Bloc Voting for Electoral Accountability

Abstract

How do citizens hold local politicians accountable? I argue that citizens, especially through neighborhood associations, can use bloc voting as a bottom-up, grassroots strategy to pressure politicians for public services. Politicians monitor polling station voting, and communities switch allegiance if politicians do not deliver. I measure the perceived and actual relationships between community characteristics, bloc voting, and water access—an essential resource prone to political manipulation. I analyze an original household survey and conjoint experiment merged with electoral data in rural Brazil, and qualitative interviews illustrate theoretical mechanisms. Bloc voting is more likely in communities with high trust and participation, and bloc voting improves water access for association members. However, this strategy is only worthwhile for communities that can demonstrate their vote at their polling station. In contrast to top-down explanations of bloc voting, I highlight the interaction of collective action and electoral institutions for accountability and public service provision.

INTRODUCTION

In much of the world, access to essential public services can vary dramatically from neighborhood to neighborhood. In one rural Brazilian community, residents blamed low participation in the community association and lack of support from city officials when the well’s water pump remained broken for months. Ten kilometers down the road, residents of an otherwise similar community said that if their pump breaks, their association president contacts the city councilor that the community supported last election, who gets it fixed.

Scholars studying accountability and service provision find evidence that politicians target services to groups that vote in coordinated blocs (Gottlieb and Larreguy 2020). They argue that politicians or group leaders act as brokers to coordinate or coerce the bloc vote in a top-down fashion (Auerbach and Thachil 2018; Baldwin 2015; Koter 2013; Larreguy, Marshall, and Querubin 2016; Novaes 2017; Stokes et al. 2013). However, these theories cannot explain community-based mobilization that I observed during fieldwork.

I argue that groups can use bloc voting as a bottom-up, grassroots strategy. By combining two key mechanisms of collective action (Olson 1965) and voting (Przeworski, Stokes, and Manin 1999), bloc voters pressure politicians to provide public services. In local elections, a small group often plays a pivotal role, and if politicians fail to deliver, the bloc switches its votes to a different candidate. The credible threat of switching makes it more likely for politicians to respond to demand-making between elections.

However, this strategy is not possible everywhere; coordination mechanisms and electoral institutions must align. Bloc voting requires groups to coordinate a secret, individual action, and it requires politicians to monitor a group’s vote; therefore, it is most likely where group members have high trust and participation in local associations and vote at their own polling station. Community associations in particular provide a platform for residents to organize around voting and public services such as the local water system, health clinic, or roadways. This strategy is most likely in democratic settings with unreliable public service provision, limited government resources, and local civic engagement. These conditions are most prevalent in lower-income democracies in the Global South but may also hold in non-democracies and the Global North.

My study takes place in rural, semi-arid Northeast Brazil and evaluates community-level, sub-municipal variation that is difficult to measure. I focus on household water service, which is an essential and often scarce resource that requires public investment and is prone to political manipulation (Björkman 2015; Carlitz 2017; Herrera 2017). My theory emerged from 104 qualitative interviews and consultation with rural residents, local leaders, bureaucrats, and scholars in the state of Ceará, Brazil in 2016 and 2017.¹ I test my hypotheses from this inductive theory-building process through multiple methods. To differentiate between top-down and bottom-up explanations for bloc voting, I use a conjoint survey experiment to evaluate how residents perceive the relative importance of community characteristics. Next, I evaluate the relationship between community trust, bloc voting, and water security through an original household survey in 120 rural communities merged with polling station electoral data. Lastly, I study long-term voting patterns across Ceará during five municipal elections. I unpack mechanisms with additional data from 42 interviews with residents, election officials, and city council staffers in 2022.

In contrast to top-down explanations of bloc voting, I find that rural residents perceive community participation to be more important for bloc voting than political endorsements by an association leader acting as a broker. Bloc voting is most likely where residents have stronger trust and vote at the same polling station. These distinctions matter for daily life—bloc voting is associated with more reliable, secure water access, but only for association members. Some groups saw bloc voting as so important that they petitioned to have a polling station in their community so residents could clearly demonstrate their support for a specific candidate. Large-scale electoral data show that communities are consistent in using bloc voting over time, and many communities switch allegiance across elections. These results indicate that social and institutional factors interact to shape bloc voting, and communities are credible in their threats to switch their electoral support if they do not get the services they need.

I contribute a novel explanation of why we observe bloc voting, and my findings show how grassroots bloc voting is a mechanism through which marginalized groups hold politicians accountable. I use in-depth fieldwork with rural residents to show how groups with seemingly little influence can strategically improve their living conditions. However, benefits may not be shared universally within a community, and groups unable to execute the bloc voting strategy were frustrated that they were left behind. Residents that succeeded in this strategy were frustrated that they had to mobilize extensively for basic public services. The finding that accountability for organized groups drives unequal outcomes within and across communities has important implications for democratic quality.

I also contribute an explicit focus on group dynamics and unpack the social and institutional conditions under which some groups use bloc voting to improve public service access. In contrast, most scholarship on local distributive politics and service provision studies individuals (Hicken and Nathan 2020), and earlier work on clientelism focused on top-down, coercive relationships among individual voters, brokers, and politicians (Stokes et al. 2013). Recent work studies bottom-up or demand-side clientelistic strategies by individuals (Borges Martins da Silva 2023; Kao, Lust, and Rakner 2017; Nichter 2018; Nichter and Nunnari 2022; Oliveros 2016; Pellicer et al. 2022). Likewise, many studies on demand- or claim-making and constituency service focus primarily on individuals (Bussell 2018; Calvo and Murillo 2019; Kruks-Wisner 2018) or group leaders as brokers (Auerbach 2019; Auerbach and Thachil 2018; Brierley and Nathan 2021; Krishna, Rains, and Wibbels 2020; Larreguy, Marshall, and Querubin 2016; Paniagua 2022; Zarazaga 2014). My focus on neighborhood groups thus emphasizes the role of collective action in local distributive politics, with key implications for equity in democratic accountability and essential service provision for marginalized groups.

THEORY

Most scholarship in distributive politics focuses on politicians and brokers. I contribute that citizens use bloc voting as a grassroots strategy to hold local politicians accountable for providing public services; this is a long-term relationship between group members, group leaders, and politicians that combines actions before elections with demand-making between elections (Figure 1).

Figure 1. Accountability Cycle

Before the local election, a group of voters decides to coordinate its votes in a particular candidate. The group leader—also a member of the group—invites candidates to speak to members, and candidates compete for the group’s votes. Members consider the campaign promises, discuss their options with the leader, and decide whom to support. After voting ends, members, leaders, and politicians see which groups followed through and which candidates won by monitoring public data for aggregate votes of small, identifiable groups at polling stations.

Between elections, groups make demands on the state by reaching out to candidates and local politicians. Winning candidates are now elected politicians, and losing candidates are embedded in local social and political networks and act as intermediaries to politicians at different levels of government. Politicians or losing candidates decide whether or not to respond to a group’s requests through unilateral actions or pressure on bureaucrats. Since the group is a potential source of pivotal votes in the next election, other politicians may target services to coordinated groups to attract instead of punish them. As the next election approaches, the group evaluates politicians’ promises and actions, and the cycle repeats.

I focus on community or neighborhood associations, which can shape voting and collective action due to their geospatial focus. They are a common local organization across the Global South and North (Auerbach 2019; Berry, Portney, and Thomson 2002; Boulding and Holzner 2021; Constantino, Cooperman, and Muñoz 2023; Read 2012). Most associations hold regular meetings and leadership elections, and they are often citizens’ most frequent interaction with democratic processes. Association leaders, who are residents themselves, play dual roles as development brokers that advocate for community interests and independent vote brokers unattached to a particular party (Holland and Palmer-Rubin 2015; Muñoz 2014; Novaes 2017; Rizzo 2019).

Many studies document the relationship between community associations and politicians (Section B11 of the Supplementary Material; see Cooperman 2023). However, small groups vary significantly in their ability and willingness to engage in collective action, with implications for political behavior and public service provision (Collier and Handlin 2009; Grindle 2007; Paniagua 2022; Rains 2021). I contribute a novel bottom-up mechanism for bloc voting and show how local electoral institutions shape groups’ ability to use this strategy.

Mechanisms and Hypotheses

Democratic accountability often focuses on two mechanisms of voting (electoral) and collective demand-making (coordination), each with its limitations. Most studies focus on one mechanism, but I combine them to argue that bloc voting for accountability is most likely to succeed where groups coordinate and have their aggregate votes monitored. I outline mechanisms in Figure 2 and test hypotheses regarding concepts in the bolded boxes.

Figure 2. Theory Map

Ability to Coordinate Group Action

In the coordination mechanism, citizens organize in groups to use collective action to pressure politicians in public or private (Olson 1965). However, free-riding jeopardizes coordination efforts, and individual action is often not worth it. Nevertheless, communities with high participation in civil society organizations and strong feelings of unity and reciprocity are more likely to advocate for better public services or provide them independently (Putnam, Leonardi, and Nanetti 1993). Strong leadership, informal accountability, and community institutions enable groups to make collective choices and monitor and sanction members (Olson 1965; Ostrom 1990; Tsai 2007).

High activity associations have high participation in group meetings, which enable members discuss their collective vote choice (Gottlieb 2016), and group promises are more credible due to trust and reciprocity (Ravanilla, Haim, and Hicken 2022). Low activity associations have weaker feelings of trust, unity, and reciprocity. Where communities have multiple groups due to spatial distribution of residents or interpersonal divisions, it will be more challenging to coordinate group actions that span the community.

Bloc voting is one of a variety of collective actions that active groups pursue, including participating in council meetings, visiting government offices, and protesting. Bloc voting is appealing because it takes less time or energy than other actions, especially if members would vote anyway. On its own, bloc voting signals the potential of the group as a future vote base; it also increases the effectiveness of other collective actions or demands from members or leaders. Still, bloc voting requires time, discussion, and challenging coordination.

Hypothesis 1. Groups with higher trust and association participation are more likely to concentrate their votes in one local candidate.

Ability to Monitor Group Vote

In the electoral mechanism, voting in free and fair elections is the primary tool for individuals to hold office-seeking politicians accountable in democracies (Manin, Stokes, and Przeworski 1999). Citizens use their votes to reward or punish incumbents and/or select a challenger who promises to fulfill voter preferences, including service provision. However, voting is an infrequent blunt instrument, which makes it hard to hold politicians accountable for policy outcomes, especially public goods shared across users (Przeworski, Stokes, and Manin 1999). An individual vote is rarely pivotal, so politicians have little interest in the actions of a given voter.

Politicians care about groups, and collective voting behavior is often publicly observable at electoral sections or polling stations with fewer than two hundred registered voters (Auerbach 2019; Gottlieb and Larreguy 2020; Rueda 2017). Electoral institutions, especially the distribution of polling stations, affect politicians’ ability to monitor broker behavior (Larreguy, Marshall, and Querubin 2016).

I contribute that electoral institutions also affect community members’ ability and incentive to coordinate. Depending on how polling stations are distributed across the municipality, community members may vote at one polling station (with or without residents of neighboring communities) or be divided across multiple polling stations. When community members vote at their own polling station, the outcome of bloc voting is more observable and impactful.

Hypothesis 2. Groups where members vote at the same polling station are more likely to concentrate their votes in one local candidate.

Impact of Bloc Voting on Service Provision

Once a group succeeds in bloc voting, electoral and coordination mechanisms shape service provision. First, the group is a valuable vote base. Politicians make decisions based on the prior election’s results and expectations for the next election (Gottlieb and Larreguy 2020). If Community A concentrated its one hundred votes in Politician A, then Politician A will want to keep those votes in the next election. Politician B may also have an incentive to respond to Community A’s demands and secure future votes. Second, the group signals it can coordinate and mobilize between elections; politicians do not want to face public protest and possibly lose support from those or other voters.

The definition of a sufficient or effective bloc vote varies by context. Politicians want a large vote base (Auerbach 2019) and a reliable one, but these features may be at odds due to challenges of coordinating large groups (Olson 1965). Coordinated small groups may be more likely to protest and make public demands.

Bloc voting signals a group’s coordination capacity and its potential as a future vote base, so politicians are more likely to prioritize investment in public services to communities that bloc vote.

Hypothesis 3. Individuals in groups that concentrated their vote in one local candidate have better public service access.

For bloc voting to incentivize politicians to provide services, communities must credibly commit to bloc voting in the future and credibly threaten to switch candidates. Community activity and leadership are “sticky” and change slowly (Sampson and Graif 2009), so I expect bloc voting behavior to be consistent over time. My theory assumes that communities switch their loyalty even if their previous most voted candidate is running; I explore this empirically.

Scope Conditions

Bloc voting for accountability is most likely as a grassroots strategy for local elections in contexts that have scarce resources and unreliable service provision, free and fair elections with polling station results, and local civic engagement and participation.

My theory applies to local public goods and services that provide salient collective benefits and are (or could be) delivered by the state; applicable services are revocable or require maintenance, targetable to certain groups and excludable, and attributable to a politician (Batley and Mcloughlin 2015; Post, Bronsoler, and Salman 2017). Depending on service provision systems, they include water, pavement and transportation, health and education, electricity, trash collection, and so forth.

Bloc voting takes time and effort, so it is most likely as a grassroots strategy in contexts that have unreliable service provision and where politicians allocate scarce resources. These conditions are more likely in lower-income settings, but there is high variation even within the United States (US) (Wutich et al. 2022). Variation in political participation and geographic representation shape local government responsiveness to service requests and service provision in the US and Europe (Epstein, Bode, and Connolly 2023; Harjunen, Saarimaa, and Tukiainen 2023; Rasmussen and Reher 2019). My theory applies in rural or urban areas where politicians can prioritize services by village, neighborhood, or street (Auerbach 2019; Paniagua 2022; Post 2018).

Electoral accountability is most likely in democracies with regular, free, and fair elections and where there are multiple competitive candidates or parties. Voters must be able to credibly threaten to switch their allegiance in a future election. My argument applies to regional or nationalized parties that use programmatic or non-programmatic strategies, as long as programmatic politicians have discretion over which eligible communities receive services first. In semi-authoritarian or authoritarian countries, public service provision can vary depending on local political relationships (Martinez-Bravo et al. 2022; Tsai 2007), but politicians may be less responsive than in democracies. Bloc voting is most likely and effective in local elections since small groups can more easily contact candidates and are more likely to be pivotal voters.

Politicians target services and monitor voters within territorial areas, so the strategy is most effective where polling station data are publicly available. Still, communities organize vote blocs along racial, ethnic, or religious identities (Hajnal and Trounstine 2014), and politicians target club goods and monitor groups that vote for identity-based parties (Cammett and MacLean 2014).

The bottom-up elements of the strategy require that citizens organize freely and can form local associations. Latin America has robust civil society and political participation across rich and poor sectors (Avritzer 2007; Collier and Handlin 2009), where 77% of poor individuals participate in a community organization (Boulding and Holzner 2021). However, neighborhood associations exist throughout the world in democratic and non-democratic contexts and in high- and low-income countries (Auerbach 2019; Berry, Portney, and Thomson 2002; Read 2012).

CONTEXT

Located in the poorest region of Brazil, the Northeast state of Ceará had well-regarded bureaucratic reforms in the 1990s (Tendler 1997). Nevertheless, service provision varies across and within municipalities, which have around 50–300 rural communities outside the city center, some of which are very isolated. Communities are groups of people who self-identify as residents of a sub-municipal geographic area (e.g., neighborhood or locality). Rural communities often have 20–200 households; they are quite homogeneous in terms of income and race.

Community Associations

Community associations mobilize citizens for collective goals and development programs, and most focus on water resource management in this semi-arid region (Enéas da Silva et al. 2013). In 2010, community associations were one quarter of non-profit organizations, with an average of 22 community associations per municipality in the state of Ceará (IBGE 2012); in rural areas, I observed higher numbers because registered associations are required to access many government development programs. A participatory map created during a water management workshop shows a typical community (Figure 3) (Cooperman, McLarty, and Seim 2021). Households are black squares, and the community association is the prominent building in the center.

Figure 3. Participatory Map

Source: Photo by research assistant, 2018.

Associations conduct initiatives such as mobilizing the community to build rainwater cisterns, organizing collective work days, and enrolling community members in government programs. Associations improve legibility in both directions: they help rural citizens make demands on the state and enable the state to get information and target services. Associations are not substitutes for the state; they supplement public provision and provide local expertise (Lopez and Abreu 2014).

Most associations hold regular elections to appoint a leadership board. The association leaders in my sample are not traditional wealthy, landed elites; while they tend to be older, male, have more household assets, and be more educated than members, the mean leader still falls below a middle school education level (Supplementary Table B1).

Associations vary significantly in social capital and participation, most likely due to state-driven and society-driven processes (Fox 1996). Most Brazilian community associations were created in the 1990s to 2000s through state-society partnerships for local development. Associations leveraged social capital developed by existing organizations (Teixeira 2008) such as Christian Base Communities (CEBs) that advocated for human rights during the military dictatorship (Braga and Barreira 1991; Mainwaring 1984) and the Landless Workers Movement (MST) (Wolford 2010). I find that communities with a higher share of residents that participated in activities with CEBs in the past tend to have higher community trust (Supplementary Table B2).

Elections and Electoral Institutions

Municipal elections occur every 4 years and are staggered by 2 years from state/federal elections. Voting is mandatory for literate individuals aged 18–70, and turnout in the 2016 municipal election averaged 82% across Brazil, 85% in Ceará, and 82% in the sample (Supplementary Figure B6).

Brazil’s Tribunal Superior Eleitoral administers elections within zones that generally correspond to municipalities outside state capitals. The zone’s electoral judge divides it into sections allocated to a voting machine (urna) at a polling station (local de votação). A rural polling station often hosts one or two sections, and urban polling stations host multiple sections.

The zone’s electoral judge decides the distribution of polling stations. Polling stations are usually in public buildings and must meet standards for electricity, safety, physical conditions, and accessibility (Brazilian Electoral Code: Law No. 4.737/1965, Art. 135–138). They must support an electoral section with a minimum of 50 registered voters. Citizens can submit a request for a new polling station that meets these requirements to the electoral judge. Voters are assigned to a section at a polling station near their residence and can transfer if they move (Section B7 of the Supplementary Material).

Votes are publicly reported for each voting machine, which usually has one section. Election staff post poll tapes with candidate totals at each machine (boletim de urna) in front of the polling station after the polls close. Bloc voting revolves around the polling station and focuses on the most voted candidate. Rural residents told me which council candidates were the “most voted” by their community, including citing accurate polling station vote counts. A rural resident said, “Our community has two sections that vote at the school. As soon as people vote, the whole world knows…People working for different politicians know how many votes their politician should expect, and the politician will find out how many votes he got” (Interview 59).

Brazilian municipalities have a mayor and a city council. Since the mayor runs the entire municipality, councilors are more accessible, especially for rural residents (Nichter 2018). Councilors are elected at-large via open-list proportional representation, so candidates target specific communities or seek votes throughout the municipality. This system creates high numbers of candidates, and the median municipality in Ceará had 53 candidates for 13 city council seats in 2016. The difference between winning and losing a city council seat can be fewer than five votes.

Public Services

Communities usually request services for which the government is responsible but has not provided due to lack of information, lack of resources, or selective performance. The city council is primarily responsible for discussing municipal laws and budget, and councilors request that the mayor allocate funds for public services to specific neighborhoods. These requests are formalized through a solicitation letter (requerimento/indicação) or through direct contact between councilors and municipal bureaucrats, the mayor, and/or state or federal deputies. These actors have electoral incentives to respond, even municipal bureaucrats, who may use these relationships to launch political careers (Boas et al. 2022).

Residents use collective action and targeted requests to improve public service provision. Community members and leaders in Northeast Brazil protest on the radio, visit city hall, and appeal to politicians, bureaucrats, or organizations to increase their chances (Medeiros 2012; Teixeira 2008). Though long-lasting water resources like cisterns reduce residents’ reliance on clientelistic politicians (Bobonis et al. 2022; Frey 2022), residents said they relied on association leaders and councilors to maintain essential services like wells, roadways, and lights.

RESEARCH OVERVIEW

Brazil has rich data for its 5,570 municipalities, but little data exist for sub-municipal phenomena. Rural communities are not official administrative units; census tracts merge most rural communities.

I thus collected original data through household surveys, which were conducted in 2017 and 2019 in 120 communities across 10 municipalities during a field experiment about community water resource management (Cooperman, McLarty, and Seim 2021; Slough et al. 2021). The survey sample is composed of rural communities with existing associations that use wells in the semi-arid region of Ceará.² In each community, Brazilian enumerators surveyed (1) heads of households in the most populated area; (2) heads of households in more isolated areas; (3) association leaders; (4) water system operators; and (5) landowners. Surveys were conducted face-to-face, and all respondents provided oral consent.

I first report the methods and results from a conjoint experiment to contrast residents’ perceptions of bottom-up versus top-down bloc voting. Next, I report the methods and results from the observational analysis of survey results merged with electoral data. Last, I report methods and results using section-level data across the state from 2000 to 2016. To illustrate the main mechanisms, I include anecdotes from interviews that I conducted in similar rural communities (Supplementary Tables A17–A22), which I triangulate with electoral results and public records.

Perceptions of Bloc Voting: Conjoint Experiment

Do rural residents think that community characteristics matter for bloc voting? I use a conjoint experiment to provide causal evidence that rural residents perceive that bottom-up mechanisms of community coordination cause bloc voting.³

Conjoint experiments have a complex design that is difficult to implement in rural field settings with low literacy rates (Cooperman, McLarty, and Seim 2022). I, therefore, created pictogram booklets⁴ with five attributes of a community association that are typical, but also vary, in this context: (1) active participation in association meetings, (2) leadership that is responsive to community members, (3) endorsement of a local political candidate by the association president, (4) high competition for the position of association president, and (5) turnover in the association leadership. All combinations are plausible and independent, which led to 32 unique profiles.

The enumerator first read a preface then turned the physical booklets to the profiles randomly selected by the tablet (Figure 4). As the respondent held the two booklets on their lap, the enumerator read the text (Supplementary Table B12) of the profiles’ options so that respondents had consistent interpretations of the pictograms. Finally, the enumerator asked the respondent four follow-up questions, including: “In your opinion, in municipal elections, which of these communities would organize to vote for just one candidate?”⁵

Figure 4. Sample Profiles

The conjoint experiment was conducted with rural residents and association leaders in the 2019 survey; the full sample has 1,745 respondents in 120 communities. I estimate the average marginal component effect on the choice outcome (Hainmueller, Hopkins, and Yamamoto 2014). Standard errors are clustered by respondent ID.

Participation Drives Bloc Voting

Residents perceive a strong impact of participation in association meetings (Hypothesis 1): going from weak (0) to strong participation (1) leads to a 27-percentage-point (pp) increase in the perceived likelihood of bloc voting (Figure 5). Residents perceive a strong impact of having responsive leadership: going from weak responsiveness (0) to strong responsiveness (1) leads to a 17-pp increase in the perceived likelihood of bloc voting (Figure 5).

Figure 5. Direct Effects of Community Association Features on Bloc Voting

Note: Outcome reflects whether the respondent selected a community profile with that characteristic as more likely to engage in bloc voting. Results show average marginal component effects, $ n=2,478 $ . Standard errors clustered by respondent: 1,239 clusters. Plot shows 95% confidence interval. No controls. Coefficients in Supplementary Table A4.

These findings highlight that residents perceive that community participation and leadership responsiveness are much stronger drivers of bloc voting than formal endorsement of a candidate by the association president, which leads to an 8-pp increase.⁶ If bloc voting were only coerced from above, residents would likely perceive endorsement to be a more influential category.

In qualitative interviews, residents described the benefits and challenges of coordinating their vote as a community. With many candidates and one vote, citizens look to familial, social, and organizational networks for information (Smith 2018). Active community associations often allow city council candidates to present at monthly meetings. Leaders remind group members of the benefits of coordinating their vote to have a representative in municipal government (Reis 1988); one association president said he organizes a community meeting to coordinate the group so that they will have more bargaining power vis-à-vis councilors to obtain permanent, collective resources such as a well (Interview 128).

Residents in both high and low trust communities perceive participation to be the strongest driver of bloc voting (Supplementary Table A5), which is consistent with my interviews. A rural resident said that the community discussed whom they wanted to vote for: “Yes, we [got together to vote for him]. We see what the [candidate] talks about, and then we give a vote of confidence in that person and then will see if the person [follows through]” (Interview 60). Still, that resident emphasized their autonomy in the voting process (Interview 60). A resident of a different community similarly said, “The community does meet to discuss candidates, but no one is obliged to vote for anyone. Everyone makes a decision on their own” (Interview 115).

Respondents in other communities were aware that other communities used bloc voting as a grassroots strategy but lamented that their community was unable to. Many explained that group members were disengaged from community affairs and succumbed to clientelistic pressures. In a community that dispersed its votes, a resident said, “The community does not have councilor. We tried to have one here, but we didn’t succeed in the last election…Others always come from afar, who knows from where, asking for votes, even buying votes here…Community members don’t discuss things in association meetings…The association is not very active” (Interview 63). I find empirical support for the negative relationship between trust and vote-buying in Supplementary Tables B9 and B10, described later.

Bloc Voting: Observed Relationships

I next evaluate the observed interaction between collective action and electoral mechanisms using 2017 survey data with 1,990 rural residents and association leaders merged with 2016 municipal election results.

To measure bloc voting, I first calculate each candidate’s polling station vote share: the number of votes candidate A received at polling station p divided by the total number of votes for all city council candidates at polling station p. ⁷ Next, I identify a community’s primary polling station by selecting the modal response among respondents’ reported 2016 polling station. My main measure of bloc voting is the maximum of the vote share for every candidate at the community’s primary polling station (Vote Share in Most Voted CC Member). This measure had specific salience in rural communities in my interviews. As alternative measures, I calculate a Herfindahl index and a two-candidate concentration index.

I account for the ability to monitor the group’s vote by calculating the Share of Respondents at the Polling Station most mentioned in the community (Supplementary Figures B3 and B4). I calculate the total number of valid votes for city council candidates at the polling station (Total Votes at Polling Station [ln]).

To measure coordination, Community Trust in Others is a binary variable that is 1 if respondent says people in the community are very trustworthy or trustworthy, and 0 if not very or not at all.

Community relations, bloc voting, and polling station distribution may be endogenous in the long run. When residents observe successful bloc voting via poll tapes, this could increase their trust, or vice versa. High trust communities may also be more likely to petition to have their own polling stations. However, my interviews suggest that petitions are not widespread, and exogenous factors related to site conditions are primary drivers of polling locations; community trust and share of respondents at the main polling station are not significantly correlated ( $ \rho =-0.09 $ , $ p=0.33 $ ).

In the observational survey data merged with election and geospatial data, some variables are collected at the individual level, whereas others are measured at the community level. In model 1, I aggregate fine-grained individual-level data to the community level to test hypotheses about coordination and electoral mechanisms:

(1) $$ \begin{array}{l}Blo{c}_{cm}={\beta}_1Trus{t}_{cm}+{\beta}_2{PollShare}_{cm}\\ {}\hskip42pt +\hskip2pt {\beta}_3Trus{t}_{cm}\times PollShar{e}_{cm}+\boldsymbol{\Omega} {\mathbf{X}}_{cm}+{\alpha}_m+{\epsilon}_{cm},\end{array} $$

where $ Blo{c}_{cm} $ is the vote share for the most voted city council candidate at the primary polling station in community c in municipality m, $ Trus{t}_{cm} $ is the mean community value for trust in others, $ {PollShare}_{cm} $ captures the share of respondents voting at the primary community polling station, $ {\mathbf{X}}_{cm} $ is a series of control variables, and $ {\alpha}_m $ is a municipal fixed effect.

In models 1 and 2 (next section), I use municipal fixed effects because my theory focuses on variation within municipalities. All models use ordinary least squares regression.

Trust Drives Bloc Voting

Consistent with the coordination mechanism in Hypothesis 1, I find that, on average, communities with higher trust in others are more likely to have higher bloc voting at their main polling station (column 1 of Table 1, $ p<0.05 $ ). However, my interviews suggest that groups with high trust would only be willing to pursue bloc voting if they are able to clearly demonstrate their vote by voting at the same polling station. I, therefore, focus on the interaction between the coordination and electoral mechanisms (Hypothesis 2) in column 2.

Table 1. Coordination and Electoral Mechanisms Interact for Bloc Voting

Note: Includes municipal fixed effects. Controls not shown for: elites attend meetings, current/former CC member lives in community, vote not perceived secret, more than one association, leader proposes ideas, constant leader, total votes at polling station, distance to city center, and household assets index (see Supplementary Table A7). * $ p<0.1 $ , ** $ p<0.05 $ , *** $ p<0.01 $ .

When I interact Trust with Share at Main Polling Station, the interaction term is positive ( $ p=0.07 $ ). When 92% of respondents vote at the main polling station (one standard deviation above the mean), the predicted coefficient on Trust is 0.26: when community members are concentrated at one polling station, higher trust is associated with bloc voting (Figure 6). In contrast, when 52% of respondents vote at the main polling station (one standard deviation below the mean), the predicted coefficient on Trust is 0.03; when community members are dispersed across polling stations, higher trust is not associated with bloc voting. I find consistent results for the Herfindahl index, but results are not statistically significant for vote share in the top two most voted candidates (Supplementary Table A8).

Figure 6. Bloc Voting

Note: Model from column 2 of Table 1.

The results are consistent with my theory that it is challenging for politicians to monitor the collective votes of community members if they are spread across multiple polling stations; therefore, it is not strategic for communities to coordinate their votes. A rural resident explained, “A community without its own voting machine (urna) is not seen by [politicians]” (Interview 129).

Staff members at the municipal and state electoral offices confirmed that some residents request their own polling stations to prove the community’s voting behavior and reduce transportation time; they said that electoral staff verify and approve requests that meet the minimum physical and population requirements (Interviews 134, 137, and 139). A bureaucrat at the state Election Board (TRE-CE) said,

We try to install polling stations where people live. [We have lots of contact with community association leaders] so that residents can vote for candidates that will effectively serve that community…People have a strong identification with ‘the councilor for the neighborhood’ so that the person will bring improvements to the neighborhood. For the identity of the group to be respected, they want to vote near where they live…It’s very common for candidates for city council or deputy to do their own parallel totals for sections in communities in the areas that they ‘serve’ (Interview 134).⁸

Residents shared similar stories. In a few communities that I visited, residents requested to have their own polling station to demonstrate their loyalty to a candidate (Interviews 61, 113, 114, 118, 119, 128, and 129). Residents said the process was complicated but they ultimately succeeded. They reported needing 53 valid voters because the electoral zone office encouraged them to have more than 50 in case people die or move. They are concerned that if they lose voters, their section will get combined with a neighboring community, so they stay vigilant (Interview 129).

Bloc voting is less likely at larger polling stations ( $ p<0.05 $ , Supplementary Table A7), which are likely to include citizens from either multiple small communities or one large community. In the first case, the votes of different communities are reported together, and politicians are unable to monitor the votes of a specific community. Some residents said they previously voted at a polling station in a neighboring community with over five hundred voters and felt they could not prove their allegiance to a particular candidate. They petitioned to have their own polling station (Interviews 61, 118, and 119), which electoral records corroborate.

Large communities are also much harder to coordinate, especially if they are geospatially divided into smaller neighborhoods or divided along interpersonal lines. Having more than one association could make it harder to coordinate around one candidate; empirically, it is negatively correlated with bloc voting but not statistically significant ( $ p>0.1 $ , Supplementary Table A7).

Limited Evidence for Top-Down Bloc Voting

Is this bloc voting coerced in a top-down fashion? If economic elites coerce political behavior, we should observe more bloc voting when economic elites participate in association meetings. This variable is not statistically significant ( $ p>0.1 $ , Supplementary Table A7).

If elites coerce individuals to follow the group choice by threatening to monitor their individual behavior, we should also observe more bloc voting where voters do not perceive that their individual vote is secret. I note a key distinction between legally monitoring group behavior at the polling station and illegally monitoring or threatening to monitor individual vote choice. The former is a precondition for a grassroots bloc voting, whereas the latter is often associated with (though not required for) individual clientelism and vote-buying (Hicken and Nathan 2020; Stokes 2005). In interviews, rural residents that perceived that their individual vote is not secret were more susceptible to pressure from politicians for individual vote-buying and had less perceived political efficacy. Lack of vote secrecy is positively associated with clientelism and vote-buying ( $ p<0.05 $ , Supplementary Tables B9 and B10) but is not significantly associated with bloc voting ( $ p>0.1 $ , Supplementary Table A7).

The presence or proximity to political elites could influence vote concentration through bottom-up or top-down mechanisms. Community members are more likely to coalesce their bloc vote around candidates with whom they have closer social, geographic, or familial ties (Cruz, Labonne, and Querubin 2017; Ravanilla, Davidson, and Hicken 2022). A political elite living in the community could coerce members but also provides an obvious person around whom to mobilize pre-election and make demands post-election. I find that vote concentration is higher in communities where a higher share of respondents said that a current or past councilor lives in the community ( $ p<0.01 $ , Supplementary Table A7).⁹ Candidates in municipal political dynasties could be more effective at coercing votes (Boas, Hidalgo, and Melo 2019; Querubin 2016).¹⁰ Still, residents evaluate candidates based on their actions, not their political networks (Interview 66). Even controlling for local political elites, trust predicts higher bloc voting.

I also evaluate dependent variables of clientelism and general vote-buying in the community (Section B4 of the Supplementary Material). If clientelistic politicians coerce bloc voting in the same community, we would see a positive correlation between bloc voting and clientelism/vote-buying; instead, I find a negative, statistically insignificant correlation (Supplementary Table B8). Clientelism and vote-buying are more likely in communities with lower trust and where the vote is not perceived to be secret (Supplementary Tables B9 and B10). Vote-buying practices are more likely in communities where a higher share of respondents vote at the main polling station, which could reflect increased ability to monitor broker behavior at the group level (Supplementary Table B10).

Even in a coercive, clientelistic environment (Nichter 2018), bloc voting plays a distinct bottom-up role driven by an interaction of collective action and ability to demonstrate the vote.

Robustness Checks

Could concentrated voting simply reflect that more cohesive and trusting groups have similar preferences and, therefore, independently select the same candidate? This may be true, but if it were the only explanation, we would not see any impact of the electoral mechanism. With multiple candidates per party and a median of 53 candidates, it is unlikely that candidates differentiate themselves enough that 30%–50% of voters would independently coalesce around one.

A sensitivity analysis (Cinelli and Hazlett 2020) of column 1 of Table 1 uses omitted variables with similar impact as key theoretical and empirical predictors: the size of the modal polling station and having a resident current or former councilor. With an omitted variable the same size, the point estimates on trust decrease from 0.14 to 0.1 (Supplementary Figure B7) or 0.13 (Supplementary Figure B8), respectively. When I add categories of control variables to bivariate models of the relationship between trust and bloc voting (Lenz and Sahn 2021), point estimates on trust decrease but remain substantively and statistically significant in the full model (Supplementary Tables B3 and B4).

A limitation of this analysis is that I use household survey data from 2017 to predict voting behavior in 2016.¹¹ While leadership and social dynamics tend to change slowly, voting behavior could drive community characteristics. I thus analyze a much smaller household survey that I implemented weeks before the 2016 municipal election in Ceará with 411 respondents in 104 small communities with and without associations in rural and urban areas. Communities with associations and especially with high satisfaction in the association are more likely to concentrate their votes in the election ( $ p<0.1 $ , Supplementary Table B15).

Water Access: Observational Relationships

The final stage of my theory (Figure 2) predicts that groups that bloc vote will have better public service access. In a drought-prone area, associations coordinate voting preferences and manage rural development programs, many of which focus on water and drought relief (Campos and Studart 2008; Cooperman 2022). Residents routinely said that water is the biggest challenge facing the community, and they rely on a mix of water sources as shown in the participatory map (Figure 3). Rainwater cisterns (dots) are next to houses (squares). Wells surround the community, and a water tower supplies the community’s pipes with untreated well water. Residents use storage tanks and the river for subsistence farming and livestock. Access to each source can break down: the well pump breaks, cisterns get contaminated, or reservoirs or wells dry out.

Water service provision varies across and within communities: piped systems exist only in some areas, water trucks go to certain households, and access to cisterns or drought relief is sometimes contingent on formal association participation. I evaluate whether bloc voting improves water access across the entire community (universal) or based on household participation in the association (selective).

Water access is challenging to operationalize. Most organizations ask whether the household has piped water or an improved water source (WHO 2015), but these data cannot capture how often a system breaks down. An index is more appropriate in settings with multiple sources depending on use and season (Jepson et al. 2017).

I developed a Water Index to capture security and reliability for each household based on interviews with rural residents and leaders. I use measures for access (piped water, rainwater cistern, and satisfaction with access), security (days with water in last month and no reliance on emergency water truck), and satisfaction with quality. Measures are self-reported on the household survey, and I create a z-score index with equal weights (Supplementary Table A3).

In model 2, I analyze individual-level variation in water access, and I evaluate the independent and interactive effects of bloc voting and association membership:

(2)

where $ WaterInde{x}_{icm} $ is the water access security and reliability index for respondent i in community c in municipality m, $ BlocVotin{g}_{cm} $ is the vote share for the most voted city council candidate at the primary polling station, $ AssocMembe{r}_{icm} $ is the household membership in the association, $ {\mathbf{X}}_{icm} $ is a series of control variables, and $ {\alpha}_m $ is a municipal fixed effect. Since individual observations within the same community are not independent, I cluster standard errors at the community level.

Bloc Voting Improves Members’ Water Access

Bloc voting is associated with more secure, reliable water services (Hypothesis 3), but this finding only holds for households with association members (column 2 of Table 2).¹² Water access does not depend on whether the most voted city council candidate won or lost (columns 3 and 4 of Table 2). Bloc voting leads to selective distribution, since community residents who are not association members do not benefit from bloc voting.

Table 2. Water Access and Vote Concentration

Note: Includes municipal fixed effects and clustered standard errors at community level: 120 clusters. Dependent variable is an index in standard deviations. Controls not shown for gender, age, household assets index, and type of respondent (see Supplementary Table A11). * $ p<0.1 $ , ** $ p<0.05 $ , *** $ p<0.01 $ .

I expect that bloc voting helps association members through both the electoral and coordination mechanisms. In the study area, community associations are the primary venue for both coordinating voting behavior and distributing water services. In the coordination mechanisms, bloc voting signals an association’s ability to coordinate group action. Since coordinated associations are more likely to make demands between elections, politicians prioritize them for constituency services. In the electoral mechanism, the association’s votes are a current or potential vote base. The salient level and measure of bloc voting will vary by context. In the study area, having even 30% of the community vote for a single candidate is perceived as bloc voting.

Residents noted that politicians have helped with access to water resources, drought relief, paved roads, and ambulance services. In a community that petitioned to have its own polling station, residents said they leveraged the electoral relationship created through bloc voting to get public services. In interviews in 2017, one resident said the association was able to get a well and trash collection (Interview 60), and another said the community’s candidate promised before the 2016 election to pave the road (Interview 61). Through archival research at the city council, I confirmed that the councilor requested in 2017 that the municipal government pave the community road. I observed the completed road in 2022, and residents attributed it to that councilor and his connections with a state deputy (Interviews 110, 118, and 119).

Communities that bloc vote leverage candidates’ personal or partisan ties to elected officials or bureaucrats. Most parties have weak identities at the municipal level in Brazil (Feierherd 2020), and networks are often based on personal relationships. Winning and losing municipal candidates are embedded in broader networks and use their electoral base to campaign 2 years later for state or federal deputies, who reward them with government resources (Avelino, Biderman, and Barone 2012; Novaes 2017; Pessoa Júnior 2022; Vieira 2012). Individuals and opposition groups also bypass local officials to receive benefits (Bueno 2018; Bussell 2018; Grindle 2007). One association leader said that deputies are like “nuns’ ears” because no one sees them, but the community leveraged the relationship between an elected councilor and a deputy to get a new well (Interview 124). In another community, a resident said in 2017, “The guy who arranged the well over there was [Name of most voted candidate] who ran but wasn’t elected - he got it through a deputy” (Interview 70). That candidate was elected to the city council in 2020 and received even more votes from the community in 2020 than 2016. In interviews in the same community in 2022, residents said the councilor helped fix potholes and get water storage (Interviews 106–108, 116, and 117).

Still, other residents said politicians did not help at all, which they attributed to a variety of factors. One resident said the community lacked development because of low participation and widespread vote-buying practices leading residents to disperse their votes. She believed community members should not blame everything on their politicians and should not complain if they do not mobilize (Interview 63). Residents in communities that did not bloc vote are aware of the potential benefits: one said it was challenging to get government resources because the community “does not have a councilor” (Interview 68), and councilors deny requests by citing low vote counts based on poll tapes (Interview 112).

My theory describes a long-term relationship, and it is possible that better water access among association members leads to coordinated voting, not the other way around.¹³ A community could reward an incumbent or candidate who helped with local services. This endogeneity and reverse causality are consistent with my theory and reflect the long-term cycle (Figure 1) whereby community members hold politicians accountable for public service promises, and politicians respond to voters’ preferences and seek future votes.¹⁴

Switching in Electoral Data

Lastly, I verify key theoretical assumptions that communities have consistent voting behavior and credibly threaten to switch candidates. I analyze data from 15,326 sections in 182 municipalities across Ceará during five municipal elections: 2000–16.¹⁵ Figure 7 demonstrates that vote concentration stays remarkably consistent over time.¹⁶ While it dips at the top of the distribution, a section with 75% of its votes for the most voted candidate in one election is still likely to give around 50% of its votes to the most voted candidate in the next election.

Figure 7. Voting Behavior over Time

Note: Local polynomial regression fit lines (loess) calculated using “loess” ( $ n=48,465 $ section-years where previous election is 2000, 2004, 2008, or 2012). Plot shows 95% confidence intervals. Does not include controls, municipal fixed effects, or clustered standard errors.

Bloc voting is most consistent where the previous most voted candidate ran again and was chosen again to be the section’s most voted candidate (“Ran and Same”). It is still remarkably consistent where the community switches its vote (“Ran Not Same”) or the prior choice is not available (“Did Not Run”) (see Supplementary Table A14). These findings fit with my observations from fieldwork that community characteristics are sticky.

However, this stickiness does not inherently extend to candidates. I find evidence that communities have bargaining power because they can switch, which I define as the most voted council candidate in election $ t=1 $ being different from the most voted council candidate in the previous election $ t=0 $ , even though the previous one ran again (“Ran Not Same”).

For those sections whose previous most voted candidate was in the race, 56% switch and coordinate around a different candidate.¹⁷ Sections are less likely to switch if they bloc voted in the prior election and coordinated around the person; sections are more likely to switch if they disperse their vote, since the community did not coordinate around a specific person (Supplementary Table A16).

High bloc voting sections (40–60 pp) still switched 24% of the time, and very high bloc voting sections (60–100 pp) switched 11% of the time, on average (Supplementary Table A13). In the high and very high bloc voting sections, switching is unaffected by whether the prior top candidate won (Supplementary Table A16 and Supplementary Figure A4), which is consistent with interviewees saying losing candidates helped the community. While these patterns would also be consistent with top-down coercive bloc voting, they validate key assumptions in my theory that communities can and do switch allegiance between elections. They indicate that voters have agency vis-à-vis politicians; they may even be better off not forming strong linkages with parties.

In interviews, rural residents were willing to switch if candidates did not follow through: “The shortest career is in football. It’s the same with a council candidate…if he didn’t do anything [for his supporters or potential supporters], he’s done” (Interview 75). Others explained that council candidates must provide benefits to the communities where they made promises and received votes; otherwise, they will not get votes there in future (Interviews 77, 109, and 110). Politicians must also use their position to help other communities to get additional votes (Interview 75).

CONCLUSION

How can citizens hold politicians accountable for providing public services? I argue that organized groups of citizens, especially through community associations, use bloc voting to effectively pressure politicians to provide basic public services. By combining an original household survey, a conjoint survey experiment, electoral data, and geospatial data, I measure the perceived and actual relationships between community characteristics, bloc voting, and water services.

In contrast to top-down explanations of bloc voting, residents perceive that communities with high association participation are more likely to bloc vote; participation is a larger driver of bloc voting than political endorsement by the group leader. Coordination and electoral mechanisms interact to shape a grassroots bottom-up strategy, and bloc voting is most likely where residents have stronger community trust and vote at the same polling station. Bloc voting shapes outcomes: it is associated with more secure water access among association members.

My theory and findings have several implications. First, organized marginalized communities have agency even in clientelistic settings. Most studies about vote-buying and pork politics focus on politicians’ strategies and suggest that parties or brokers drive bloc voting. I advance our understanding of collective action and distributive politics through evidence that civil society groups help marginalized residents coordinate and use their votes to influence the distribution of public services. By speaking directly with over a hundred rural residents and community leaders across widely varying communities, I unpack grassroots dynamics that research designs focused on politicians or aggregate voting behavior do not capture. Civic participation through community associations is common in this context (Boulding and Holzner 2021), and future research should evaluate the extent to which this grassroots bloc voting dynamic is exceptional or prevalent in other areas.

Second, the strategy’s grassroots nature means that variation in civic participation may disproportionately influence equity of public service access. Residents in communities with low trust and participation felt that bloc voting by other communities left them further behind. Some residents in “successful” communities that leveraged their bloc vote for better services were frustrated that they had to do so in the first place. They believed they should not have to use bloc voting to get secure drinking water, which was their fundamental right. Future research should explore the perceived fairness of bloc voting for accountability and its actual impact on inequality between and within communities.

Third, I highlight the role of electoral institutions in shaping or constraining grassroots strategies, which could partially explain why recent work has not found a link between service quality and accountability (Bland et al. 2023). Community members that demonstrate their voting behavior can leverage their coordination capacity and bargain with their bloc vote. Multiple communities that I visited even petitioned to have their own polling station to do just that! The flip side is that communities whose members are dispersed across polling stations, or who vote with other communities, have no incentive to pursue this strategy. Still, organized communities coordinate between elections by calling on councilors, protesting, and leveraging political networks.

Future research could use interviews and surveys with residents and groups to understand the conditions under which individuals and groups use bloc voting or other forms of political participation. Extensions of this work could explore how institutional arrangements, including different types of electoral systems and how citizens are assigned to polling stations, affect the ability and willingness of different groups to use bloc voting or other forms of political participation to improve accountability at different levels of government.

Lastly, my findings contribute to a growing literature on water politics. Water scarcity is a growing concern for a majority of the world’s population, and more than half of the world’s poor live in drought-prone areas (Mearns and Norton 2010). While water is highly salient for many communities, other public or club goods may be more relevant in different contexts, by which I mean rural/suburban/urban environments or different states and countries. Future research should explore how bloc voting dynamics vary by service within the same context and for water in other contexts (Kramon and Posner 2013).

This article points to the interrelationship of collective action and distributive politics for key public service outcomes. By unpacking how civil society organizations participate in local politics and how electoral institutions motivate or constrain their actions, we can better understand the political economy of development in developing democracies and design public policies to ensure that all citizens have access to essential services.