-Cayeros, Estévez, and Magaloni 2016; De La O 2013, 2015; Manacorda, Miguel, and Vigorito 2011), whereas other studies do not find a positive relationship between CCTs and electoral behavior (Bohn 2011; Imai, King, and Rivera 2017; Corrêa 2015).
Although a heated debate continues over whether CCTs influence electoral behavior, we lack a comprehensive examination of the relationship between CCTs and political participation. Indeed, existing research on CCTs and political participation focuses almost exclusively on voter turnout, even though other forms of political activity are important in their own right and have meaningful consequences in the region. Do CCTs lead to increased participation in multiple political activities, or is the positive effect of CCTs confined to a select form of political participation (voting)? If CCTs contribute to broad political participation, how do CCTs lead to more participation in a wide variety of political activities?
Following Booth and Seligson (1978), political participation is defined here as “behavior influencing or attempting to influence the distribution of public goods” (6; italic in the original). This definition recognizes that political participation encompasses a wide range of electoral and nonelectoral activities. I define broad political participation as involvement in multiple modes of political activities.
My central argument is that CCTs lead to broad political participation among beneficiaries, and that both the program transfers and the conditionalities play an important role in shaping political activity. I define “conditionalities” as the requirements that program beneficiaries must meet in order to continue receiving benefits from CCT programs.1 The transfers provide key material resources that facilitate political participation, while the conditionalities offer important nonmaterial resources that boost involvement in political activities.
This article contributes to the literature on political behavior and policy feedback in two ways. First, it challenges the findings from the existing literature on the political behavioral consequences of program conditionalities. Whereas previous research suggests that CCT conditionalities, and means-tested program conditionalities more generally, reduce political participation or have no effect (Layton and Smith 2015; Watson 2015), I utilize an alternative research design to argue that CCT conditionalities spur increased involvement in political activities. Second, the article improves our theoretical understanding of why some means-tested social programs increase political participation, while others have no effect or even a negative effect on participation (Soss 1999; Campbell 2003; Mettler and Stonecash 2008; Bruch, Ferree, and Soss 2010; Garay 2007; Holzner 2010). Social programs can have multiple components, and each component has the potential to contribute to the overall effect of the program on political behavior.
What is the relationship between CCTs and political participation? If the relationship is positive, how do CCTs increase political participation? To answer these questions, I emphasize the importance of program transfers and conditionalities in shaping decisions to participate in politics.
In line with rational choice models, I argue first that individuals consider the expected utility of performing an act when deciding whether to participate in politics. The basic equation for deciding whether to participate in a specific political activity is Ua = p(Ba) – Ca, where “a” is the political activity; “Ua” is the expected utility of participating in the political activity; “p” is the probability that the political activity will make a difference in the outcome; “Ba” is the expected benefit from the outcome; and “Ca” is the cost of participating in the political activity.2 Additionally, an individual may receive benefits from simply participating in the political activity, generally denoted by a “D” term (Downs 1957; Riker and Ordeshook 1968).
How do individuals determine the expected costs and benefits of performing a political act? Following the institutional theory of political participation (Holzner 2010), I argue that the state plays a central role in shaping individuals’ cost-benefit calculations. In particular, the state has the ability to distribute politically relevant resources through its policies, and these resources directly influence the expected costs (Ca) and benefits (Ba) of political participation (Holzner 2010).
Among the actions of the state, conditional social assistance programs are particularly influential in shaping an individual’s expected utility for performing a political act.3 There are two reasons for this influence. First, the desirable transfers provide an incentive for program beneficiaries to defend or expand the social program (Pierson 1993; Campbell 2003; Holzner 2010), directly increasing the expected benefits of political participation. By raising the expected benefits of political participation, the transfer component is expected to drive more political activity among program beneficiaries. Second, the conditionalities affect the expected costs of political participation by altering the beneficiaries’ stock of nonmaterial, politically relevant resources.4 For example, all conditionalities reduce the beneficiaries’ available time to spend on nonessential activities such as political participation, making political participation more costly. Importantly, though, conditionalities can also provide beneficiaries with nonmaterial resources that reduce the costs of political participation. The overall effect of the conditionalities thus depends on the program design.
In the particular case of CCTs in Latin America, I argue that these programs will drive increased political participation among beneficiaries, and that both the cash transfers and the conditionalities will shape decisions to participate in politics. Indeed, CCT programs in Latin America give sizeable cash transfers directly to beneficiaries, increasing the incentive for these beneficiaries to participate more in politics in order to defend or expand the desirable transfers.5 The increased incentives should raise the expected benefits of participating in political activities and lead to more political participation among beneficiaries.
In addition, the CCT conditionalities provide beneficiaries with important nonmaterial resources that reduce the costs of political participation and facilitate more involvement in politics. Although the exact set of conditionalities can vary slightly across CCT programs in Latin America, CCT beneficiaries typically are required to attend regular checkups at local health facilities, attend regular health education workshops and other group meetings with program officials, and engage with local school officials (Fiszbein and Schady 2009; Cecchini and Madariaga 2011). The increased social interactions with local state officials and community members in an organized setting lead to more opportunities for beneficiaries to practice civic skills, which are defined as “organizational and communications abilities that facilitate political life” (Verba, Schlozman, and Brady 1995, 4). In particular, beneficiaries are likely to exercise communication and organizational skills when participating in required workshops and meetings with program officials and other beneficiaries; when interacting with public service employees in local health and education facilities; when keeping track of deadlines and pending requirements; and when addressing any problems that may arise in the program.
Because CCT beneficiaries encounter opportunities to exercise communication and organizational skills through the conditionalities, beneficiaries gain regular experiences in building and maintaining these important skills. According to Verba, Schlozman, and Brady (1995), civic skills are crucial resources for political involvement because they increase an individual’s capacity to participate in political activities and thus reduce the costs of political participation. Not surprisingly, civic skills have been shown to be strong determinants of political activity in a wide range of contexts, including developed and developing countries (Verba, Schlozman, and Brady 1995; Finkel 2002). Civic skills are especially critical for more demanding forms of political participation that require communicating and/or working with others.
I expect adult female CCT beneficiaries, in particular, to gain politically relevant resources from participating in the program. A key feature of the CCT program is that an adult female generally is required to be the head participant in each family. As the head program participant, the adult female receives the cash transfers and ultimately is responsible for fulfilling the program conditions (Fiszbein and Schady 2009). In order to satisfy certain conditions or resolve some problems regarding the program, the presence and active involvement of the head program participant may even be required. Of course, male beneficiaries may assist with fulfilling the conditions, so they too may gain opportunities to practice civic skills through the program. However, as the head program participants, the female beneficiaries will have more opportunities than the male beneficiaries to exercise civic skills. The increased opportunities to exercise civic skills will facilitate even more political participation for female beneficiaries, given that their costs of participating in political activities will be even further reduced.
My argument offers the following predictions regarding CCTs, otherwise similar unconditional cash transfers (UCTs), and political participation:
Hypothesis 1: (H1a) CCTs will be positively associated with political participation, and especially with participation in more demanding political activities; (H1b) the positive association between CCTs and political participation will be larger than the positive association between UCTs and political participation; and (H1c) the positive association between CCTs and political participation will be larger for female beneficiaries.
My argument also makes the following predictions regarding the potential pathways that connect CCT conditionalities and political participation:
Hypothesis 2: (H2a) CCTs will be positively associated with civic skills; (H2b) the association between CCTs and civic skills minus the association between UCTs and civic skills will be a positive value;6 and (H2c) the positive association between CCTs and civic skills will be larger for female beneficiaries.
The literature on social assistance programs offers two competing explanations for why CCTs may contribute to increased political participation among beneficiaries. Holzner (2010) argues that government assistance programs contribute to increased external political efficacy among beneficiaries, because the transfers demonstrate that the government is paying attention to, and addressing, their needs.7 The increased external political efficacy, according to this argument, will facilitate more political participation among program beneficiaries. Hunter and Sugiyama (2014) argue that CCTs in Brazil boost feelings of social inclusion and personal agency, which in turn facilitate political involvement. They emphasize that the particular policy features of the Brazilian CCT program are responsible for the positive program experiences and subsequent changes in citizenship. To the extent that CCTs in other contexts have similar policy designs, Hunter and Sugiyama would expect similar consequences for citizenship.
I test the hypotheses using two sources of data: original survey data from Mexico and Latin American Public Opinion Project (LAPOP) survey data.8 Mexico is an ideal country for this study, because it has a CCT program (Oportunidades) and a UCT program (Programa de Apoyo Alimentario, or PAL) that provide a similar amount of cash to program beneficiaries, serve almost identical beneficiary populations, and have the same administrative agency and structure.9 Indeed, PAL is a means-tested, unconditional cash transfer program that uses nearly the same beneficiary selection criteria as Oportunidades at the household level, but it targets more rural areas where poor citizens may have more difficulty accessing local health and education services.10 PAL, with 3.7 million beneficiaries in total, is a smaller program than Oportunidades, which has nearly 26 million beneficiaries.11 Both programs are administered through the national Oportunidades agency.
Importantly for our purposes, the one crucial difference between these programs in Mexico is that the CCT program has a conditionalities component while the UCT program does not. The similarities across these programs allow for important natural controls when comparing program effects, which in turn provide increased leverage when trying to isolate the effect of CCT conditionalities. Furthermore, the beneficiary selection process in Mexico uses an objective formula—based on household and geographic criteria—that is free from political manipulation (Levy 2006; Díaz-Cayeros, Estévez, and Magaloni 2016), thus alleviating concerns of bias due to politicized selection criteria.
The original survey is from July 2014 and contains a representative sample of adults (aged eighteen or older) in three Mexican municipalities, including two in the state of Puebla (Oriental and San Felipe Teotlalcingo) and one in the state of Mexico (Ixtlahuaca).12 The selected municipalities offer several advantages for my study. First, they have a very high rate of poverty (nearly 80 percent of residents are below the poverty line), including a large percentage of poor residents who are not CCT or UCT beneficiaries.13 This setup creates an extensive set of natural controls for my analysis of program effects, where the differences between program beneficiaries and nonbeneficiaries are minimized. Indeed, the survey likely captures many poor residents, including both beneficiaries and nonbeneficiaries, whose poverty scores were close to the formula-based program eligibility line. Second, the municipalities contain a sizeable percentage of CCT beneficiaries and UCT beneficiaries, allowing for a comparative analysis of program effects. Third, because the beneficiaries of both programs reside in the same municipalities, the case selection minimizes any potential differences in rurality and local services across the two groups of beneficiaries. This advantage is especially important for comparing the effects of the CCT and UCT programs in Mexico because, as mentioned above, geographic factors determine whether each qualified household is eligible for the CCT program or for the UCT program. Within the selected municipalities, the physical distance between CCT- and UCT-eligible areas is very small, separating neighborhoods that are otherwise very similar. The similarities create natural controls for my analysis.
The fourth advantage of selecting these municipalities for my survey is that the CCT and UCT beneficiaries in these areas receive similar amounts of cash transfers through the programs, with CCT beneficiaries receiving an average of 660 pesos (about US$51) per month and UCT beneficiaries receiving an average of 540 pesos (about US$42) per month.14 Given the other similarities across the programs, including the beneficiary selection criteria and the administrative agency and structure, the similar transfer amounts allow us to isolate the impact of the CCT conditionalities.
The case selection for my survey clearly attempts to minimize the differences between CCT beneficiaries and nonbeneficiaries and between CCT and UCT beneficiaries. Is this goal achieved in practice? As an empirical test of whether the nonbeneficiaries and UCT beneficiaries are otherwise similar to the CCT beneficiaries in the three Mexican municipalities, I use the original dataset to compare the values of new program beneficiaries and nonbeneficiaries on a variety of variables. New beneficiaries are defined as program recipients who have participated in the program for less than one year, which is the smallest time interval available. Significance tests are utilized in the analysis, with the nature of the dependent variable determining which exact test is used. Despite some limitations, the analysis provides a reasonable, near-baseline assessment of the similarities between groups. The results are listed in the online appendix (Tables A1 and A2).
Tables A1 and A2 provide strong evidence that new CCT beneficiaries are similar to both nonbeneficiaries and new UCT beneficiaries on a variety of variables. The small, not statistically significant difference in locality population between new CCT beneficiaries and new UCT beneficiaries is particularly striking, given that UCTs are targeted to more rural areas. This finding suggests that my strategy to minimize the physical distance between CCT- and UCT-eligible areas was effective. The absence of statistically significant differences in socioeconomic variables further illustrates the high level of poverty in these municipalities, while the nonsignificant findings on the political variables provide additional support that CCTs are not politicized in Mexico. Overall, the evidence from these two tables increases our confidence that nonbeneficiaries and UCT beneficiaries can serve as valid counterfactuals for CCT beneficiaries in the analysis.
One limitation of the original Mexican survey data is that there still may be differences in rurality between CCT and UCT beneficiaries, despite my efforts to minimize these differences. If these contextual differences influence whether program beneficiaries participate in politics, then it is unclear whether conditionalities or rurality are driving the effects on political participation.
Another potential limitation of the Mexican survey data is the threat of spillover effects. More specifically, the small physical distance between program beneficiaries and poor nonbeneficiaries in the municipalities creates opportunities for the poor nonbeneficiaries to acquire positive program effects. Although I acknowledge this threat, it is important to highlight that spillover effects would work against finding positive estimates of CCT and UCT program effects. In other words, spillover effects would create a more difficult test of the hypotheses.
The second main source of data is the LAPOP survey data. The LAPOP survey was conducted in eighteen Latin American countries in 2012, and it includes a nationally representative sample of adults (aged eighteen or older) in each country. The LAPOP surveys in nine countries contain questions on CCT participation, general social program assistance, and involvement in a wide variety of political activities, allowing for extensive analysis of the relationship between program participation and political participation.15
The following empirical analysis is divided into three parts. First, I analyze the relationship between program participation (in CCTs and otherwise similar UCTs) and participation in a wide range of political activities. This analysis draws from the original Mexican survey data. Next, I consider the particular pathways that connect CCTs and political participation, and I examine whether CCT conditionalities contribute to increased civic skills. This part of the analysis utilizes the data from my original survey in Mexico. The third part, which draws from the LAPOP survey data, analyzes the relationship between program participation and political participation in other areas of Latin America. More detailed descriptions of all variables and their summary statistics are included in the appendix.
To examine the relationship between CCTs, UCTs, and political participation (H1a, H1b, and H1c), I primarily use the original survey data from Mexico. The original survey contains questions on CCT participation, UCT participation, a standard set of individual and household characteristics, and involvement in twenty-three political activities.
The dependent variables in this analysis are a set of indices for the different modes of political participation. Through the use of factor analysis, prior research on the dimensions of political participation in Latin America (Mexico in particular) identifies six distinct modes of political participation: voting, campaign activism, contacting public officials, community activism, civil society engagement, and protest (Booth and Seligson 2009; Klesner 2009). I follow this work and construct a separate index for each of these six modes. The additive index for each mode is made up of between two to six survey questions, with each survey question rescaled to have equal weight in the index and a maximum value of one.
I use confirmatory factor analysis to test whether the indices correspond to different dimensions of political participation in these Mexican municipalities, and the results are reported in the appendix (Table A3). The results provide strong support that each index represents a separate dimension of political participation in the Mexican survey data. The root mean square error of approximation (RMSEA) values are all below or close to .06, and the comparative fit index (CFI) values are all above or close to .95, providing no cause for concern (Hu and Bentler 1999).
Studies suggest that factor analysis may lead to the selection of too many dimensions or factors (overfactoring), which in turn can create nonmeaningful factors and estimates that are not as stable (see Fabrigar et al. 1999, 277–281; Zhang 2014, 341). Due to these potential concerns, I construct two alternative sets of dependent variables. The first alternative set of dependent variables simply consists of each separate political activity, and the second set consists of two large categories of political activities (electoral and nonelectoral political participation). For the second alternative set of dependent variables, an additive index is created for both electoral and nonelectoral political participation. As robustness checks, I repeat all of the following analyses on political participation using each of the alternative sets of dependent variables, and I report the results in the appendix (Tables A4–A10).
The independent variables in this analysis are CCT participation,16 UCT participation, and a standard set of control variables. CCT participation and UCT participation are each dichotomous variables for whether someone in the household is a beneficiary of these programs. The control variables include income,17 education level, the natural log of age, a gender variable, whether the respondent has children, and religiosity. The locality-level control variables of population, poverty, and education are also included in the models, in order to account for potential differences across localities.18
Using the Mexican survey data, I employ multilevel linear models with random intercepts (Rabe-Hesketh and Skrondal 2008) to assess the relationship between program participation and political participation.19 A multilevel model is a “regression … in which the parameters—the regression coefficients—are given a probability model…. The feature that distinguishes multilevel models from classical regression is in the modeling of the variation between groups” (Gelman and Hill 2007, 1). For data that is nested in groups, a key advantage of a multilevel model is that it allows for within-group correlation, which in turn reduces concerns regarding biased standard errors (Rabe-Hesketh and Skrondal 2008, 129–130). Each multilevel model of political participation has three levels: the individual level, the locality level, and the municipality level. The analysis serves two main purposes: to test whether CCTs are positively associated with political participation (H1a); and to compare the impact of CCT participation and UCT participation on political participation in this same context (H1b). The results are listed in Table 1.
|(1) Voting||(2) Campaign activism||(3) Contact public officials||(4) Community activism||(5) Civil society engage||(6) Protest|
In line with my expectations, there is a positive and statistically significant relationship between CCTs and several modes of political participation. CCTs are positively associated with contacting public officials, community activism, civil society engagement, and voting. For the average respondent, the estimated effect of CCT participation corresponds to an increase of 26 percent in contacting public officials, an increase of 28 percent in community activism, an increase of 21 percent in civil society engagement, and an increase of 6 percent in voting. As my argument predicts, the impact of CCTs is greater for more demanding forms of political participation, including community activism, contacting public officials, and civil society engagement. The estimated impact of CCTs is much smaller for voting, which is considered a low-cost activity (Niemi 1976; Aldrich 1993).
As for UCTs, the results indicate that UCT participation is positively and significantly associated with only one mode of political participation: voting. These findings offer strong support for hypothesis (H1b), which predicts that CCTs will have a larger positive association with political participation than UCTs. The evidence suggests that CCT conditionalities matter for political participation, because the individuals who only receive the cash transfers—the UCT beneficiaries—increase their participation in only one mode of political activity. It also is not surprising, at least when viewed from the lens of my argument, that this one mode of participation is the low-cost activity of voting. As expected, CCT conditionalities have a greater impact on high-cost forms of political participation.
Because UCTs are positively associated with voting, the results provide at least some support for the argument that material program transfers boost political participation. Still, as predicted, it appears that the CCT conditionalities also play an important role in driving increased political participation among program beneficiaries.
One potential concern regarding the results is that there may be an overlap between the measures of the dependent variables and the CCT requirements. If the political participation measures simply capture activities that CCT beneficiaries need to perform in order to continue in the program, then the analyses above are not valid tests of my argument. To address this concern, I remove several political activities from the measures of the following three indices of political participation: (1) contacting public officials (the activities of contacting a state official or agency and contacting a public program official are removed); (2) community activism (the activities of helping to resolve a community problem and attending meetings for a community improvement group are removed); and (3) civil society engagement (the activities of attending school and health organization meetings are removed). I then repeat the analyses from Table 1 with these alternative measures of the three political participation indices, and I report the results in the appendix (Table A11). The main findings are robust to using alternative measures for two of the three political participation indices (contacting public officials and community activism).20 However, CCT participation is no longer statistically significant in the model of civil society engagement, which suggests that at least part of the estimated CCT effect is due to the program requirements. Overall, the findings indicate that the required CCT activities are not directly responsible for the positive associations between CCTs and multiple modes of political participation.
The Mexican survey data also allows for an analysis of whether the positive association between CCTs and political participation is larger for female beneficiaries (H1c). An interaction term of CCT program participation and gender is added to the models from Table 1, and the results are reported in Table 2.21 Contrary to my expectations, the evidence suggests that there is not a statistically significant difference in the estimated effects of CCTs across gender for most of the political participation modes. In other words, the positive CCT effects are working through both female and male beneficiaries for most modes of political participation, despite the females being the head program participants. The one exception is for the mode of contacting public officials, where the estimated effect of CCTs is much larger for female beneficiaries. I revisit, and attempt to explain, these unexpected results at the end of the next subsection.
|(1) Voting||(2) Campaign activism||(3) Contact public officials||(4) Community activism||(5) Civil society engage||(6) Protest|
I now consider the pathway that connects CCTs, and particularly the conditionalities component, to political participation. My argument predicts that CCTs will be positively associated with civic skills (H2a), and that the association between CCTs and civic skills minus the association between UCTs and civic skills will be a positive value (H2b). Furthermore, I expect the positive association between CCTs and civic skills to be larger for female beneficiaries (H2c). To test these hypotheses, I once again utilize the original 2014 Mexican survey data.
The dependent variables for this analysis are several measures of civic skills. Rather than relying on subjective assessments of civic skills, I follow Verba, Schlozman, and Brady (1995) and ask respondents about their recent experiences exercising specific civic skills at work, through a social program, or in other nonpolitical activities. The idea behind the measure is that if adults practice these types of skills outside of their political activities, then they are able to develop or maintain the skills for future political participation. This type of measure also has the added benefit of being more objective—and thus leading to less measurement error—than self-assessments of civic skills.22 In the survey, respondents were asked how often they performed each of the following organizational and communications activities in the past year: attended a meeting and participated in the discussion; organized a meeting or set up an appointment; wrote a letter or filled out forms; and asked someone to clarify information. The respondents then selected one of the following options: never, one or two times a year, one or two times a month, or once a week. These survey questions are used to create four separate civic skill variables as well as an additive index of civic skills, with each civic skill rescaled to have equal weight in the index and a maximum value of one.
In the following analysis, I use the same set of independent variables as in the previous subsection, including CCT participation, UCT participation, the individual-level control variables, and the locality-level control variables. Multilevel linear models with random intercepts are utilized to assess the relationship between program participation and civic skills.23 Table 3 reports the results of these models.
|(1) Attended a meeting and participated in the discussion||(2) Organized a meeting or set up an appointment||(3) Wrote a letter or filled out forms||(4) Asked someone to clarify information||(5) Civic skills index|
Before discussing the results, I highlight some potential endogeneity concerns with this analysis. It may be the case that individuals who exercise civic skills are more likely to enroll in CCT programs, which challenges the causal direction of the relationship between CCTs and civic skills. As a result, I am unable to claim that the following analysis provides tests of a causal mechanism. Still, there is a compelling reason to think that the threat of endogeneity is relatively low in these models. The available evidence from Mexico suggests that CCT beneficiaries are selected using objective criteria (Levy 2006; Díaz-Cayeros, Estévez, and Magaloni 2016), and that the take-up rate among eligible households is 97 percent (Gertler 2000).24 Consequently, there is little room for individuals to self-select into the program.
As predicted, Table 3 indicates that the relationship between CCTs and each measure of civic skills is positive and statistically significant. The results suggest that the CCT program creates opportunities for beneficiaries to attend meetings and participate in the discussions; organize meetings or set up appointments; write letters or fill out forms; and ask other people to clarify information. These experiences allow CCT beneficiaries to develop and/or maintain important organizational and communication skills, which in turn reduce the costs of political participation and likely facilitate more political activity.
Moreover, as expected, the difference between the estimated CCT effect and the estimated UCT effect is a positive value in each model from Table 3. Interestingly, UCTs are positively associated with exercising one civic skill (attending meetings and participating in discussions), and also are positively associated with the overall index of civic skills. In both instances, the size of the estimated UCT coefficient is much smaller than the estimated CCT coefficient, indicating that CCTs lead to more experiences practicing civic skills. These findings suggest that the cash transfers and the conditionalities each lead to more opportunities for program beneficiaries to practice civic skills, but that the conditionalities have a much larger and broader impact. In terms of the substantive effect, it is estimated that CCTs lead to a 59 percent increase in the exercising of civic skills for the average respondent, while UCTs only lead to an increase of 29 percent.
Once again, it is important to address the concern about whether the measures of the dependent variables (civic skills) simply are capturing CCT requirements. I therefore construct an alternative index of civic skills that excludes the exercise of two specific skills: (1) attending a meeting and participating in the discussion, and (2) writing a letter or filling out forms. I then repeat the analyses from Table 3 with the alternative index of civic skills, and I report the results in the appendix (Table A18). The main findings are robust to using the alternative index of civic skills, indicating that CCTs provide opportunities to exercise civic skills beyond the basic activities that are required to continue in the program.25
Overall, Table 3 provides support for the hypothesized pathway connecting CCTs and increased political participation. The results are consistent with the argument that CCTs, and particularly the conditionalities, produce important nonmaterial resources for beneficiaries, including opportunities to exercise civic skills. Based on the evidence, it appears that these nonmaterial resources may lower the costs of political activity for beneficiaries and create a pathway to increased political participation, particularly for more demanding activities. Potential endogeneity concerns, discussed above, challenge the causal direction of the relationship, however. CCTs and civic skills are positively and significantly associated, but I am unable to determine whether changes in civic skills are a part of the causal mechanism that produces changes in political participation.
To test the last hypothesis (H2c), I add an interaction term for CCTs and gender to the models from Table 3.26 The results are presented in Table 4, and they suggest that female beneficiaries generally exercise more civic skills than male beneficiaries, as expected. The results thus offer more support for my argument. Interestingly though, the estimated CCT effects for male beneficiaries are still positive and statistically significant for all four types of civic skills, indicating that CCTs have a positive impact on civic skills for male beneficiaries as well. According to this evidence, the CCT program is pushing male beneficiaries, not just female beneficiaries, to exercise civic skills. For one form of civic skill (asking someone to clarify information), there is not even a statistically significant difference in the estimated CCT effects across gender. This result implies that CCTs have the same positive impact on exercising that particular civic skill for both male and female beneficiaries.
|(1) Attended a meeting and participated in the discussion||(2) Organized a meeting or set up an appointment||(3) Wrote a letter or filled out forms||(4) Asked someone to clarify information||(5) Civic skills index|
The findings on CCTs, gender, and civic skills offer some important insight into the unexpected findings on political participation from the previous subsection. According to the evidence from Table 4, CCTs are positively associated with political participation for both male and female beneficiaries, because the program pushes both genders to exercise more civic skills. In other words, male beneficiaries appear to be helping the female beneficiaries fulfill the program conditions, which in turn offers the males some important opportunities to exercise civic skills. Even though female beneficiaries, as the head participants in the program, receive the cash transfers and ultimately are responsible for fulfilling the conditions, the male beneficiaries are exercising some crucial skills for political participation as well.
Although the above results support my argument, they do not rule out the possibility of other potential pathways from CCTs to increased political participation. Several competing arguments, which were described in the theory section, suggest that CCTs will contribute to increased political participation through other mechanisms. The original 2014 Mexican survey data contains measures of these alternative pathways, permitting detailed analysis of the competing explanations. I use multilevel models to analyze the relationship between CCTs and several political attitudes, including two measures of external political efficacy, three measures of internal political efficacy, and a measure of interest in politics. The two external political efficacy variables offer measures of sociopolitical inclusion; the three internal political efficacy variables provide measures of personal efficacy; and the political interest variable offers an additional measure of psychological engagement with politics. I report the results in the appendix (Table A24).
The results from Table A24 provide no support for the competing explanations. According to this evidence, the alternative pathways are not responsible for the changes in political participation.
The analyses presented above only consider data from Mexico. Do CCTs, and especially the conditionalities, boost political participation in other Latin American contexts? To test the generalizability of my argument, I utilize data from the 2012 LAPOP survey, which contains questions on CCT participation and twenty-one forms of political participation in nine Latin American countries. Unfortunately, the LAPOP survey does not contain questions on civic skills, so the hypotheses on civic skills cannot be evaluated with this data.
The dependent variables—the six indices of political participation from the previous subsection—are measured using nearly the same set of political activities for each respective index. Confirmatory factor analysis corroborates that each index represents a separate dimension of political participation in the 2012 LAPOP survey data.27
The independent variables in this analysis are CCT participation, UCT participation, and a standard set of control variables. CCT participation is a dichotomous measure of whether someone in the household participates in the CCT program. Although the cash transfer amounts and program requirements are not identical across Latin American countries, the CCT programs in Latin America share a large number of similarities and several programs are modeled after the CCT program in Mexico (Fiszbein and Schady 2009; Cecchini and Madariaga 2011).
UCT participation is a dichotomous measure of whether someone in the household regularly receives (non-CCT) government assistance in the form of money or products. One potential concern with the UCT measure is that it likely captures transfer programs that differ from CCTs in ways beyond the conditionality of the transfers. To minimize these other differences between UCT and CCT programs, and also to address concerns about potential differences across the selected recipients, I only allow households that are below the poverty line to be counted as participating in UCTs. Thus, the UCT measure at least provides a rough proxy for participation in unconditional antipoverty programs that transfer desirable resources to poor households.
Using the variables from the 2012 LAPOP data, I employ multilevel linear models with random intercepts to analyze whether CCTs and UCTs are positively associated with political participation.28 Each multilevel model has three levels: individual, municipality, and country. The results are listed in Table 5.
|(1) Voting||(2) Campaign activism||(3) Contacting public officials||(4) Community activism||(5) Civil society engage||(6) Protest|
As predicted, the results indicate that CCT participation is positively and significantly associated with several modes of political participation, including campaign activism, contacting public officials, civil society engagement, and protest. My argument also predicts that CCTs will have a larger impact on more demanding political activities, and this too finds support in the results. CCTs have no relationship with voting, but are positively associated with high-cost activities such as campaigning and participating in civic organizations. For the average respondent, the estimated percentage increase in political participation ranges from 8 to 41 percent.
Furthermore, according to these results, CCT conditionalities likely play an important role in boosting political participation among beneficiaries in Latin America. UCTs do not have a statistically significant relationship with any of the political participation modes, which implies that the transfers alone are not responsible for the estimated positive impacts of CCTs. It appears that the combination of transfers and conditionalities boosts political participation, not just in Mexico but throughout Latin America.29
Although the LAPOP evidence supports my predictions, the results from this subsection should be viewed with more caution. The potential differences between the CCT and UCT variables in these other countries, as well as the lack of natural controls in the nationally representative samples, may bias the results. Another potential threat to inference is that some programs outside of Mexico may be politicized. In other words, several of the main advantages of the original Mexican survey data are absent in the LAPOP data, creating greater concerns about the estimated CCT effects. Moreover, because LAPOP does not include questions on civic skills, it is not possible to analyze the predicted pathway that connects CCTs and political participation. Overall, the LAPOP results provide preliminary evidence that CCTs are boosting political participation throughout Latin America, but more detailed data is needed to confirm these results and to test the particular pathway(s) to increased political participation in other Latin American countries.
The results from the Mexican context, and perhaps throughout Latin America, suggest that the pathway from conditional means-tested programs to increased political participation contains nonmaterial resource effects. These findings have important implications for debates on how the welfare state shapes mass political participation. In particular, the results pose a challenge to theoretical arguments that emphasize the importance of interpretive effects in explaining the consequences of social programs for political participation (Pierson 1993; Soss 1999; Watson 2015).
The evidence presented above also raises an important question: why do conditional means-tested programs appear to boost political participation in developing countries, but they discourage political participation in advanced industrialized countries (Watson 2015; Soss 1999)? One potential answer to this question builds on insight from Bruch, Ferree, and Soss (2010) that not all means-tested programs are the same. Indeed, the nature of the conditionalities for means-tested programs may be different across contexts, and the different conditionalities may contribute to different effects on politically relevant resources. Whereas conditionalities in Latin America provide opportunities for beneficiaries to gain key civic skills and thus appear to overcome any reductions in available time for political participation, the conditionalities in the United States and Europe may not offer beneficiaries the same types of opportunities. For example, in the United States, research on conditional means-tested programs reveals that beneficiaries have negative interactions with program officials and little opportunity for acquiring key political resources (Soss 1999). Given that the conditionalities in these United States programs still require time commitments from beneficiaries and reduce their available time to spend on political participation, the conditionalities likely create a net loss in politically relevant resources among beneficiaries in the United States. The different resource effects thus may help explain the different consequences for political participation across contexts.
As CCTs continue to expand in Latin America and throughout the global south, we must take into account their consequences for broad political participation. Earlier research focuses on the effects of CCTs on voter turnout, but this narrow measure of political participation misses other important consequences and may even be misleading.
In particular, the conditionalities component of CCTs deserves more of our attention. According to the available evidence, CCT conditionalities matter for political participation, as citizens acquire important nonmaterial resources that reduce the costs of participation and facilitate more activity. This finding may help to explain why some means-tested programs lead to increased political participation, while other programs have no effect or even a negative effect on political involvement. By breaking apart the effects of the different program components, we can resolve apparent puzzles in the study of social programs and political participation.
Through conditionalities, the welfare state can play a critical role in providing opportunities for citizens to develop and/or maintain civic skills. When governments target these opportunities to certain individuals through social programs, they likely influence which citizens receive nonmaterial resources. More importantly, the governments shape the costs of political participation for these citizens, particularly for more demanding political activities. Future research that analyzes the mobilizing effects of government transfers must take into account any conditions that are attached to the transfers, as well as their potential to lead to different effects across modes of participation.
1I define and use the word “conditionalities” instead of “conditions” because “conditions” may mean the initial program eligibility conditions, the program continuation requirements, or the contextual conditions for the program (see Zucco, Luna, and Baykal 2019).
2For a more detailed description of the basic model, see Holzner (2010, 32–33) and Aldrich (1993, 247–251). The model was originally developed by Downs (1957) for voter turnout, but it can be adapted for any form of political activity.
4See Pierson (1993) for a discussion on the resource effects of policies and the distinction between nonmaterial resource effects, material resource effects, and interpretive effects. By recognizing the importance of politically relevant resources for political participation, this part of the argument draws from the civic voluntarism model (Verba, Schlozman, and Brady 1995).
5The size of the cash transfers varies across countries and time, but studies suggest that the transfers represent from 7 percent to 31 percent of consumption in beneficiary households in Latin America (see Fiszbein and Schady 2009, 105).
6The differences in wording between hypotheses H1b and H2b are due to the different predictions regarding the program transfer component. Whereas my argument predicts that the transfer component will be positively associated with political participation, it makes no predictions regarding the transfer component and civic skills.
7For more information on the distinction between external and internal political efficacy, see Niemi, Craig, and Mattei (1991).
8The AmericasBarometer by LAPOP, www.LapopSurveys.org. All additional files, including the appendix, data sets, and replication files, are available online at Harvard Dataverse (dataverse.harvard.edu).
9Note that the CCT program in Mexico originally was named Progresa and now is called Prospera. For an overview of these programs, see http://prospera.gob.mx/.
10For information about the rules for each program, see https://www.prospera.gob.mx/Portal/wb/Web/reglas_de_operacion.
11For this data, see http://www.sedesol.gob.mx/es/SEDESOL/Padron_de_Beneficiarios.
12I designed the survey. The survey was collected by Parametría, a research firm in Mexico. The sample size for each municipality is 380 respondents (1,140 respondents in total). The sampling error is ±5 percent. The survey technical information and questionnaire are available upon request.
13The source of the poverty data is the 2010 Análisis y Medición de la Pobreza, Consejo Nacional de Evaluación de la Política de Desarrollo Social, http://www.coneval.gob.mx/Medicion/Paginas/Medici%C3%B3n/Medicion-de-la-pobreza-municipal-2010.aspx.
14The cash transfer amount as a percentage of household income also is similar across programs in these areas, with the average cash transfer amount representing 54 percent of household income for CCT beneficiaries and about 46 percent of household income for UCT beneficiaries. The source of this data is the original survey.
18The source of the locality-level data is the 2010 Mexican census, Instituto Nacional de Estadística y Geografía (INEGI). See http://www.inegi.org.mx/sistemas/consulta_resultados/iter2010.aspx.
19Multilevel models are used in the analyses, because tests reveal that there exists sufficient variation across groups (see Table A12). As a robustness check, I use ordinary least squares (OLS) linear regression models with clustered standard errors and include dummy variables for localities and municipalities. The results are reported in Table A13.
20Given that the political participation variables are additive indices of political activities, it is not surprising that the size of the estimated effects is smaller when using the alternative measures with fewer activities. The substantive effect, measured as the percentage increase for the average respondent, is similar in size for each of the alternative measures.
21As a robustness check, I again use OLS linear regression models with clustered standard errors and include dummy variables for localities and municipalities (see Table A16). For the results of the multilevel models that use the female variable in the interaction term, see Table A17.
22For a discussion, see Verba, Schlozman, and Brady (1995).
23Tests reveal that there exists sufficient variation across groups for multilevel models, except for one of the civic skills (see Table A19). The main results are robust to using an OLS linear regression model for this civic skill (see Table A20). As an additional robustness check, I use OLS linear regression models with clustered standard errors and include dummy variables for localities and municipalities (Table A21).
24Initially, the take-up rate in urban areas was much lower, because a household targeting process was not used in the first few years. The selection process in urban areas was later changed to incorporate a household targeting component (see Angelucci and Attanasio 2009).
26The results are robust to using OLS linear regression models with clustered standard errors (see Table A22). For the results of the multilevel models that use the female variable in the interaction term, see Table A23.
27For the results, see Table A25. As robustness checks, I again analyze the effects of program participation on each separate political activity and on indices of electoral and nonelectoral participation (see Tables A26–A32).
28Tests reveal that there exists sufficient variation across groups to use multilevel models (see Table A34). As a robustness check, I use OLS linear regression models with clustered standard errors and include dummy variables for countries (see Table A35).
This article is derived from my dissertation (Schober 2015). Previous versions of this paper were presented at the International Congress of the Latin American Studies Association in 2013, the Annual Meetings of the American Political Science Association in 2014 and 2015, and the Junior Scholars Forum at the Stanford Center on Philanthropy and Civil Society (PACS) at Stanford University in 2015. This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. (NSFDGE1106401), the Duke Global Health Institute Doctoral Dissertation Fieldwork Grant, and the Dissertation Fellowship from the Center for the Study of Philanthropy and Voluntarism at Duke University. For their insightful comments and suggestions, I thank Karen L. Remmer, John H. Aldrich, Guillermo Trejo, David Soskice, Erik Wibbels, Sarah Brooks, Frances Hagopian, Elizabeth J. Zechmeister, Cesar Zucco, Evelyne Huber, Rob Reich, Walter W. Powell, Paul Brest, Johanna Mair, Emma Saunders-Hastings, Yan Long, Anirudh Krishna, Christopher Johnston, André Blais, anonymous reviewers, and the editors of the Latin American Research Review. I thank Parametría for collecting the survey. I also thank the Latin American Public Opinion Project (LAPOP) and its major supporters (the United States Agency for International Development, the Inter-American Development Bank, and Vanderbilt University) for making the LAPOP data available. I am solely responsible for any errors.
Gregory S. Schober is an Assistant Professor in Rehabilitation Sciences at The University of Texas at El Paso. His research interests are at the intersection of social policy, political and civic behavior, and global health. He is the coauthor of published articles in Political Behavior, Journal of Development Studies, and other journals.
Aldrich, John H. 1993. “Rational Choice and Turnout.” American Journal of Political Science 37 (1): 246–278. DOI: https://doi.org/10.2307/2111531
Angelucci, Manuela, and Orazio Attanasio. 2009. “Oportunidades: Program Effect on Consumption, Low Participation, and Methodological Issues.” Economic Development and Cultural Change 57 (3): 479–506. DOI: https://doi.org/10.1086/596616
Baez, Javier E., Adriana Camacho, Emily Conover, and Román A. Zárate. 2012. Conditional Cash Transfers, Political Participation, and Voting Behavior. World Bank Policy Research Working Paper 6215. http://elibrary.worldbank.org/content/workingpaper/10.1596/1813-9450-6215. DOI: https://doi.org/10.1596/1813-9450-6215
Bohn, Simone R. 2011. “Social Policy and Vote in Brazil: Bolsa Família and the Shifts in Lula’s Electoral Base.” Latin American Research Review 46 (1): 54–79. DOI: https://doi.org/10.1353/lar.2011.0003
Booth, John A., and Mitchell A. Seligson. 1978. “Images of Political Participation in Latin America.” In vol. 1 of Political Participation in Latin America: Citizen and State, edited by John A. Booth and Mitchell A. Seligson, 3–33. New York: Holmes and Meier Publishers.
Booth, John A., and Mitchell A. Seligson. 2009. The Legitimacy Puzzle in Latin America: Political Support and Democracy in Eight Nations. Cambridge: Cambridge University Press. DOI: https://doi.org/10.1017/CBO9780511818431
Bruch, Sarah K., Myra Marx Ferree, and Joe Soss. 2010. “From Policy to Polity: Democracy, Paternalism, and the Incorporation of Disadvantaged Citizens.” American Sociological Review 75 (2): 205–226. DOI: https://doi.org/10.1177/0003122410363563
Campbell, Andrea L. 2003. How Policies Make Citizens: Senior Political Activism and the American Welfare State. Princeton: Princeton University Press. DOI: https://doi.org/10.1515/9781400841318
Cecchini, Simone, and Aldo Madariaga. 2011. Conditional Cash Transfer Programmes: The Recent Experience in Latin America and the Caribbean. Santiago, Chile: United Nations, ECLAC. DOI: https://doi.org/10.2139/ssrn.1962666
Corrêa, Diego Sanches. 2015. “Conditional Cash Transfer Programs, the Economy, and Presidential Elections in Latin America.” Latin American Research Review 50 (2): 63–85. DOI: https://doi.org/10.1353/lar.2015.0020
De La O, Ana L. 2013. “Do Conditional Cash Transfers Affect Electoral Behavior? Evidence from a Randomized Experiment in Mexico.” American Journal of Political Science 57 (1): 1–14. DOI: https://doi.org/10.1111/j.1540-5907.2012.00617.x
De La O, Ana L. 2015. Crafting Policies to End Poverty in Latin America: The Quiet Transformation. New York: Cambridge University Press. DOI: https://doi.org/10.1017/CBO9781316105528
Díaz-Cayeros, Alberto, and Beatriz Magaloni. 2009. “Aiding Latin America’s Poor.” Journal of Democracy 20 (4): 36–49. DOI: https://doi.org/10.1353/jod.0.0115
Díaz-Cayeros, Alberto, Federico Estévez, and Beatriz Magaloni. 2016. The Political Logic of Poverty Relief: Electoral Strategies and Social Policy in Mexico. New York: Cambridge University Press. DOI: https://doi.org/10.1017/CBO9781316492710
Fabrigar, Leandre R., Duane T. Wegener, Robert C. MacCallum, and Erin J. Strahan. 1999. “Evaluating the Use of Exploratory Factor Analysis in Psychological Research.” Psychological Methods 4 (3): 272–299. DOI: https://doi.org/10.1037/1082-989X.4.3.272
Finkel, Steven E. 2002. “Civic Education and the Mobilization of Political Participation in Developing Democracies.” Journal of Politics 64 (4): 994–1020. DOI: https://doi.org/10.1111/1468-2508.00160
Fiszbein, Ariel, and Norbert Schady. 2009. Conditional Cash Transfers: Reducing Present and Future Poverty. Washington, DC: World Bank. DOI: https://doi.org/10.1596/978-0-8213-7352-1
Fried, Brian J. 2012. “Distributive Politics and Conditional Cash Transfers: The Case of Brazil’s Bolsa Família.” World Development 40 (5): 1042–1053. DOI: https://doi.org/10.1016/j.worlddev.2011.09.022
Garay, Candelaria. 2007. “Social Policy and Collective Action: Unemployed Workers, Community Associations, and Protest in Argentina.” Politics and Society 35 (2): 301–328. DOI: https://doi.org/10.1177/0032329207300392
Gelman, Andrew, and Jennifer Hill. 2007. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge: Cambridge University Press. DOI: https://doi.org/10.1017/CBO9780511790942
Gertler, Paul. 2000. “Final Report: The Impact of Progresa on Health.” Washington, DC: International Food Policy Research Institute. www.ifpri.org/publication/impact-progresa-health.
Holzner, Claudio A. 2010. Poverty of Democracy: The Institutional Roots of Political Participation in Mexico. Pittsburgh: University of Pittsburgh Press. DOI: https://doi.org/10.2307/j.ctt6wrdtx
Hu, Li-tze, and Peter M. Bentler. 1999. “Cutoff Criteria for Fit Indexes in Covariance Structure Analysis: Conventional Criteria Versus New Alternatives.” Structural Equation Modeling 6 (1): 1–55. DOI: https://doi.org/10.1080/10705519909540118
Hunter, Wendy, and Natasha Borges Sugiyama. 2014. “Transforming Subjects into Citizens: Insights from Brazil’s Bolsa Família.” Perspectives on Politics 12 (4): 829–845. DOI: https://doi.org/10.1017/S1537592714002151
Hunter, Wendy, and Timothy J. Power. 2008. “Rewarding Lula: Executive Power, Social Policy, and the Brazilian Elections of 2006.” Latin American Politics and Society 49 (1): 1–30. DOI: https://doi.org/10.1353/lap.2007.0005
Imai, Kosuke, Gary King, and Carlos Velasco Rivera. 2017. “Do Nonpartisan Programmatic Policies Have Partisan Electoral Effects? Evidence from Two Large Scale Randomized Experiments.” Working paper. http://gking.harvard.edu/publications/do-nonpartisan-programmatic-policies-have-partisan-electoral-effects-evidence-two.
Klesner, Joseph L. 2009. “Who Participates? Determinants of Political Action in Mexico.” Latin American Politics and Society 51 (2): 59–90. DOI: https://doi.org/10.1111/j.1548-2456.2009.00048.x
Layton, Matthew L., and Amy Erica Smith. 2015. “Incorporating Marginal Citizens and Voters: The Conditional Electoral Effects of Targeted Social Assistance in Latin America.” Comparative Political Studies 48 (7): 854–881. DOI: https://doi.org/10.1177/0010414014565889
Lindert, Kathy, Emmanuel Skoufias, and Joseph Shapiro. 2006. Redistributing Income to the Poor and the Rich: Public Transfers in Latin America and the Caribbean. Social Protection Discussion Paper no. 605. Washington, DC: World Bank. http://siteresources.worldbank.org/SOCIALPROTECTION/Resources/SP-Discussion-papers/Safety-Nets-DP/0605.pdf
Manacorda, Marco, Edward Miguel, and Andrea Vigorito. 2011. “Government Transfers and Political Support.” American Economic Journal: Applied Economics 3 (3): 1–28. DOI: https://doi.org/10.1257/app.3.3.1
Mettler, Suzanne, and Jeffrey M. Stonecash. 2008. “Government Program Usage and Political Voice.” Social Science Quarterly 89 (2): 273–293. DOI: https://doi.org/10.1111/j.1540-6237.2008.00532.x
Niemi, Richard G. 1976. “Costs of Voting and Nonvoting.” Public Choice 27 (1): 115–119. DOI: https://doi.org/10.1007/BF01718955
Niemi, Richard G., Stephen C. Craig, and Franco Mattei. 1991. “Measuring Internal Political Efficacy in the 1988 National Election Study.” American Political Science Review 85 (4): 1407–1413. DOI: https://doi.org/10.2307/1963953
Pierson, Paul. 1993. “When Effect Becomes Cause: Policy Feedback and Political Change.” World Politics 45 (4): 595–628. DOI: https://doi.org/10.2307/2950710
Riker, William H., and Peter C. Ordeshook. 1968. “A Theory of the Calculus of Voting.” American Political Science Review 62 (1): 25–42. DOI: https://doi.org/10.1017/S000305540011562X
Rocha Menocal, Alina. 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 (3): 513–538. DOI: https://doi.org/10.1017/S0022216X01006113
Soss, Joe. 1999. “Lessons of Welfare: Policy Design, Political Learning, and Political Action.” American Political Science Review 93 (2): 363–380. DOI: https://doi.org/10.2307/2585401
Watson, Sara. 2015. “Does Welfare Conditionality Reduce Democratic Participation?” Comparative Political Studies 48 (5): 645–686. DOI: https://doi.org/10.1177/0010414014556043
Zhang, Guangjian. 2014. “Estimating Standard Errors in Exploratory Factor Analysis.” Multivariate Behavioral Research 49 (4): 339–353. DOI: https://doi.org/10.1080/00273171.2014.908271
Zucco, Cesar. 2013. “When Payouts Pay Off: Conditional Cash Transfers and Voting Behavior in Brazil 2002–10.” American Journal of Political Science 57 (4): 810–822. DOI: https://doi.org/10.1111/ajps.12026
Zucco, Cesar, Juan Pablo Luna, and O. Gokce Baykal. 2019. “Do Conditionalities Increase Support for Government Transfers?” Journal of Development Studies, published online February 18. DOI: https://doi.org/10.1080/00220388.2019.1577388