Linkage Switches in Local Elections: Evidence From the Workers’ Party in Brazil

Abstract

How do politicians from programmatic parties govern when their programmatic commitments are inconsistent with voters’ expectations? This article demonstrates that anticipated electoral pressures can lead politicians from programmatic parties to resemble those from nonprogrammatic parties. When politicians from programmatic parties believe that voters will evaluate their performance using nonprogrammatic criteria, they engage in linkage switches: They initially campaign on programmatic platforms, but prioritize nonprogrammatic actions in office. I support this claim with evidence from Brazilian municipalities (1996-2012), where the Workers’ Party (PT) employed programmatic appeals in an otherwise nonprogrammatic political context. Using a regression discontinuity design and the semi-automated text analysis of campaign platforms, I demonstrate that PT mayors initially ran on programmatic platforms, but once in office, they responded to anticipated electoral pressures by prioritizing actions that were inconsistent with the party’s programmatic orientation.

Keywords

political parties Latin American politics subnational politics representation and electoral systems decentralization

Developing countries have devolved unprecedented resources to local governments over the past three decades. This “quiet revolution” in governance means that local politicians’ actions are increasingly important for citizens’ well-being throughout the global south (Campbell, 2003). However, there is deep disagreement about the normative implications of these shifts: Observers were initially optimistic that decentralization reforms would lead to more responsive governance, but recent work suggests that they have had heterogeneous consequences—even within countries (Faguet, 2012; Grindle, 2007).

One commonly cited culprit for failures of decentralization in the developing world is the prevalence of local capture. Research suggests that decentralization reforms can empower corrupt local elites who use the selective provision of goods to gain power and escape accountability, thereby undermining the benefits of decentralization (Bardhan & Mookherjee, 2000; Fox, 2007; Platteau, 2004). In response, others have suggested that strong, programmatic parties can mitigate these problems by constraining politicians’ actions and providing meaningful policy alternatives for voters (Keefer, 2007; Manor, 1999). Through these channels, programmatic parties are expected to shift political competition away from patronage and clientelism, which are widely acknowledged to be undesirable (Hicken, 2011; Robinson & Verdier, 2013).

However, it is unclear whether this is necessarily the case. These claims about programmatic parties often implicitly assume that politicians are bound by their campaign promises, but in practice, politicians might face different incentives in and out of office. If this is the case, politicians who use different strategies to attain power might govern the same once they have power. Specifically, in contexts dominated by nonprogrammatic linkages, politicians who initially run on programmatic platforms might subsequently adopt nonprogrammatic linkage strategies to win reelection.

Here, I draw on original evidence from Brazilian municipalities to explore whether this is the case: To what extent do politicians from programmatic parties govern differently than those from nonprogrammatic parties? Brazil serves as an useful case to explore this question because it has historically had nonprogrammatic parties at the local level, with one exception: the Workers’ Party (PT). This makes it possible to test if and how the election of a mayor from a programmatic party matters in an otherwise nonprogrammatic political context. To do so, I use three complementary strategies: the qualitative analysis of internal party documents, including guides that PT leaders prepared for local candidates; a regression discontinuity (RD) design, which compares outcomes across municipalities where PT candidates won and lost close elections; and the semi-automated unsupervised text analysis of campaign platforms.

The main finding is that anticipated electoral sanctions can cause partial convergence across linkage types: PT mayors are sincere in their programmatic commitments, but they rightly believe that their reelection is contingent on the provision of selective benefits and visible public works projects. Confronted with tensions between their party’s programmatic commitments and these nonprogrammatic electoral pressures, PT mayors not only implement some programmatic policies that they believe are electorally advantageous, but also employ the same nonprogrammatic strategies as mayors from other parties. The use of nonprogrammatic linkage strategies by PT mayors constitutes linkage switching: PT mayors initially run on programmatic platforms, but prioritize nonprogrammatic linkage strategies in office, which they emphasize when running for reelection.

Through the analysis, I make three distinct contributions: First, I build on previous studies of democratic representation and policy switches by identifying a related phenomenon—linkage switches—in which politicians who initially campaign using one type of linkage strategy adopt a different strategy after gaining power. In contrast to policy switches (Stokes, 2001), which occur when politicians reverse policy positions within a single policy dimension, linkage switches involve a categorical shift in the type of linkage strategies employed by a candidate. Still, they have troubling implications for democratic representation because they involve a disjuncture between politicians’ campaign promises and their actions in office.

Second, I provide an empirical test of whether politicians from programmatic and nonprogrammatic parties govern differently. The use of a regression discontinuity (RD) design and the unique structure of Brazil’s party system makes it possible to estimate the causal effect of electing a mayor from a programmatic party in an otherwise nonprogrammatic political context, and the inclusion of a comprehensive set of outcomes—including detailed measures of public service provision within issue areas—makes it possible to assess how local politicians prioritize and target policies.

And third, I contribute to ongoing debates about the evolution of the PT in Brazil (Amaral & Power, 2016). By focusing on the complete universe of municipalities where the PT governed, rather than isolated success stories, this article provides a more complex account of the party’s historical trajectory than previous work. The findings suggest that the PT’s local administrations were not as exceptional as commonly assumed. This has implications for the theoretical lessons that can be drawn from the party’s history. Whereas previous work stresses the PT’s long-term transformation (Hunter, 2010), the present analysis suggests that the act of gaining power itself can help explain some shifts in the party’s linkage strategies.

Parties, Local Elections, and Linkage Switches

Political parties fulfill a variety of important functions in democracies. They educate citizens, recruit candidates, mediate between opposing views, simplify political debates, and aggregate preferences (Scarrow, 1967). And when parties play these roles effectively, they promote political stability and help ensure effective representation for voters (Przeworski, Stokes, & Manin, 1999; Sartori, 1976).

All of these benefits apply at the local level, where parties play crucial roles at each stage of the democratic process: In elections without incumbents, party labels are helpful heuristics for voters, who are often uninformed about local candidates. Then, in between elections, local parties check the actions of party members in office, recruit new leaders, and facilitate policy diffusion by sharing successful innovations across jurisdictions (Scartascini, Stein, & Tommasi, 2010; Van Cott, 2008). And in terms of accountability, local parties organize political competition into defined groups that can be sanctioned for their performance, which is particularly helpful in contexts with term limits (Lupu & Riedl, 2013; Manor, 1999).

However, not all parties are able to play these important representational roles. First, parties need to be strong—they need to have stable roots in society, mass support, and independent organizations (Mainwaring & Scully, 1995).¹ These characteristics allow parties to develop reputations that are independent from specific candidates, which fosters political stability (Mainwaring, 2018).

And second, for parties to effectively represent voters, they need to employ certain linkage strategies, defined here as the types of appeals that parties use to attract electoral support. While there are several elaborate typologies of linkage strategies (Kitschelt & Wilkinson, 2007; Kitschelt, 2000), I follow Stokes, Dunning, Nazareno, and Brusco (2013) and distinguish between programmatic strategies, which involve policies based on formalized and public criteria, and nonprogrammatic strategies, in which either of these criteria are absent.² I use this typology because it provides a relatively straightforward set of criteria for empirical research and corresponds to the main normative debates about linkage strategies. Still, it should be noted that programmatic and nonprogrammatic linkage strategies are ideal types, and most parties incorporate a mix of these elements (Luna, 2014).³

Previous work suggests that programmatic linkage strategies promote good governance through several channels (Cheeseman et al., 2014). During campaigns, programmatic appeals make politics intelligible to voters by providing a clear sense of the policies that a given candidate will pursue in office (Downs, 1957). And later, they provide voters a standard by which to evaluate politicians who seek reelection (Scartascini et al., 2010). Furthermore, programmatic policies are viewed as intrinsically good because they avoid the inefficiencies and distributive distortions associated with nonprogrammatic strategies, such as patronage and clientelism (Hicken, 2011; Robinson & Verdier, 2013).

However, it is unclear whether candidates who employ different linkage strategies when campaigning will necessarily govern in different ways. This ambiguity arises because the term linkage strategies is often used interchangeably to refer to politicians’ campaign appeals and their actions in office, which forecloses the possibility that politicians who campaign using different linkage strategies might converge to the same linkage strategy in office. This assumption is particularly problematic at the local level, where shifting information asymmetries between politicians and voters can provide incentives for politicians to stress different types of linkages in and out of office.

To understand why this might be the case, it is instructive to walk through different stages of the local electoral representation process: First, in open elections—those without incumbents—the key information asymmetry is that candidates know their own policy preferences, but voters have to infer candidates’ preferences from their campaign promises (Ashworth, 2012). Because citizens recognize that politicians have an incentive to pander, the credibility of candidates’ promises is conditioned by their reputations (Keefer & Vlaicu, 2008): Only politicians with personal, partisan, or familial histories of fulfilling clientelist requests or providing pork can make credible nonprogrammatic promises, and only candidates from parties with established programmatic reputations can make credible programmatic appeals. This means that candidates often have an incentive to reveal their true type in open elections.

Once a politician is in office, however, a new information asymmetry emerges: Voters do not have complete information about the full set of actions that an incumbent has pursued. This information asymmetry is particularly stark at the local level, where there are few independent media sources and nongovernmental organizations often lack the resources to effectively monitor politicians. As a result, citizens often rely on direct, firsthand observations when assessing an incumbent’s performance (Johannessen, 2017), which can push politicians to prioritize actions that are easy to observe and for which they can claim credit. These characteristics increase the likelihood that voters will reward an incumbent for a given action (Batley & Mcloughlin, 2015; Mani & Mukand, 2007).

Faced with these electoral pressures, reelection-seeking politicians from programmatic parties not only have an incentive to prioritize visible and easily attributable policies that are consistent with their programmatic orientation, but they also face incentives to adopt policies that are inconsistent with their programmatic orientation. In low-information environments, nonprogrammatic linkage strategies such as patronage, pork, and clientelism are easy for voters to observe and attribute to a specific politician, and voters are often accustomed to evaluating a politician’s performance based on nonprogrammatic criteria. Thus, even when politicians initially employ programmatic appeals, they might encounter electoral incentives to prioritize nonprogrammatic policies in office.

When politicians are more responsive to anticipated electoral pressures than their initial promises, this can lead to linkage switches: a phenomenon in which politicians prioritize different linkage strategies in and out of office. Linkage switches are different in key respects from the two most influential models of how programmatic parties govern: the responsible party government model (Committee on Political Parties, 1950) and the policy switches model (Campello, 2014; Stokes, 2001). The main differences are outlined in Figure 1.

Figure 1.

Models of programmatic party governance.

In the responsible party government model, candidates initially run on programmatic platforms, they implement their platforms in office, and then they run on the consequences of those policies (Ranney, 1954). The model proposed here differs in that it expects politicians who run on programmatic platforms will abandon those platforms once in office. This expectation follows from the assumption that reelection-seeking politicians are forward-looking and believe that voters care more about the consequences of policies than candidates’ fidelity to their initial campaign platforms.

The assumption that politicians are forward-looking is also shared by the policy switches model (Stokes, 2001). The difference is that the linkage switches model involves the prioritization of a different type of linkage strategy, rather than a policy reversal within a single programmatic dimension. It expects that politicians who campaign using programmatic appeals will try to act consistent with their programmatic commitments, but will prioritize nonprogrammatic linkage strategies in office.

Theoretically, linkage switches and policy switches are motivated by different incentives. According to Stokes (2001), policy switches occur because politicians believe that voters are misinformed about the future consequences of policies and will eventually come to embrace policies they once opposed. In contrast, linkage switches occur because politicians believe that citizens are uninformed about—or will not prioritize—programmatic actions. In both models, politicians are responsive to anticipated electoral pressures, but the nature of those pressures determines the type of switch that occurs.

Case Selection: The PT in Brazil

To test this argument, I turn to Brazil, which serves as a useful context to explore these questions because of its institutional structure and the unique position of the PT within the Brazilian party system.

In Brazil, local governments are powerful and mayors are relatively autonomous from other political actors. These institutional characteristics increase the precision with which it is possible to study the effect of mayors on policy. In terms of duties, Brazilian mayors administer programs across several social policy domains, including health care, primary education, social assistance, housing, and public transit.⁴ And within municipalities, mayors are the dominant political actors, as local legislators primarily view themselves as intermediaries for their constituents, rather than independent policy-makers (Almeida & Lopez, 2011).

Given these powers, it is unsurprising that previous work finds that short-term mayoral interventions matter. For instance, Ferraz, Finan, and Moreira (2012) show that municipal corruption reduces test scores and leads to worse student outcomes, Gonçalves (2014) demonstrates that the implementation of participatory budgeting (PB) reduces infant mortality, and de Janvry, Finan, and Sadoulet (2012) provide evidence that schools in municipalities with reelection-eligible mayors have higher school attendance than those with term-limited mayors.

Brazil is also a useful case because of its party system. Historically, observers have viewed the country’s party system as uniquely weak and nonprogrammatic—even within Latin America (Kitschelt et al., 2010; Mainwaring, 1995, p. 354). This has changed in the past two decades, as Brazil’s party system has become stronger and more programmatic (Hagopian, Gervasoni, & Moraes, 2009; Lyne, 2005; Mainwaring, 2018), but political competition continues to be predominately nonprogrammatic at the local level (Epstein, 2009).

The most notable exception is the PT. As Keck (1992) noted in the midst of the party’s rise, the PT was initially an anomaly in Brazil because it “had a solid base in labor and social movements, took seriously the question of representation (both in internal organization and with regard to electoral constituencies), and couched its appeals in programmatic terms” (p. 3). In contrast, other Brazilian parties at the time were “highly permeable, elitist, and personalistic” (Keck, 1992, p. 3). These early insights largely still apply. Despite its recent “normalization” (Hunter, 2010), the PT continues to be the most ideologically cohesive major party in Brazil (Kitschelt et al., 2010).⁵

The contrast between the PT and other Brazilian parties is clearest at the local level, where the PT sought to build its reputation. In 1996, the party developed a platform for mayoral candidates entitled, “the PT way of governing” (Partido dos Trabalhadores: Secretaria de Assuntos Institucionais, 1996), which included commitments to citizen participation, pro-poor policies, and clean governance. Together, these commitments constituted a programmatic approach to local elections that was viewed as a radical departure from the linkage strategies employed by other parties (Hochstetler, 2008).

Certain aspects of the PT’s distinct nature at the local level find support in the data. For instance, PT mayoral candidates tend to be uniquely rooted within their party. In the 2000 to 2012 elections, only 13% of PT candidates in the 2000 to 2012 elections were affiliated with a different party in the previous election, and only 8% left the PT to run for another party in a subsequent election (see Table 1). These rates are less than half of other parties’ rates. The most likely cause for this difference is that PT candidates were required to endorse the party’s platform, which inhibited the opportunistic affiliations common with other parties (Novaes, 2018).⁶

Table 1.

Party Switching—Brazilian Mayoral Candidates (2000-2012).

Party name	Total candidates	% switched out	% switched in
PMDB	9,921	18.7	22.6
PSDB	7,143	26.7	27.6
PT	6,521	8.0	13.1
PFL/DEM	3,956	36.9	22.6
PTB	3,852	39.7	35.5

Source. Tribunal Superior Eleitoral (author’s calculation).

Note. PT = Workers’ Party.

A similar view of the PT’s position within Brazil’s party system comes from the Democratic Accountability and Linkages Project (DALP), an expert survey that includes questions about the internal organization, ideology, and linkage strategies of parties throughout the world.⁷ The survey was conducted in countries with more than two million inhabitants and at least two national elections in the recent past and covers all parties that received over 5% of the vote in the previous legislative elections. With the data, I created indexes that capture expert perceptions of (a) the extent to which a party uses programmatic campaign appeals and (b) the strength of a party’s local organization.⁸ The results (Figure 2) reflect the consensus in the literature: At the local level, the PT is the strongest and most programmatic party in Brazil.⁹

Figure 2.

Programmatic party linkages and local party strength according to DALP surveys (2008-2009).

The PT and Anticipated Electoral Pressures

While these indicators speak to the PT’s unique organization and use of programmatic campaign appeals, it is unclear whether these characteristics translated into meaningful policy differences across PT and non-PT administrations at the local level.

This is a concern because PT mayors often took office in cities where citizens’ expectations about local politics were in tension with the PT’s programmatic orientation. For instance, several studies exploring the demand side of clientelism in Brazil illustrate that citizens pay lip service to programmatic ideals (Borges, 2018), but expect and initiate clientelist exchanges (Nichter & Peress, 2017). And ethnographic work suggests that politicians who attempt to resist these pressures quickly face a backlash (Borges, 2019). Similarly, other studies have highlighted the key role of patronage and pork for incumbent performance (Colonnelli, Teso, & Prem, 2018; Johannessen, 2017).

Given these pressures, it is unsurprising that PT officials have perceived tensions between the party’s programmatic commitments and electoral ambitions. This is evident in internal party documents, including party newsletters and candidate guides.¹⁰ For instance, the authors of a guide for PT candidates in local elections stated, “The left has a tendency to deify programmatic positions. However, as important as they might be, they don’t win elections by themselves” (Pereira & Paschoal, 1996, p. 21). They continued, “It’s important to not forget that the majority of people, contrary to our wishes, do not think like us, and because of that, actions and concepts that we might judge to be fundamental might not be for them” (Pereira & Paschoal, 1996, p. 25).¹¹

These perceived tensions between voters’ expectations and the party’s programmatic commitments had implications for the policies that PT mayors felt they had to adopt in office. For instance, one guide noted, “It is necessary to balance our strategic goals and the Brazilian reality” (Partido dos Trabalhadores: Secretaria de Assuntos Institucionais 1996, p. 3). To do so, the guide noted that PT governments would likely have to diverge from some of the party’s main ideological commitments to win reelection, suggesting, “There is a permanent political tension between an elected government that needs to respond to the everyday needs of getting stuff done and the party’s expectation of seeing its programmatic stances put into practice” (Partido dos Trabalhadores: Secretaria de Assuntos Institucionais, 1996, p. 11).

These tensions were particularly acute for the PT because its procedural commitment to popular participation was potentially at odds with its programmatic commitment to directing resources toward public services for the poor. Policies that party members believed were most beneficial for citizens, such as increased spending on health and education services, were not always prioritized by voters. This is evident in a guide for PT candidates in small towns, which noted that necessity should not always determine an administration’s policy choices. The authors write, “People don’t just want food. They have dreams and fantasies. They like pleasure, and being recognized and admired” (da Silva, 1996, p. 10). For this reason, the guide suggested that politicians might want to prioritize, “a pretty bus station because that would give a good impression to those who arrived in the city” (da Silva, 1996, p. 11). And while this excerpt could be viewed as a straightforward endorsement of participation, it is nevertheless in tension with the PT’s stated commitment to improving the material well-being of the poor by allocating resources based on need. And the same guide noted that politicians should be responsive to citizens’ demands, even when that meant goods were no longer allocated using formal and public criteria (da Silva, 1996, p. 11).

Some internal critics pushed back against advice like this. For instance, a guide for city council candidates listed a series of risks facing party members, including “giving into the popular opinion that all politicians have to build public works,” “wanting to present results at any cost,” and “making reelection the primary goal” (Pietá, 1996, p. 13). And some of the PT’s internal factions criticized party members for failing to implement the party’s program. In one case, a newsletter from the party’s democratic socialist faction noted that mayors in Rio Grande do Sul—home to some of the party’s most celebrated administrations—were affected by “the ideological degradation that affects the left [and] relegates principles to second place in the name of a pragmatism oriented around a politics of results in which the existing rules of the game dominate” (Marques, 2000, p. 6). Two other party officials echoed this concern, suggesting that the actions of affiliated mayors were “not sufficiently different from those of local traditional bourgeois parties” because mayors did not “put our program into place” (Helenda & Bastos, 2000, p. 3).

Yet, it is unclear whether these critiques were justified. The dominant narrative in the literature is that PT mayors resisted nonprogrammatic pressures and governed in a programmatic manner (Hochstetler, 2008; Hunter, 2007), but these assessments tend to be based on highly visible cases, such as the PT’s famous administrations in Porto Alegre (Abers, 2000). Studies that look at a wider range of cases reach mixed conclusions (Alves, 2018; Baiocchi, 2003). Here, I build on existing work by using a rigorous identification strategy and data from the complete universe of PT administrations to understand what the “PT way of governing” looked like in practice.

Research Design

To estimate the causal effect of electing a PT mayor, I use a regression discontinuity (RD) design that compares policies and outcomes across cities in which PT mayoral candidates win and lose close elections.¹² This approach addresses a clear threat to causal inference: PT candidates might have contested (and won) elections in cities that differed systematically from those in which they either chose not to run or ran and lost. This is plausible because the party’s initial successes came in the South of Brazil, a region with a history of relatively clean and competent governance. Thus, the PT might have earned a reputation for good governance by administering municipalities that were already well-run. The RD design addresses this concern by controlling for contextual variables (Eggers, Fowler, Hainmueller, Hall, & Snyder, 2015; de la Cuesta & Imai 2016).

The primary assumption needed to interpret the difference in expected outcomes at the discontinuity as a causal effect is continuity: the only change at the discontinuity should be the shift in treatment status.¹³ In other words, there should be no discontinuous jumps in the value of other contemporaneous variables at the exact point at which PT mayors win or lose a close election (in this case, as the margin of victory approaches 0 from above and below the discontinuity). I test the continuity assumption in two ways: First, I employ a McCrary Sorting Test (McCrary, 2008), which evaluates whether the distribution of the running variable—in this case, the PT’s margin of victory—is smooth at the discontinuity. And second, as a placebo test, I estimate RD models using lagged dependent variables as the outcomes. As these values are causally prior, they should be continuous at the discontinuity. Both tests suggest the research design is valid.

To model the relationship between vote share and outcomes at the cutoff, I use a robust nonparametric local linear regression estimation strategy (Calonico, Cattaneo, & Titiunik, 2014).¹⁴ This approach has become standard in the literature because it has better theoretical properties at the discontinuity than other approaches, such as polynomial models, and it includes a data-driven optimal bandwidth selection procedure, which minimizes concerns about researcher degrees of freedom in choosing the observations to be included in the analysis (de la Cuesta & Imai, 2016, p. 12).

The main drawback of the RD design is that it recovers a local average treatment effect (LATE) at the discontinuity, rather than a population average treatment effect (PATE) for all Brazilian municipalities. The design makes it possible to precisely estimate the causal effect of electing a PT mayor in elections decided by an infinitesimally small margin, but not in other cases. This raises concerns about external validity because electoral competitiveness might have an independent effect on how politicians act in office (Boulding & Brown, 2013; Cleary, 2007). However, I reestimated the main findings using multivariate regressions with data from all Brazilian municipalities and find similar results, which provides some evidence this is not the case. These estimates require much stronger assumptions to be interpreted as causal effects and are not included here.

Analysis

If PT mayors engaged in linkage switches in response to anticipated electoral sanctions, we should expect to see the following patterns in the data: First, PT mayors will seek to follow through on their party’s programmatic commitments, but will prioritize policies that are electorally advantageous. Second, PT mayors will adopt other linkage strategies that they perceive as electorally advantageous, even when those are inconsistent with the party’s programmatic orientation. And third, differences in policies and outcomes across municipalities governed by PT and non-PT mayors should remain constant over time, even as the party evolved over the late 1990s and early 2000s from a radical left-wing party with strong union roots to a center-left, catch-all party (Amaral & Power, 2016; Hunter, 2010).

I test these hypotheses in two steps: First, I look at two policy domains that are central to the PT’s program—participation and pro-poor policies—to see whether PT mayors prioritized electorally advantageous policies within these areas. Then, I look for direct evidence of linkage switches by testing whether PT mayors were as likely as non-PT mayors to employ pork and patronage. The results across both sets of analyses provide evidence of convergence across PT and non-PT administrations.

Participation

To start, I assess whether PT mayors are more likely than mayors from other parties to create participatory governance institutions (PGIs). Specifically, I estimate whether the election of a PT mayor leads to the adoption of two types of PGIs: participatory budgeting (PB) and municipal councils (MCs).¹⁵

Institutional differences make this a productive comparison. PB is a highly visible program in which citizens participate in open forums to set government spending priorities (typically for infrastructure projects) and select citizen representatives to oversee the implementation of these projects. The program was an integral part of the PT’s brand and was featured prominently in its promotional materials. Existing work finds that PT mayors (and only PT mayors) receive an electoral advantage from adopting PB (Spada, 2010).¹⁶

Conversely, MCs are issue-specific committees that meet regularly to monitor government performance, discuss priorities, and exchange information. While their composition varies by issue area and municipality, MCs generally provide fewer electoral benefits for PT incumbents because they do not involve mass participation or create highly visible outputs. The PT has more difficulty claiming credit for MCs because their roots date back to corporatist institutions created during Vargas’s Estado Novo (Tranjan, 2016).¹⁷ Still, the PT-endorsed councils as a key form of participation and they have been linked with a host of positive outcomes (Donaghy, 2013; Touchton, Sugiyama, & Wampler, 2017).¹⁸

These institutional differences suggest that if PT mayors are primarily motivated by electoral considerations, they will prioritize PB, rather than MCs. Conversely, if PT mayors are motivated by an intrinsic desire to promote participation, we should expect them to adopt both types of institutions at a higher rate than non-PT mayors.¹⁹

The results (Figure 3) are consistent with the expectation that PT mayors will prioritize PB, but not MCs. The election of a PT mayor is associated with the adoption of both PB and MCs, but the effect is only substantively and statistically significant for PB. Whereas the election of a PT mayor practically guarantees the adoption of PB,²⁰ it only leads to the adoption of 0.2 more councils. This estimate is substantively small because there are more opportunities to adopt councils than PB (the average municipality had fewer than five of the 12 possible councils during the period considered here), and it is not statistically significant.²¹ Thus, PT mayors follow through on the party’s commitment to participation, but prioritize the institution that provides a clear electoral return. Notably, there are no clear temporal trends, which is consistent with the expectation that PT mayors’ actions are shaped by anticipated electoral pressures, rather than the strength of the party’s programmatic orientation.

Figure 3.

Effect of Workers’ Party mayors on council and participatory budgeting adoption (regression discontinuity estimates, 1996-2012, 95% confidence intervals).

Pro-Poor Policy

Given the close link between the PT and PB, these patterns might not be surprising, but similar patterns exist for pro-poor policies, the PT’s second programmatic commitment. This can be seen across the domains of housing, health policy, and education.

To start, I focus on the impact of PT mayors on housing policy, an area in which mayors can initiate new programs, hire employees, and allocate funds. I estimate the effect of electing a PT mayor on the adoption of four types of housing programs: the construction of units, land grants, the provision of building materials, and the regularization of titles. I focus on these programs because previous work demonstrates that they are both substantively important and manipulable by municipal governments (Donaghy, 2013, p. 51).²²

If PT mayors are primarily motivated by anticipated electoral pressures, I expect that they will prioritize construction over less visible housing initiatives. Land titles and the provision of building materials have the potential to be electorally beneficial, but the construction of public housing provides more electoral benefits due to its visibility—a fact acknowledged by local officials (e.g., Rolnik, Iacovini, & Klintowitz, 2014). And this is borne out in the data: Of the municipalities considered here, 64% have construction programs, but only 27% provide land, 39% provide materials, and 22% have programs to regularize titles.

Consistent with expectations, the RD results (Figure 4) show that PT mayors are more likely to construct housing than mayors from other parties, but have no discernible effect on the other three programs. The magnitude of this difference is striking: the election of a PT mayor leads to a 38 percentage point increase in the probability of implementing a public housing program. Temporal trends are not statistically significant,²³ which is consistent with the claim that PT mayors seek to follow through on their programmatic commitments, but remain responsive to electoral pressures when doing so.

Figure 4.

Effect of Workers’ Party mayors on housing programs (regression discontinuity estimates, 2000-2012, 95% confidence intervals).

Results in health care and education are also consistent with expectations. To understand the effect of PT mayors on health policy, I look at service provision, access, and outcomes. For services, I consider the proportion of a municipality’s population that is covered by community health agents and family health teams.²⁴ For access, I consider the proportion of births in which the mother had at least seven prenatal visits, which is a common indicator for access to care (Fujiwara, 2015). And for outcomes, I use infant mortality rates, a common measure that mayors can affect through short-term interventions (Gonçalves, 2014).²⁵ Because the PT’s commitment to pro-poor policies means that disadvantaged citizens should be the main beneficiaries of its administrations, I also test for the effect of PT mayors on inequalities in health access and outcomes.²⁶

In terms of service provision, I expect that PT mayors will have a larger effect on the expansion of health agents than health teams because agents go door-to-door in their communities. As for access and outcomes, infant mortality and prenatal visits reflect crucial dimensions of health system performance, but I expect that they will not be prioritized by PT administrations because they are difficult for citizens to monitor and attribute to mayoral initiatives.

The results (Figure 5) are consistent with expectations. There is some evidence that PT mayors expanded access to Programa Saúde da Família (PSF) programs, but time trends are in the opposite direction as would be expected if explained by the programmatic nature of the party. Instead, PT mayors only became more likely to hire community health agents and family health teams after the PT won the presidency in 2002, at which point the national party could claim credit for the initiative. And in terms of access and outcomes, there are no consistent differences across municipalities in which PT and non-PT mayors won close elections: Infant mortality and prenatal appointment rates are similar in terms of both overall levels and inequalities. Time trends are inconclusive.

Figure 5.

Effect of Workers’ Party mayors on health programs (regression discontinuity estimates, 2000-2012, 95% confidence intervals).

Similar patterns emerge for primary education, which is also administered by Brazilian municipalities. To measure school quality, I focus on class size, which has been identified as a key determinant of student achievement in other contexts (Case & Deaton, 1999). And to capture the effect of mayoral initiatives that are not measured across municipalities, I look at pass rates.²⁷ As with health care, I assess the distributional effect of electing a PT mayor by calculating a GINI index for class size and pass rate inequalities in a given municipality.²⁸ For all indicators, I expect that there will be no differences across PT- and non-PT-governed municipalities because these outcomes are relatively hard to observe for most citizens—especially those without children in the school system.

In line with expectations, the results (Figure 6) do not show any consistent differences in education policy across PT and non-PT administrations. Municipalities governed by PT and non-PT administrations have similar student–teacher ratios, pass rates, and levels of educational inequality. The time trends for student–teacher ratios and pass rates also go in opposite directions: schools in cities where PT mayors were elected in 1996 and 2000 not only had slightly lower student–teacher ratios, but also had lower pass rates. None of these estimates are statistically significant.

Figure 6.

Effect of Workers’ Party mayors on education programs (regression discontinuity estimates, 2000-2012, 95% confidence intervals).

Pork

The preceding results suggest that PT mayors prioritized some programmatic policies that were electorally advantageous, but otherwise resembled non-PT mayors. However, to directly test whether PT mayors engaged in linkage switches, it is necessary to turn to actions that are clearly inconsistent with the party’s programmatic orientation.

One example of this is pork. When the PT was out of power at the national level, it was a fierce critic of pork, which is defined here as the allocation of funds for localized projects based on electoral criteria. For instance, in newsletters published during Brazil’s 1996 mayoral elections, PT activists and officials expressed outrage at the Cardoso administration’s use of federal funds to help allies at lower levels of government. In one editorial, a party member wrote, “The most problematic attitude taken by [the national] government to support allied candidates is the delivery of public resources to state and local governments administered by the party’s friends. These funds are not being distributed according to necessity” (Montenegro, 1996, pp. 2-3). And while these concerns read as sour grapes from a party out of power, they also reflect the PT’s programmatic opposition to spending money on public works projects at the expense of social services. The PT’s initial national platform explicitly claimed to oppose “the conservative political culture of investing in highly visible public goods, abandoning other priorities, draining resources for public services” (Partido dos Trabalhadores: Secretaria de Assuntos Institucionais, 1996, p. 8).

Yet, after the PT gained the presidency in 2002, PT mayors received disproportionate transfers from the federal government that were primarily used for visible public works projects and capital purchases. As shown in Figure 7, the election of a PT mayor led to a more than R$ 30 increase (per capita) in federal funds after the PT gained national power in 2002. Yet, there was no corresponding increase in transfers from state governments, which suggests that federal transfers are not a proxy for mayoral competence.²⁹

Figure 7.

The effect of Workers’ Party mayors on voluntary intergovernmental transfers (regression discontinuity estimates, 1996-2012, 95% confidence intervals).

Still, there are reasons to believe that these transfers did not constitute pork. It is possible that federal transfers were directed toward co-partisan mayors because they were more willing to advance the national government’s programmatic agenda than opposition mayors. Furthermore, even if transfers were allocated based on electoral considerations at the national level, the funds might have been directed toward programmatic policies at the local level.

Yet, neither alternate explanation seems consistent with the evidence. As shown in Figure 8, PT mayors did not use their budget surplus to shift money toward social spending. Instead, they only allocated additional resources to capital purchases, free goods, and public works projects.³⁰ And party officials openly acknowledged the key role that these transfers played in the party’s subsequent local electoral success. As one postelection report by a PT member slyly noted, “[after Lula’s election] the central government began to treat municipal governments equally, which increased the competitiveness of PT administrations” (Ferreira, 2004, p. 1).

Figure 8.

Effect of Workers’ Party mayors on spending (regression discontinuity estimates, 1996-2012, 95% confidence intervals).

Thus, the PT’s politicized use of federal transfers represented a departure from the party’s programmatic orientation at both the national and local levels. At the national level, the PT directed federal funds toward PT mayors—a scheme in which mayors were active participants. And at the local level, mayors spent the money on visible projects, capital purchases, and selective benefits—not social spending.

Patronage

Similar patterns hold for patronage, which is defined here as the allocation of public employment based on electoral considerations. Patronage is difficult to identify empirically because it can take several different forms, but PT and non-PT administrations had similar hiring patterns across several different indicators.

The most straightforward form of patronage involves allocating government jobs as a reward for partisan loyalists. I do not test for this here, but recent work suggests that PT mayors are—if anything—more likely to provide government jobs to declared partisans than mayors from other parties (Brollo, Forquesato, & Gozzi, 2018). However, this does not necessarily indicate patronage because PT mayors might have a programmatic interest in hiring workers who share their ideological commitments.

For this reason, I focus on a second variant of patronage: targeting jobs toward swing voters. Without detailed information about the political behavior of all public employees, it is difficult to classify any individual hire as patronage, but this form of patronage has several observable implications for aggregate hiring patterns. To start, patronage might occur through sheer volume, in which politicians give jobs to as many people as possible. This would lead to high levels of hiring immediately after an election. Alternately, politicians who rely on patronage to secure electoral support might disproportionately use short-term contracts to retain power over public employees. Or finally, patronage might be targeted toward low-skill workers, who are cheaper to buy off and will have less of a detrimental effect on an administration’s bureaucratic capacity (Brierley, 2018).

The presence of any single one of these patterns would not definitively indicate the presence of patronage because each pattern could be explained by alternate causes, but there are few alternate channels through which patronage could plausible occur. For this reason, I view this a hoop test: Differences in hiring patterns across PT and non-PT administrations would not necessarily indicate differences in the use of patronage, but the absence of differences would provide compelling evidence that PT and non-PT administrations do not differ in their use of patronage.

To test for differences in hiring patterns, I draw on data from Brazil’s Ministry of Work and Employment (MTE), which requires that all employers—including municipal governments—provide information about employees, including occupational codes, educational attainment, and hire dates. With this individual-level data, I created three distinct municipal-level indicators of patronage: (a) rates of new hires, (b) the difference in turnover for low-skill workers (compared with all workers), and (c) the prevalence of temporary contracts for low-skill hires. These indicators correspond to the different aggregate patterns of patronage outlined above.

The results (Figure 9) demonstrate that PT administrations have broadly similar hiring practices as non-PT administrations. Across all employment indicators, PT and non-PT administrations are statistically and substantively indistinguishable. As with the prior analyses, there are no clear temporal patterns.

Figure 9.

Effect of Workers’ Party mayors on hiring patterns (regression discontinuity estimates, 2000-2012, 95% confidence intervals).

Robustness Checks

These results are robust to a range of model specifications, measures, and coding choices. Robustness tests include

Excluding the PSDB, which could be classified as a strong and programmatic party.

Only comparing the PT with clientelist parties (defined as those on the left of Figure 2).

The exclusion of outliers that might skew the results.

Robustness checks are included in the supplemental information.

Linkage Switches in Candidate Platforms

The preceding patterns suggest that PT mayors were as likely to use nonprogrammatic strategies as non-PT mayors from 1996 to 2012, but it is possible that PT candidates emphasized the same strategies as other candidates before entering office. The archival documents presented earlier suggest that this was not the case, but it remains possible that the patterns uncovered in the RD analysis reflect preelection policy convergence, not linkage switches.

As an additional test, I employ the semi-automated text analysis of campaign platforms. Since 2009, all Brazilian mayoral candidates have been required to file a platform at least 3 months before the first round of municipal elections. These platforms serve as an official statement of candidates’ promises and are often publicized in media outlets.³¹ Here, I draw on campaign platforms from Brazil’s 2012 mayoral elections to test (a) whether PT candidates initially use more programmatic appeals than other candidates and (b) whether PT candidates become less likely to use programmatic appeals when running for reelection.

To create the data set, I scraped platforms for all candidates in Brazil’s 2012 mayoral elections from the Supreme Electoral Court’s (TSE) website, then I ran an OCR script to convert all files into .txt documents. I cleaned the documents using a series of standard techniques: I made characters lowercase; stemmed all words; removed stopwords, numbers, and punctuation; and removed idiosyncratic terms that appeared in fewer than 500 documents. Finally, I linked the cleaned documents with a data set that includes candidate characteristics (party and incumbency status) and municipal characteristics (region, economic inequality, human development, and population).³²

With the cleaned data, I estimated a Structural Topic Model (STM; Lucas et al., 2015). The STM is a semi-automated, unsupervised approach to text analysis in which a researcher selects the number of topics and then estimates the content and prevalence of those topics as a function of user-selected covariates.³³ There are three characteristics of the model worth noting: First, the STM is an unsupervised approach, which means that topics emerge from the data, rather than a predefined coding scheme. This addresses concerns about researcher degrees of freedom. Second, the STM is a mixed-membership model because it allows for each document to contain multiple topics. This is different than single-membership models, which assign a single topic to each document (Grimmer & Stewart, 2013). The mixed-membership approach is a better fit for long, complex documents like party platforms, which contain multiple topics. And third, the STM is preferable to other mixed-membership approaches because it incorporates information from covariates during the estimation and interpretation of the model.

In estimating the STM, I am more interested in the language that candidates use to justify policy proposals than the presence of any specific policy. As defined earlier, programmatic linkage strategies involve the use of formal and transparent criteria, so I expect that the PT will be unique in how it frames its policy proposals. That said, candidate platforms provide a hard test for the theory because they represent a sanitized version of a candidate’s appeals: Candidates cannot openly advertise their use of illegal linkage strategies (e.g., clientelism and patronage), so all candidates have an incentive to propose programmatic policies in the platforms, even when they do not actively campaign on those policies.

Still, meaningful differences emerge across parties. Of the topics identified by the STM, two topics roughly map onto the PT’s main programmatic commitments (Inclusion and Participation), and one maps onto the construction of Public Works.³⁴ A list of associated keywords for each topic can be found in Table 2. Robustness checks and additional information about topics and keywords, including full model results and examples of representative platforms, can be found in the supplemental information.

Table 2.

Selected Keywords by Topic.

Topic	Keywords
Inclusion	proposal, services, social, disability
Participation	society, participation, public, governance
Public works	acquisition, construction, enlarge, maintenance

Note. Topics are hand-labeled based on keywords.

To test whether PT candidates are more likely to stress inclusion and participation when they initially run for office than when they run for reelection, I modeled the prevalence of each topic as a function of the interaction between a candidate’s incumbency status and party membership, controlling for population, levels of development, economic inequality, and region. With the STM results, I estimated the predicted prevalence of each topic by incumbency status and party membership, holding control variables at their medians.

The results (Figure 10) are consistent with theoretical expectations. In open elections, PT candidates are more likely than candidates from other parties to emphasize inclusion and participation, but the rhetoric used by PT candidates shifts when they run for reelection, at which point they emphasize public works.³⁵ And while incumbents from other parties also emphasize public works more than nonincumbents, the magnitude of the difference for PT candidates means that the appeals of PT incumbents resemble the appeals of other parties’ incumbents more than for challengers. Notably, the other major party in Brazil that could potentially be classified as programmatic (the PSDB) sees a similar shift toward emphasizing visible public works projects when running for reelection.

Figure 10.

Estimated prevalence of topics in mayoral candidate platforms by party and incumbency status in 2012 (Structural Topic Model).

Thus, the text analysis provides additional evidence of linkage switches: Nonincumbent PT candidates use different appeals than nonincumbents from other parties, but this shifts when PT mayors run for reelection, at which point they abandon their unique appeals and—along with other incumbents—stress public works. The abstract nature of the topics extracted from the platforms means that the results should be interpreted with caution, but the emphasis on pork among PT incumbents is consistent with the findings from both the RD analysis and archival documents. Viewed together, all forms of evidence suggest that linkage switches occur and are driven by anticipated electoral pressures.

Conclusion

To review, this study explored how local politicians from programmatic parties campaign and govern in nonprogrammatic political contexts. To do so, it drew on evidence from Brazil, where the PT has historically stood out as a relatively programmatic party in an otherwise nonprogrammatic party system. The results show that PT mayors adopted policies in response to anticipated electoral pressures. Some of these policies were consistent with the party’s programmatic orientation, but others—such as the use of patronage and pork—were not. I provide evidence that these actions constituted linkage switches: PT candidates initially campaigned on programmatic platforms, but prioritized nonprogrammatic actions in office, which they subsequently emphasized when seeking reelection.

Despite large literature on citizen–politician linkages and policy switches, this is the first study to identify how shifting incentives can push politicians to prioritize different types of linkage strategies in and out of office. In this way, the concept of linkage switches extends previous work on citizen–voter linkages by acknowledging the temporal instability of those linkages. And while the difficulty inherent in distinguishing between types of linkage strategies means that these shifts do not necessarily involve a wholesale abandonment of a programmatic linkage strategy on the part of PT mayors,³⁶ I find evidence of shifts in PT candidates’ appeals when they run for reelection that are consistent with their actions in office.

These insights suggest two directions for future research: To start, scholars could extend the framework presented here and map out other types of linkage switches. This study focused on switches from programmatic strategies toward pork and patronage, but different incentives might encourage different types of linkage switches. For instance, politicians might switch from programmatic to clientelist strategies in the face of citizen demands (Nichter & Peress, 2017).³⁷ Alternately, a party that gains power using clientelist appeals might pivot toward programmatic policies to build a broader coalition, especially when faced with an electorate that disapproves of clientelism (Weitz-Shapiro, 2012).

Future research should also investigate the conditions under which linkage switches occur, why they take the forms they do, and their implications for democracy: Are linkage switches more common at the local or national level? Do term limits inhibit linkage switches by pushing parties to adopt longer time horizons? Under which conditions do parties choose to ignore future electoral pressures? Does the internal organization of parties matter for linkage switches? Do reverse linkage switches occur when incumbents are term-limited? How do voters respond to these switches? And what are the long-term implications of linkage switches for parties and the broader political system? These questions suggest a fruitful research agenda that could shed light on an understudied phenomenon with important implications for understanding democratic representation.

Table 3.

Pooled regression discontinuity Results for Effect of Electing a PT Mayor on All Dependent Variables.

	Coefficient	Low CI	High CI	SE	Z	p value	N	BW
Councils	0.21	−0.22	0.65	0.22	0.95	.34	1,561	0.32
Participatory budgeting	0.84	0.60	1.09	0.13	6.69	.001*	315	0.28
Construct housing	0.40	0.23	0.58	0.09	4.54	.001*	1,297	0.22
Provide land	0.06	−0.07	0.19	0.07	0.92	.36	1,297	0.42
Provide materials	0.05	−0.12	0.22	0.09	0.57	.57	1,297	0.29
Regularize titles	0.09	−0.06	0.23	0.07	1.16	.24	1,297	0.29
Community health agent	0.08	0.003	0.15	0.04	2.04	.04	1,561	0.28
Health teams coverage	0.02	−0.06	0.11	0.04	0.49	.62	1,561	0.25
Infant mortality	−0.22	−2.83	2.39	1.33	−0.17	.87	1,227	0.28
Infant mortality inequality	−0.07	−2.82	2.68	1.40	−0.05	.96	885	0.36
7+ prenatal visits	0.004	−0.04	0.05	0.02	0.15	.88	1,237	0.33
7+ prenatal visits inequality	0.004	−0.02	0.03	0.01	0.28	.78	887	0.33
Student–teacher ratio	−0.31	−1.54	0.93	0.63	−0.49	.63	1,321	0.38
Student–teacher GINI	0.001	−0.01	0.01	0.01	0.15	.88	834	0.26
Pass rate	0.001	−0.02	0.03	0.01	0.11	.91	1,368	0.34
Pass rate GINI	−0.002	−0.01	0.01	0.01	−0.33	.74	834	0.30
Federal rransfers (R$)	31.22	11.68	50.77	9.97	3.13	.002	1,294	0.26
State transfers (R$)	1.76	−11.52	15.05	6.78	0.26	.79	1,294	0.27
Free goods spend	0.002	−0.001	0.005	0.002	1.05	.30	1,294	0.31
Public works spend	0.02	−0.002	0.03	0.01	1.75	.08	1,290	0.22
Capital purchase spend	0.01	0.01	0.02	0.003	4.78	.001*	1,290	0.21
Social spend	−0.01	−0.04	0.01	0.01	−1.23	.22	1,549	0.33
New hires	0.02	−0.01	0.06	0.02	1.20	.23	1,295	0.45
Rate of low-skill hires	0.04	−0.07	0.15	0.06	0.77	.44	1,052	0.43
Use of temp. contracts	0.03	−0.08	0.13	0.05	0.50	.62	1,281	0.33

Note. Bias-corrected estimator with robust 95% confidence intervals (Calonico, Cattaneo, & Titiunik, 2014). PT = Workers’ Party; CI = confidence interval; BW = bandwidth.

p < .001.

Supplemental Material

Linkage_Switches_in_Local_Elections_-_Supplemental_Information – Supplemental material for Linkage Switches in Local Elections: Evidence From the Workers’ Party in Brazil

Supplemental material, Linkage_Switches_in_Local_Elections_-_Supplemental_Information for Linkage Switches in Local Elections: Evidence From the Workers’ Party in Brazil by Peter G. Johannessen in Comparative Political Studies

Supplemental Material

replication-code – Supplemental material for Linkage Switches in Local Elections: Evidence From the Workers’ Party in Brazil

Supplemental material, replication-code for Linkage Switches in Local Elections: Evidence From the Workers’ Party in Brazil by Peter G. Johannessen in Comparative Political Studies

Footnotes

Author’s Note

This paper benefitted from valuable feedback on earlier drafts by Francisco Cantu, Marty Gilens, Alisha Holland, Pablo Querubín, Glauco Peres da Silva, Grigo Pop-Eleches, Jason Seawright, Deborah Yashar, and Manuel Vogt.

Declaration of Conflicting Interests

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

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by funding from the Kellogg Institute for International Studies at the University of Notre Dame; the Ash Center for Democratic Governance and Innovation at the Harvard Kennedy School; and the Mamdouha S. Bobst Center for Peace and Justice at Princeton University.

ORCID iD

Peter G. Johannessen

Supplemental Material

Supplemental material for this article is available online at the CPS website

Notes

Author Biography

Peter G. Johannessen is a Democracy Postdoctoral Fellow with the Ash Center for Democratic Governance and Innovation at Harvard University’s Kennedy School of Government. His main areas of interest are democratic representation, local politics, and decentralization.

References

Abers

R. N.

(2000). Inventing local democracy: Grassroots politics in Brazil. Boulder, CO: Lynne Rienner.

Almeida

Lopez

F. G.

(2011). Representação Política Local: Padrões de Atuação dos Vereadores em Quatro Cidades Mineiras [Local Political Representation: Action Patterns of City Councilors in Four Cities in Minas Gerais]. Rio de Janeiro: Instituto de Pesquisa Econômica Aplicada.

Alves

J. A.

(2018). Transformation or substitution? The Workers’ Party and the right in Northeast Brazil. Journal of Politics in Latin America, 10, 99-132.

Amaral

O. E.

Power

T. J.

(2016). The PT at 35: Revisiting scholarly interpretations of the Brazilian Workers’ Party. Journal of Latin American Studies, 48, 147-171.

Ashworth

(2012). Electoral accountability: Recent theoretical and empirical work. Annual Review of Political Science, 15, 183-201.

Baiocchi

(2003). Radicals in power: The Workers’ Party and experiments in urban democracy in Brazil. London, England: Zed Books.

Bardhan

Mookherjee

(2000). Capture and governance at local and national levels. The American Economic Review, 90, 135-139.

Batley

Mcloughlin

(2015). The politics of public services: A service characteristics approach. World Development, 74, 275-285.

Borges

(2018a). Vote buying and impression management: Performing Largesse in Northeast Brazil (Working Paper). Retrieved from https://www.marianaborges.info/uploads/7/5/6/5/75650241/borges_apsa_2018.pdf

10.

Borges

(2019). When voters help politicians: Understanding elections, vote buying, and voting behavior through the voters’ point of view (Working paper). Retrieved from https://www.marianaborges.info/uploads/7/5/6/5/75650241/when_voters_help_politicians_180808.pdf

11.

Boulding

Brown

D. S.

(2013). Political competition and local social spending: Evidence from Brazil. Studies in Comparative International Development, 49, 197-216.

12.

Brierley

(2018). Patronage, meritocracy and political party machines (Working paper). Retrieved from https://www.sarahbrierley.com/uploads/2/7/7/1/27711295/brierley_merit_machine_9jun18.pdf

13.

Brollo

Forquesato

Gozzi

J. C.

(2018). To the victor belongs the spoils? Party membership and public sector employment in Brazil (Working paper). Retrieved from https://warwick.ac.uk/fac/soc/economics/staff/fbrollo/tothevictor.pdf

14.

Calonico

Cattaneo

M. D.

Titiunik

(2014). Robust nonparametric confidence intervals for regression-discontinuity designs. Econometrica, 82, 2295-2326.

15.

Calonico

Cattaneo

M. D.

Titiunik

Farrell

M. H.

(2018). rdrobust: Robust data-driven statistical inference in regression-discontinuity designs. R package version 0.99.4. Retrieved from http://CRAN.R-project.org/package=rdrobust

16.

Campbell

(2003). The quiet revolution: Decentralization and the rise of political participation in Latin American cities. Pittsburgh, PA: University of Pittsburgh Press.

17.

Campello

(2014). The politics of financial booms and crises: Evidence from Latin America. Comparative Political Studies, 47, 260-286.

18.

Case

Deaton

(1999). School inputs and educational outcomes in South Africa. Quarterly Journal of Economics, 114, 1047-1084.

19.

Cheeseman

Luna

J. P.

Kitschelt

Paget

Rosenblatt

Sample

Wang

Y.-T.

(2014). Politics meets policies: The emergence of programmatic political parties. Stockholm, Sweden: International Institute for Democracy and Electoral Assistance.

20.

Cleary

M. R.

(2007). Electoral competition, participation, and government responsiveness in Mexico. American Journal of Political Science, 51, 283-299.

21.

Colonnelli

Teso

Prem

(2018). Patronage in the allocation of public sector jobs (Working paper). Retrieved from https://scholar.harvard.edu/files/edoardoteso/files/edoardoteso_jmp.pdf

22.

Committee on Political Parties. (1950). Toward a more responsible two-party system. American Political Science Review, 3, 488-499.

23.

da Silva

R. P

. (1996). Conquistar as Prefeituras do Interior: Contribuição ao Ano Eleitoral [Winning Municipalities in the Countryside: A Contribution to the Electoral Year]. São Paulo, Brazil: Diretório Nacional.

24.

de Janvry

Finan

Sadoulet

. (2012). Local electoral incentives and decentralized program performance. Review of Economics and Statistics, 3, 672-685.

25.

de la Cuesta

Imai

(2016). Misunderstandings about the regression discontinuity design in the study of close elections. Annual Review of Political Science, 19, 375-394.

26.

Donaghy

M. M.

(2013). Civil society and participatory governance: Municipal councils and social housing programs in Brazil. New York, NY: Routledge.

27.

Downs

(1957). An economic theory of political action in a democracy. Journal of Political Economy, 65, 135-150.

28.

Eggers

A. C.

Fowler

Hainmueller

Hall

A. B.

Snyder

J. M.

(2015). On the validity of the regression discontinuity design for estimating electoral effects: New evidence from over 40,000 close races. American Journal of Political Science, 59, 259-274.

29.

Epstein

D. J.

(2009). Clientelism versus ideology: Problems of party development in Brazil. Party Politics, 15, 335-355.

30.

Faguet

J.-P.

(2012.). Decentralization and popular democracy: Governance from below in Bolivia. Ann Arbor: University of Michigan Press.

31.

Ferraz

Finan

Moreira

D. B.

(2012). Corrupting learning: Evidence from missing federal education funds in Brazil. Journal of Public Economics, 96, 712-726.

32.

Ferreira

(2004). Prefeituras do PT são as Mais Reeleitas [PT Administrations are the Most Reelected]. PT Notícias [PT News], 7(156), 1.

33.

Fox

J. A.

(2007). Accountability politics: Power and voice in rural Mexico. Oxford, UK: Oxford University Press.

34.

Fujiwara

(2015). Voting technology, political responsiveness, and infant health: Evidence from Brazil. Econometrica, 83, 423-464.

35.

Gonçalves

(2014). The effects of participatory budgeting on municipal expenditures and infant mortality in Brazil. World Development, 53, 94-110.

36.

Grimmer

Stewart

B. M.

(2013). Text as data: The promise and pitfalls of automatic content analysis methods for political texts. Political Analysis, 21, 267-297.

37.

Grindle

M. S.

(2007). Going local: Decentralization, democratization, and the promise of good governance. Princeton, NJ: Princeton University Press.

38.

Hagopian

Gervasoni

Moraes

J. A.

(2009). From patronage to program: The emergence of party-oriented legislators in Brazil. Comparative Political Studies, 42, 360-391.

39.

Helenda

Bastos

(2000). A Vitória do PT Vermelho [The Victory of the Red PT]. Em Tempo: Publicação da Tendência Democracia Socialista [In Time: Publication of the Democratic Socialist Faction] 318/319, 4-5.

40.

Hicken

(2011). Clientelism. Annual Review of Political Science, 14, 289-310.

41.

Hochstetler

(2008). Organized civil society in Lula’s Brazil. In Kingstone

P. R.

Power

T. J.

(Eds.), Democratic Brazil Revisited (pp. 33-53). Pittsburgh, PA: University of Pittsburgh Press.

42.

Hunter

(2007). The normalization of an anomaly: The Workers’ Party in Brazil. World Politics, 475, 440-475.

43.

Hunter

(2010). The transformation of the Workers’ Party in Brazil, 1989-2009. Cambridge, UK: Cambridge University Press.

44.

Johannessen

(2017, September). Visibility and local electoral accountability. Paper presented at Annual Meeting of the American Political Science Association, Washington, DC.

45.

Keck

M. E.

(1992). The Workers’ Party and democratization in Brazil. New Haven, CT: Yale University Press.

46.

Keefer

(2007). Clientelism, credibility, and the policy choices of young democracies. American Journal of Political Science, 51, 804-821.

47.

Keefer

Vlaicu

(2008). Democracy, credibility, and clientelism. The Journal of Law, Economics, & Organization, 24, 371-406.

48.

Kitschelt

(2000). Linkages between citizens and politicians in democratic polities. Comparative Political Studies, 33, 845-879.

49.

Kitschelt

Hawkins

K. A.

Luna

J. P.

Rosas

Zechmeister

E. J.

(Eds.). (2010). Latin American party systems. Cambridge, UK: Cambridge University Press.

50.

Kitschelt

Wilkinson

S. I.

(Eds.). (2007). Patrons, clients and policies: Patterns of democratic accountability and political competition. Cambridge, UK: Cambridge University Press.

51.

Klašnja

Titiunik

(2017). The incumbency curse: Weak parties, term limits, and unfulfilled accountability. American Political Science Review, 111, 129-148.

52.

Lavareda

Telles

(2015). Como o eleitor escolhe seu prefeito: Campanha e voto nas eleições municipais [How voters choose their mayor: Campaign and vote in municipal elections]. Rio de Janeiro, Brazil: Editora FGV.

53.

Lucas

Nielsen

R. A.

Roberts

M. E.

Stewart

B. M.

Storer

Tingley

(2015). Computer-assisted text analysis for comparative politics. Political Analysis 23, 254-277.

54.

Luna

J. P.

(2014). Segmented representation: Political party strategies in unequal democracies. Oxford, UK: Oxford University Press.

55.

Lupu

Riedl

R. B.

(2013). Political parties and uncertainty in developing democracies. Comparative Political Studies, 46, 1339-1365.

56.

Lyne

M. M.

(2005). Parties as programmatic agents: A test of institutional theory in Brazil. Party Politics, 11, 193-216.

57.

Mainwaring

(1995). Brazil: Weak parties, feckless democracy. In Mainwaring

Scully

T. R.

(Eds.), Building democratic institutions: Party systems in Latin America (pp. 354-398). Stanford, CA: Stanford University Press.

58.

Mainwaring

(2018). Party systems in Latin America: Institutionalization, decay, and collapse. New York, NY: Cambridge University Press.

59.

Mainwaring

Scully

T. R.

(Eds.). (1995). Building democratic institutions: Party systems in Latin America. Stanford, CA: Stanford University Press.

60.

Mani

Mukand

(2007). Democracy, visibility and public good provision. Journal of Development Economics, 83, 506-529.

61.

Manor

(1999). The political economy of democratic decentralization. Washington, DC: World Bank.

62.

Marques

(2000). Mudanças no PT Gaúcho [Changes in Rio Grande do Sul’s PT]. Em Tempo: Publicação da Tendência Democracia Socialista [In Time: Publication of the Democratic Socialist Faction], 313, 6-7.

63.

McCrary

(2008). Manipulation of the running variable in the regression discontinuity design: A density test. Journal of Econometrics, 142, 698-714.

64.

Montenegro

(1996, February 18). Brincando com o Dinheiro do Povo. Jornal Nãção Brasil.

65.

Montero

A. P.

(2012). A reversal of political fortune: The transitional dynamics of conservative rule in the Brazilian Northeast. Latin American Politics and Society, 54, 1-36.

66.

Nichter

Peress

(2017). Request fulfilling: When citizens demand clientelist benefits. Comparative Political Studies, 50, 1086-1117.

67.

Novaes

L. M.

(2018). Disloyal brokers and weak parties. American Journal of Political Science, 62, 84-98.

68.

Partido dos Trabalhadores: Secretaria de Assuntos Institucionais. (1996). O Modo Petista de Governar: Programa de Governo. São Paulo, Brazil: Diretório Nacional do PT.

69.

Pereira

Paschoal

E. S.

(1996). Preparando a Batalha Eleitoral [Preparing for the electoral battle]. São Paulo, Brazil: Diretório Nacional do PT.

70.

Pietá

(1996). O Modo Petista de Governar: Vereadores. São Paulo, Brazil: PT-SP.

71.

Platteau

J.-P.

(2004). Monitoring elite capture in community-driven development. Development and Change, 35, 223-246.

72.

Przeworski

Stokes

Manin

(1999). Democracy, accountability, and representation. New York, NY: Cambridge University Press.

73.

Ranney

(1954). The doctrine of responsible party government: Its origins and present state. Urbana: University of Illinois Press.

74.

Roberts

M. E.

Stewart

B. M.

Tingley

(2016). stm: R package for Structural Topic Models. R package version 1.3.3. Retrieved from https://cran.r-project.org/web/packages/stm/vignettes/stmVignette.pdf

75.

Roberts

M. E.

Stewart

B. M.

Tingley

Lucas

Leder-Luis

Gadarian

S. K.

Rand

D. G.

(2014). Structural topic models for open-ended survey responses. American Journal of Political Science, 58, 1064-1082.

76.

Robinson

J. A.

Verdier

(2013). The political economy of clientelism. The Scandinavian Journal of Economics, 115, 260-291.

77.

Rolnik

Iacovini

R. F. G.

Klintowitz

(2014). Habitação em municípios paulistas: construir políticas ou “rodar” programas?.” Revista Brasileira de Estudos Urbanos e Regionais, 16, 149-165.

78.

Samuels

(2006). Sources of mass partisanship in Brazil. Latin American Politics and Society, 48(2), 1-27.

79.

Sartori

(1976.). Parties and party systems: A framework for analysis. Cambridge, UK: Cambridge University Press.

80.

Scarrow

H. A.

(1967). The function of political parties: A critique of the literature and the approach. The Journal of Politics, 29, 770-790.

81.

Scartascini

Stein

Tommasi

(2010). How democracy works: Political institutions, actors, and arenas in Latin American policymaking. Washington, DC: Islamic Development Bank (IDB).

82.

Spada

(2010, September). The effects of participatory democracy on political competition: The case of Brazilian participatory budgeting. Paper presented at Annual Meeting of the American Political Science Association. Washington, DC.

83.

Stokes

S. C.

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

84.

Stokes

S. C.

Dunning

Nazareno

Brusco

(2013). Brokers, voters, and clientelism: The puzzle of distributive politics. Cambridge, UK: Cambridge University Press.

85.

Touchton

Sugiyama

N. B.

Wampler

(2017). Democracy at work: Moving beyond elections to improve well-being. American Political Science Review, 111, 68-82.

86.

Tranjan

J. R.

(2016). Participatory democracy in Brazil: Socioeconomic and political origins. Notre Dame, IN: University of Notre Dame Press.

87.

Van Cott

D. L

. (2008). Radical democracy in the Andes. New York, NY: Cambridge University Press.

88.

Weitz-Shapiro

(2012). What wins votes: Why some politicians opt out of clientelism. American Journal of Political Science, 56, 568-583.

Supplementary Material

Please find the following supplemental material available below.

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

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

38.82 MB

0.00 MB

0.46 MB