# Abstract

This article addresses the subject of clientelism with reference to an unprecedented set of data resulting from a survey of 422 municipal councilors in 44 Brazilian municipalities. Our aim was to verify whether a propensity for clientelist behavior was uniformly distributed among the councilors surveyed and, in the event that it was not, to identify factors that could explain any variations. Our analysis revealed that clientelism – understood as the degree of exposure on the part of councilors to voter demands for individual benefits, and the councilors' willingness to attend to such demands by means of informal strategies – varies among councilors. Through a multivariate analysis of data, we concluded that these variations are related, as we expected, to such characteristics of the municipalities as population, poverty levels and political competition, and to such individual attributes of councilors as ideology and position in relation to the executive branch. We also conclude that there is an interaction between poverty and competition: competition seems to be more relevant than poverty to explain the observed variations, and its effect is intensified; the explanatory power of poverty is higher in the context of low political competition.

Clientelism; representation; city councilors; local politics; interest intermediation

This article approaches the theme of clientelism with reference to an unprecedented set of data produced by means of a survey of local councilors in the state of Minas Gerais, Brazil. Two questions guided our investigation: is the propensity for clientelism spread evenly among councilors? If not, what are factors explain the variations between them? Recent studies on the subject have presented various means of measuring clientelism, and have taken into account the difficulty of observing it directly. With the exception of ethnographic studies that seek to comprehend clientelism through direct observation, other studies have generally made use of proxies that seek to understand to what extent clientelism characterizes relations between politicians and voters. In this article, and in order to operationalize the specific variable 'interest', we made use of questions addressed to councilors regarding their interactions with voters.

In particular, we sought to verify to what extent local councilors are exposed to demands for private benefits in their respective municipalities and to what degree they are willing to invest their time and resources mobilizing informal strategies in fulfilling such demands. Next, we investigated the factors associated with the differences observed among the city councilors. The survey in question was conducted with a non-representative sample of 422 council members elected in 2012 for the 2013-2016 term from 44 municipalities in all ten planning regions of the state of Minas Gerais, Brazil. The results constitute a rare and valuable dataset providing insights into the daily routines of local councilors that allows for testing of various current hypotheses on the determinants of clientelist behavior.

The most recent studies on the subject treat clientelism as an enduring relationship between individuals with asymmetrical access to resources, one that involves generous doses of personalism and conditionality. In other words, the delivery of goods is contingent on political behavior. In this article, we measured clientelism based on the councilors' answers to questions presented to them by a questionnaire. We considered the following as indicative of clientelist behavior: being contacted by the voters mainly at home, being contacted for the submission of demands of a private nature, and mobilizing informal strategies to fulfill these demands. Our working hypothesis was that neither the degree of exposure on the part of councilors to such private demands nor their willingness to fulfill them by means of informal strategies would be uniformly distributed between council members, and would present variations among the municipalities studied. Moreover, we envisaged the possibility that such variations as were found could be explained by some of each municipality's features, such as population size, poverty levels and political competition, as well as by the individual traits of the city councilors in relation to their profiles, ideologies and positions vis-à-vis the mayor. We expected that clientelist behavior would be more widespread among councilors from smaller and poorer municipalities, among right-wing councilors, and among those who support the mayor.

Largely, our expectations followed the arguments made by other scholars on the subject of clientelism – but these are somewhat controversial. While some scholars argue for a negative relationship (GEDDES, 1991GEDDES, Barbara (1991), A game theoretical model of reform in Latin American democracies. American Political Science Review. Vol. 85, Nº 02, pp. 371-392.), claiming that competition undermines clientelism, others hold precisely the opposite view (LEVITSKY, 2007LEVITSKY, Steven (2007), From populism to clientelism? The transformation of labor-based party linkages in Latin America. In: Patrons, clients, and policies: patterns of democratic accountability and political competition. Edited by KITSCHEKT, Herbert and WILKINSON, Steven I.. Cambridge: Cambridge University Press. pp. 206-226.). Still others claim that the relationship between the two variables is contingent and depends on the mediation of the other factors, such as poverty (WEITZ-SHAPIRO, 2012WEITZ-SHAPIRO, Rebecca (2012), What wins votes: why some politicians opt out. American Journal of Political Science. Vol. 56, Nº 03, pp. 568-583.). Therefore, since the existing literature on the subject is contradictory we did not cleave to any theoretical expectations. We sought to investigate, in the specific context of Brazilian local politics, whether electoral competition affects clientelism negatively or positively.

This article is organized as follows. In the first section, we briefly discuss the concept of clientelism based on the findings of the recent literature on the subject; in the second, we present a discussion regarding recent studies on local politics in Brazil; and in the third, we present information on the survey conducted with the councilors and the results of our data analysis. Finally, in the fourth section, we present our final considerations.

# Clientelism in recent literature: perspectives and convergences

In recent decades, there has been a renewed interest on the subject of clientelism, especially in developing countries. One of the reasons for this renewed interest is its demonstrated ability to survive and adapt to different socioeconomic, demographic, cultural, political and institutional contexts (HICKEN, 2011HICKEN, Allen (2011), Clientelism. Annual Review of Political Science. Vol. 14, pp. 289–310.; KITSCHELT and WILKINSON, 2007KITSCHELT, Herbert and WILKINSON, Steven I. (2007), Patrons, clients, and policies: patterns of democratic accountability and political competition. Cambridge: Cambridge University Press. 377 pp..; PIATTONI, 2001PIATTONI, Simona (ed) (2001), Clientelism, interests, and democratic representation. In: Clientelism, interests, and democratic representation: the European experience in historical and comparative perspective. New York: Cambridge University Press. pp. 193-212.). This adaptability has led to a questioning of the traditional perspective that tended to view clientelism as a pathology resulting from institutional flaws, an incomplete process of democratization, or as the remnant of a pre-modern past that would tend to disappear as democratic institutions gained strength and as the process of modernization advanced (HILGERS, 2011HILGERS, Tina (2011), Clientelism and conceptual stretching: differentiating among concepts and among analytical levels. Theory and Society. Vol. 40, Nº 05, pp. 567–588.). This vast literature, in dialogue with classic studies on the subject (SCOTT, 1977SCOTT, James C. (1977), Political clientelism: a bibliographical essay. In: Friends, followers and factions: a reader in political clientelism. Edited by SCHMIDT, Steffen W.; SCOTT, James C.; LANDÉ, Carl H., and GUASTI, Laura. Berkeley: University of California Press. pp. 483-505.; LANDÉ, 1983LANDÉ Carl H. (1983), Political clientelism in political studies: retrospect and prospects. International Political Science Review. Vol. 04, Nº 04, pp. 435-454.), is characterized by the pursuit of greater conceptual precision in dealing with clientelism (HICKEN, 2011HICKEN, Allen (2011), Clientelism. Annual Review of Political Science. Vol. 14, pp. 289–310.; HILGERS, 2011HILGERS, Tina (2011), Clientelism and conceptual stretching: differentiating among concepts and among analytical levels. Theory and Society. Vol. 40, Nº 05, pp. 567–588.) and by a strong methodological debate focused on the different forms of measuring it (HILGERS, 2011HILGERS, Tina (2011), Clientelism and conceptual stretching: differentiating among concepts and among analytical levels. Theory and Society. Vol. 40, Nº 05, pp. 567–588.)1.

We have observed an increasing use of mixed-method research strategies to approach the theme, including ethnography and case studies (AUYERO, 2001AUYERO, Javier (2001), Poor people's politics: peronist survival networks and the legacy of Evita. Durham: Duke University Press. 272 pp..), a comparative analysis of countries based on socioeconomic indicators produced by official government agencies, and (STOKES et al., 2013STOKES, Susan; DUNNING, Thad; NAZARENO, Marcelo and, BRUSCO, Valeria (2013), Brokers, voters and clientelism: the puzzle of distributive politics. New York: Cambridge University Press. 316 pp..), statistical analysis of opinion data obtained through voter surveys, representatives, brokers and specialists (STOKES et al., 2013STOKES, Susan; DUNNING, Thad; NAZARENO, Marcelo and, BRUSCO, Valeria (2013), Brokers, voters and clientelism: the puzzle of distributive politics. New York: Cambridge University Press. 316 pp..; KITSCHELT and WILKINSON, 2007KITSCHELT, Herbert and WILKINSON, Steven I. (2007), Patrons, clients, and policies: patterns of democratic accountability and political competition. Cambridge: Cambridge University Press. 377 pp..).

A considerable number of specialists on clientelism would agree that it designates an exchange relationship between a patron who 'delivers the goods', and clients who pay for them with their vote and/or some other kind of political support. In this sense, clientelism involves a considerable amount of personalism, and is based on face-to-face relationships between two participants. The relationship is also defined by the asymmetry between patron and client, due to uneven access to resources, and has a repetitious or iterative character. That is to say, the exchange relations involve more than one exchange, and they tend to repeat themselves. Finally, it is a relationship that involves conditionality (HICKEN, 2011HICKEN, Allen (2011), Clientelism. Annual Review of Political Science. Vol. 14, pp. 289–310.). The contingent character of the exchange implies that the benefit is conditional on a past or future vote, and/or political support. In the case of non-conditional delivery, as in a pork barrel relationship, it is not possible to assure a return in the form of votes and/or political support, just as there is no way of restricting the enjoyment of any benefits to those who have offered political support in the past.

Another important distinction refers to the practice of clientelism in the context of authoritarian regimes, or in democratic contexts with constitutional arrangements guaranteeing freedom of press, speech, and assembly, as well as the validity of universal suffrage with 'clean', periodic and competitive elections, and secret ballots. The main difference here concerns the dimensions of the hierarchy, and the asymmetry between patrons and clients regarding the degree of voter autonomy. In a context of prevalent political competition, clients see an increase in their bargaining power, and enjoy a wider margin of choice among different patrons. This increases their capacity to obtain goods, and offers more options for getting out of the relationship. Understanding, therefore, the dynamics of clientelism in conditions of wide political and electoral competition has been one of the challenges faced by a new generation of scholars interested in the subject.

In the quest to understand the main determinants of clientelism, and explain the variations found between countries and societies over time, the poverty factor is prominent. The association between clientelism and poverty has been supported by several scholars on the subject, based on aggregate data and analyses that take individuals as their unit of analysis. Based on data collected at the national level, Stokes et al. (2013)STOKES, Susan; DUNNING, Thad; NAZARENO, Marcelo and, BRUSCO, Valeria (2013), Brokers, voters and clientelism: the puzzle of distributive politics. New York: Cambridge University Press. 316 pp.. assert that there is no doubt that the poorer the country, the greater the likelihood that its policy will be markedly clientelist; that the poorer the voters the greater the chances of their engaging in this type of relationship. The greater propensity of the poor to engage in clientelistic relationships can be explained by the greater marginal utility they attach to immediate material inducements vis-à-vis benefits resulting from programmatic policies and also because their votes come at a bargain in comparison to those from higher income individuals. These factors make them more attractive targets for politicians (STOKES et al., 2013STOKES, Susan; DUNNING, Thad; NAZARENO, Marcelo and, BRUSCO, Valeria (2013), Brokers, voters and clientelism: the puzzle of distributive politics. New York: Cambridge University Press. 316 pp..; WEITZ-SHAPIRO, 2012WEITZ-SHAPIRO, Rebecca (2012), What wins votes: why some politicians opt out. American Journal of Political Science. Vol. 56, Nº 03, pp. 568-583.).

Another central variable cited to explain the incidence of clientelist relations is the size of the electorate. The argument is that small communities are more conducive to maintaining face-to-face interactions and exchange relationships that contain a good deal of personalism and require close monitoring capability. It is assumed that voter behavior can be monitored at a lower cost in rural and small communities than in large cities (NICHTER and PERESS, 2017NICHTER, Simeon and PERESS, Michael (2017), Request fulfilling: when citizens demand clientelist benefits. Comparative Political Studies. Vol. 50, Nº 08, pp. 1086-1117.). Therefore, the increase in the electorate, a fact often accompanied by other demographic and socioeconomic changes such as urbanization, increased levels of income and education, more developed means of transport and communication, makes clientelism a less attractive and more costly strategy for the political elites (BRUSCO, NAZARENO and STOKES, 2004; STOKES et al., 2013STOKES, Susan; DUNNING, Thad; NAZARENO, Marcelo and, BRUSCO, Valeria (2013), Brokers, voters and clientelism: the puzzle of distributive politics. New York: Cambridge University Press. 316 pp..; WEITZ-SHAPIRO, 2012WEITZ-SHAPIRO, Rebecca (2012), What wins votes: why some politicians opt out. American Journal of Political Science. Vol. 56, Nº 03, pp. 568-583.).

A third aspect around which we find less consensus concerns the impact of competitive politics on clientelism. How does increasing party-political competition affect clientelism? Does it tend to undermine it by stimulating the professionalization of bureaucracies and paving the way for a transition to more programmatic forms of representation and intermediation of interests (GEDDES, 1991GEDDES, Barbara (1991), A game theoretical model of reform in Latin American democracies. American Political Science Review. Vol. 85, Nº 02, pp. 371-392.), or does such competition in fact strengthen clientelism (LEVITSKY, 2007LEVITSKY, Steven (2007), From populism to clientelism? The transformation of labor-based party linkages in Latin America. In: Patrons, clients, and policies: patterns of democratic accountability and political competition. Edited by KITSCHEKT, Herbert and WILKINSON, Steven I.. Cambridge: Cambridge University Press. pp. 206-226.)? This last hypothesis assumes that scenarios marked by greater competition between parties, groups and factions increase the importance of undecided voters and the willingness of competitors to employ clientelist strategies in order to gain votes. Weitz-Shapiro (2012)WEITZ-SHAPIRO, Rebecca (2012), What wins votes: why some politicians opt out. American Journal of Political Science. Vol. 56, Nº 03, pp. 568-583., on the other hand, presents some evidence that the relationship between political competition and clientelism is contingent on voter poverty.

In this article, we have conceptualized clientelism as an asymmetrical exchange relationship in which politicians who have privileged access to public resources deliver private benefits to voters in return for political support. Considering the difficulty of measuring clientelism directly, we, following in the footsteps of other scholars, use a proxy variable that indicates the level of councilor exposure to demands for private benefits and their willingness to fulfill them by informal strategies. In addition to examining how widespread clientelist relations among the councilors are, we aim to investigate whether the three previously discussed factors – population size, poverty level and political competition – along with some individual attributes of legislators are associated with clientelism. Before outlining our research, we will briefly discuss in the next section the importance of the topic of clientelism in studies on local politics in Brazil.

# Clientelism and political representation at the local level in Brazil

The subject of clientelism seems to receive relatively more attention in studies on local politics in Brazil (KERBAUY, 2005KERBAUY, Maria Teresa Miceli (2005), As câmaras municipais brasileiras: perfil de carreira e percepção sobre o processo decisório local. Opinião Pública. Vol. 11, Nº 02, pp. 337-365.; LOPEZ and ALMEIDA, 2017LOPEZ, Felix G. and ALMEIDA, Acir (2017), Legisladores, captadores e assistencialistas: a representação política no nível local. Revista de Sociologia Política. Vol. 25, Nº 62, pp. 157-181.) than in other countries. Clientelism comes up frequently in studies that investigate municipal political dynamics, especially when the focus is on city councilors, the workings of city councils, and the relations between the executive and legislative branches in municipalities.

In fact, there seems to be something of a dispute about the place of municipalities in Brazilian political life, and their role when considering the political system as a whole. Traditionally, at least until the mid-1970s, municipal politics was associated with backwardness, tradition, and oligarchic power based on personal, strongly vertical, and asymmetrical relationships, "a territory where democratic projects were frustrated or perverted" (ALMEIDA and CARNEIRO, 2003ALMEIDA, Maria Hermínia Tavares and CARNEIRO, Leandro Piquet (2003), Liderança local, democracia e políticas públicas no Brasil. Opinião Pública. Vol. 09, Nº 01, pp. 124-147., p. 125). However, the redemocratization process in the 1970s and 1980s resulted in an increased appreciation of decentralization. Furthermore, municipalities came to be seen as a space for the renewal of the political system, the emergence of innovative forms of public management and mechanisms of popular participation with the potential to transform and democratize the relations between state and society.

The effects of the decentralization project and the municipalist discourse were felt in the new institutional arrangement enshrined in the Constitution of 1988. This legislation, besides transforming the municipalities into federative entities, granting them political, administrative, and fiscal autonomy, reserved for municipalities important prerogatives and competences in the management of social policies – in the areas of health, education, social assistance, among others – with a direct impact on citizens' lives. The municipalities were also the main beneficiaries of the decentralization of resources and constitutional transfers[2].

Unlike other countries that have adopted federalism, Brazil exhibits great institutional uniformity, leaving little room for subnational levels to define its institutional design. The Federal Constitution, in addition to recognizing the municipalities as federal entities and defining their competencies and prerogatives, also lays down the rules that must be used for local elections throughout the country. Councilors are elected by a proportional system with a personalized vote and mayors by the majority method, with a second round in municipalities with more than 200,000 voters, both for a four-year term. Beyond the electoral system, the Constitution also establishes the number of councilors per chamber, based exclusively on the population of the municipality and the expenditure limits for its city council.

Regarding the distribution of prerogatives and resources at the local level, mayors account for a large part of the decision-making prerogatives in the administrative, economic and budgetary spheres. Councilors, in addition to having the prerogatives of control and supervision of the executive branch and being responsible for approving the mayor's accounts, may submit amendments to budget and bills, within the limits of their authority. However, part of the literature on local representation in Brazil highlights the fact that such bills do not represent the bulk of councilors' legislative output. This is dominated by another type of action, in general not classified as legislative because it has no impact on the legal order, namely, so-called 'indications' – the service provided by legislators through these indications resembles what is sometimes referred to as 'constituency service' – and will be further discussed below.

Recent studies on the role of municipalities in the Brazilian political system and political representation at the local level have sought to address some central questions about local politics in a systematic way, based on empirical data. One of these questions concerns the extent of the clientelist character of relations between representatives and those they represent (ALMEIDA and CARNEIRO, 2003ALMEIDA, Maria Hermínia Tavares and CARNEIRO, Leandro Piquet (2003), Liderança local, democracia e políticas públicas no Brasil. Opinião Pública. Vol. 09, Nº 01, pp. 124-147.; KERBAUY, 2005KERBAUY, Maria Teresa Miceli (2005), As câmaras municipais brasileiras: perfil de carreira e percepção sobre o processo decisório local. Opinião Pública. Vol. 11, Nº 02, pp. 337-365.; LOPEZ and ALMEIDA, 2017LOPEZ, Felix G. and ALMEIDA, Acir (2017), Legisladores, captadores e assistencialistas: a representação política no nível local. Revista de Sociologia Política. Vol. 25, Nº 62, pp. 157-181.). These studies show that interest in local politics is on the rise among Brazilian social scientists.

An element common to these studies is the attempt to question some dichotomies and stereotypical views that survive when it comes to local politics and the role of councilors. For example, Kerbauy (2005)KERBAUY, Maria Teresa Miceli (2005), As câmaras municipais brasileiras: perfil de carreira e percepção sobre o processo decisório local. Opinião Pública. Vol. 11, Nº 02, pp. 337-365. suggests that local politics is characterized by the coexistence of two distinct rationales that include "the traditionalism and clientelist actions that have always characterized local power" and "the universal procedures that would characterize more innovative measures of local leaders" (KERBAUY, 2005KERBAUY, Maria Teresa Miceli (2005), As câmaras municipais brasileiras: perfil de carreira e percepção sobre o processo decisório local. Opinião Pública. Vol. 11, Nº 02, pp. 337-365., p. 362). Lopez and Almeida (2017)LOPEZ, Felix G. and ALMEIDA, Acir (2017), Legisladores, captadores e assistencialistas: a representação política no nível local. Revista de Sociologia Política. Vol. 25, Nº 62, pp. 157-181. similarly conclude that "local representation is exercised following both clientelist and universalist principles, with the incidence of clientelism being less than the literature and common sense usually indicate or presuppose" (LOPEZ and ALMEIDA, 2017, pLOPEZ, Felix G. and ALMEIDA, Acir (2017), Legisladores, captadores e assistencialistas: a representação política no nível local. Revista de Sociologia Política. Vol. 25, Nº 62, pp. 157-181., p. 34).

Several case studies have shown the predominance of clientelist relations between representatives and voters in small municipalities. In these, personal relations between councilors and voters, based on the exchange of private benefits for political support, predominate. Here we have a rationale based on contacts, networks, and the vision of the councilor as a 'social worker' who, by virtue of his position, acts as a mediator in obtaining particular benefits for his constituents. The most frequent activities in the daily life of councilors are making appointments, engaging in legislative activities, and working on the construction and maintenance of ties with voters, support groups, and members of the bureaucracy in order to facilitate the exercise of this mediation function in obtaining particular benefits for voters and allies (D'ÁVILA FILHO et al., 2014; KUSCHNIR, 2000KUSCHNIR, Karina (2000), O cotidiano da política. Rio de Janeiro: Relume-Dumará. 162 pp..; LOPEZ, 2004)LOPEZ, Felix G. (2004), A política cotidiana dos vereadores e as relações entre executivo e legislativo em âmbito municipal: o caso do município de Araruama. Revista de Sociologia Política. Nº 22, pp. 153-177..

The 'welfare' component of councilors' behavior is explained by the low capacity of municipal legislatures to influence the process of public policy production. At the same time, it deals with the need to deliver benefits to voters (ANDRADE, 1998ANDRADE, Régis Stephan de Castro (ed) (1998), Processos decisórios na Câmara dos Vereadores e na Assembleia Legislativa de São Paulo. In: Processo de governo no município e no estado: uma análise a partir de São Paulo. São Paulo: EDUSP/FAPESP. pp. 15-40.; D'ÁVILA FILHO et al., 2014; LOPEZ, 2004)LOPEZ, Felix G. (2004), A política cotidiana dos vereadores e as relações entre executivo e legislativo em âmbito municipal: o caso do município de Araruama. Revista de Sociologia Política. Nº 22, pp. 153-177., and may be based on the subjective motivations of the actors involved in clientelist networks (KUSCHNIR, 2000)KUSCHNIR, Karina (2000), O cotidiano da política. Rio de Janeiro: Relume-Dumará. 162 pp...

Some studies see clientelist logic in the exercise of political representation at the municipal level, since it is marked by personal relationships with little mediation via institutions such as political parties. Nonetheless, there is evidence that clientelism is not as widespread a pattern as is commonly thought. Its reach varies according to a number of factors, while more institutional dynamics, similar to those seen in other spheres of government, may be found at the municipal level (KUSCHNIR, 2000KUSCHNIR, Karina (2000), O cotidiano da política. Rio de Janeiro: Relume-Dumará. 162 pp..; LOPEZ and ALMEIDA, 2017LOPEZ, Felix G. and ALMEIDA, Acir (2017), Legisladores, captadores e assistencialistas: a representação política no nível local. Revista de Sociologia Política. Vol. 25, Nº 62, pp. 157-181.). Scholars also suggest that, despite the institutional homogeneity of the three spheres of Brazilian government, it is impossible, given the scarce literature, to make general statements about patterns of political interaction at the local level.

# Research outline and data analysis

Our analysis was carried out based on a survey on 422 municipal councilors in 44 municipalities in all ten planning regions of Minas Gerais, Brazil. Our sample is not representative of the whole set of councilors in Minas Gerais; for this reason, our conclusions cannot be generalized to all municipalities and councilors. However, we can say that there are many similarities between our sample and Minas Gerais state in general as regards municipalities and councilors elected in the 2012 election3 (See Methodological Notes – Tables 01-05).

The restriction of our sample to one state was due to practical reasons (limited funding and staff), but also due to the possibility of retaining some variables as constants and, where necessary, inquiring about local politics in diverse settings (SNYDER, 2001SNYDER, Richard (2001), Scaling down: the subnational comparative method. Studies in Comparative International Development. Vol. 36, Nº 01, pp. 93–110.). Indeed, the great demographic and socioeconomic variety of Minas Gerais allowed us to leverage this variation to investigate the problems that interest us in multiple settings.

The state of Minas Gerais is home to 853 municipalities, and its territory covers an area of 588,384.30 km2, making it larger than Spain. The state can be considered, in some aspects, as a kind of amalgam of Brazil. Despite being part of the Southeast macro-region, which is the wealthiest and most developed in the country, a significant amount of its territory presents features typical of the Northeast macro-region, the poorest and less developed. Due to its geographical extension and socioeconomic and demographic diversity – Minas Gerais is the second largest electoral district in the country – some analysts consider the state a national electoral bellwether. In the period from 1994 to 2014, presidential election results in Minas Gerais were the closest to those registered throughout Brazil4.

To select municipalities, we used two stratification criteria: the number of inhabitants/size of the municipality, and the level of poverty. Municipalities were selected in three size ranges: those with up to 20,000 inhabitants (small), from 20,000 to 100,000 (medium) and above 100,000 inhabitants (large) (IBGE, 2010IBGE (2010), Indicadores sociais municipais. Uma análise dos resultados do universo do Censo demográfico 2010. Instituto Brasileiro de Geografia e Estatística. Available at ˂https://biblioteca.ibge.gov.br/visualizacao/livros/liv54598.pdf˃. Accessed on October, 25, 2013.
https://biblioteca.ibge.gov.br/visualiza...
). There were three poverty ranges: municipalities with up to 20% of their populations categorized as poor (wealthy); from 20% to 40% poor (median poverty); and above 40% poor (poor) (PNUD, 2013PNUD, Programa das Nações Unidas para o Desenvolvimento (2013), Atlas do Desenvolvimento Humano no Brasil. Available at ˂http://www.atlasbrasil.org.br/2013/pt/o_atlas/idhm˃. Accessed on April, 30, 2013.
http://www.atlasbrasil.org.br/2013/pt/o_...
). From a total 489 councilors, we were able to interview 422, which represents 86.3% of the total5. (For more information on the fieldwork and for a full version of the questionnaire, see Methodological Notes).

Still on our methodological choices, it is important to consider a problem that is always present in surveys on sensitive issues such as clientelism and vote buying. Namely, the social desirability bias inherent in the risk of significant underestimates regarding the practice being observed. In our research, we have tried as much as possible to minimize this bias through a careful process in the creation of the questionnaire. This took two years and included two pre-tests in ten different municipalities. In fact, one of the most important lessons we learned is that councilors, broadly speaking, do not share social scientists' prejudices against clientelism. For the most part, they talk quite naturally about their daily routine and interactions with voters. In fact, as stated by Kuschnir (2000)KUSCHNIR, Karina (2000), O cotidiano da política. Rio de Janeiro: Relume-Dumará. 162 pp.. and De Rozas (2017)DE ROZAS, Victoria Ortiz (2017), Personal, particular y político: la producción del bien público en la práctica política cotidiana. Revista de Sociologia Política. Vol. 25, Nº 63, pp. 53-77., and as our experience in the field demonstrated, for a great number of councilors, the ability to respond to voter demands is a source of pride as a sign of prestige and access to administrative channels. Therefore, we have no reason to believe that underreporting has compromised the validity of our data.

Based on the available data, and as a proxy for clientelism, we designed a means of measurement that seeks to capture councilors' predispositions to clientelism, and takes into consideration the degree of exposure of city councilors to private demands, as well as their willingness to fulfill them through informal means. We analyzed the councilors' responses to five questions presented to them. We asked: 01. how often councilors were contacted by voters; 02. in which places they were mainly contacted by them; 03. the main reasons for that contact; 04. the main types of requests they received; and 05. what were their strategies for addressing voter requests6.

The following tables show the descriptive statistics of each variable and the results of association tests.

Figures 01 to 05 suggest that most of the municipal councilors interviewed were continuously exposed to demands, appeals and requests from voters in their daily lives. In response to these, they could deploy a variety of mechanisms. More than 60% of respondents were contacted by voters on a daily basis. As for the locus of such contact, answers varied greatly in view of the open-ended nature of the question. Councilors' private residences stood out as the place of choice.

Figure 01
With what frequency are you contacted by voters?

Figure 02
In which location are you contacted most by your voters?

Figure 03
What is the main reason for voters contacting you?

Figure 04
What is the most frequent request presented to you by voters?

More than 60% of respondents said such contacts were in the main requests. More than 70% stated that such requests were mostly of a private nature. As 'private benefits', we have considered employment requests, money, material inducements, school admissions, medical assistance/ambulance/hospital vacancies, legal counsel/lawyer's services, transportation, assistance in document acquisition, assistance in obtaining access to public agencies and services – as opposed to improvements in their neighborhoods and/or support for collective causes. The use of informal means to address requests is not as common as might be expected; however, it is still significant, representing over 40% of councilor's first choices. As 'informal strategies', we included privately contacting the mayor or other members of the municipal administration, and using their own resources or mobilizing a personal network, as opposed to presenting a bill, referring to the relevant agency or submitting an application or petition.

Figure 05
What is the most usual way you address voter requests?

In the process of creating our index of predisposition toward clientelism, we transformed all variables into binary ones (considering only the first response of the councilor), and analyzed the correlation between them. Table 01 shows the results of the tests of independence and statistical significance among the categorical variables.

Table 01
Chi-square independence tests results

Table 01 shows statistically significant and positive associations between maintaining daily interactions with the voters and being contacted for private benefit requests; between being contacted at home and receiving demands for private benefits; between being contacted for inducements for private benefits and the mobilization of informal strategies to meet requests; and between being contacted for requests for private benefits and these being addressed through informal strategies. Because daily interactions are very widespread among councilors, they did not seem to be a suitable criterion, so we removed this dimension when constructing the index. Indeed, daily contact does not necessarily suggest a clientelist relationship, but may be simply an activity inherent to the routine of a responsible politician7.

We also observed that both daily contact and requests at home presented a positive and statistically significant connection with the size of a municipality. This suggests that, as has been widely discussed in the literature (NICHTER and PERESS, 2017NICHTER, Simeon and PERESS, Michael (2017), Request fulfilling: when citizens demand clientelist benefits. Comparative Political Studies. Vol. 50, Nº 08, pp. 1086-1117.), the size of a municipality is important when thinking about the type of interactions between representatives and citizens in terms of their frequency and intensity, as well as the spaces in which they occur (to what extent they are face-to-face or mediated by institutions and organizations). It is also important to take into account the fact that neither the frequency with which demands are presented nor their locus us everything about the nature of the relationship.

Based on the results of the tests of independence, and on the evidence presented by the literature, we created an index composed of the sums of four variables: 01. contacted mainly at home; 02. contacted for the submission of applications; 03. contacted for requests of a private nature; 04. mobilization of informal strategies to respond to requests. We chose to keep requests at home as one of the components of the index, considering that less than a quarter of city councilors pointed to this alternative and that it presented a statistically significant association with two others (demands for private benefits and informal strategies). Although there is a correlation between being contacted at home and the municipality being small, the variations among the councilors within the small municipalities suggest that being contacted at home is not simply a function of the municipality size and can be a good indication of the degree of personalism in relationships.

The evidence gathered in the literature on the subject, and the independence tests suggested the need to give more weight to the last two variables, as these seem more expressive of clientelist exchange than the first two. By giving greater weight to the private benefits and informal responses, we are attributing value to two central aspects of clientelist exchanges: particularism and informality. By doubling the contribution of the last two variables, the resulting index presented a 0 to 06 variation. The higher the score on the index, the greater the councilor's predisposition toward clientelism. The distribution of councilors by the Predisposition Toward Clientelism Index (PCI) is presented in Figure 06.

Figure 06
PCI distribution

The distribution of the index presents a slight negative or leftward asymmetry (asymmetry = -0.237, median = 3, mean = 3.306), indicating that, in general, there is a tendency on the part of the councilors surveyed towards clientelism. The most frequent scores are 03, 04 and 05. Councilors classified with these scores represent 56.9% of the sample. The observed negative asymmetry, which makes the three lowest scores the least frequent, suggests that clientelist-friendly behavior is relatively normal for the councilors surveyed, but less frequent than might be expected considering the conventional wisdom on the behavior of city councilors in Brazil.

Furthermore, as part of our exploratory analysis, we analyzed the PCI distribution in four different scenarios with regard to poverty and electoral competition in the municipality, following a suggestion from Weitz-Shapiro (2012)WEITZ-SHAPIRO, Rebecca (2012), What wins votes: why some politicians opt out. American Journal of Political Science. Vol. 56, Nº 03, pp. 568-583.. We classified as 'poor' municipalities with more than 20% of the population living in conditions of poverty, and as having 'low competition' those in which the average vote concentration by the elected councilors in the period 1996-2012 was greater than 40% of valid votes8. The scenarios resulted from the combination of poverty and competition were: poor municipalities with competitive elections (C_P), poor municipalities with non-competitive elections (NC_P), non-poor municipalities with competitive elections (C_NP), and non-poor municipalities with non-competitive elections (NC_NP).

Political competition seems to be negatively associated to the Predisposition Toward Clientelism Index (PCI), as shown in Table 02: the PCI mean for municipalities with low competition is higher than other municipalities, be they poor or not. Moreover, poverty seems to strengthen the effects of low competition since there is an increase of 27% in the PCI mean when we compare non-poor and poor municipalities with low levels of competition, while this increase is a mere 08% among high competition municipalities. In other words, the explanatory power of poverty appears to be higher in the context of low political competition. Competition seems to be more relevant than poverty to explain index variations and intensifies the effect of poverty: there is an increase of 55% in the PCI mean when we compare poor municipalities with low and high competition, while this difference is only 29% among non-poor municipalities.

We used boxplots to compare the PCI distribution for each combination of competition by poverty level.

In Figure 07, we can see that the PCI distribution pattern presents the following characteristics: higher PCIs are usually observed in municipalities with little competition in the election (NC) in comparison to the more competitive ones (C); there is also a smaller dispersion of the PCIs in municipalities with low levels of competition; the distribution of the PCIs seems to be quite symmetrical at each level of competition and poverty; the variances of the indexes are similar in non-poor municipalities (see Table 02), regardless of competition, and are different in poor municipalities, with different degrees of competition.

Figure 07
PCI by levels of competition and poverty

Table 02
Descriptive PCI statistics by level of competition and poverty

The descriptive analysis pointed to an expected association between the PCI and all our variables of interest: 01. municipality size, with up to 20,000 inhabitants (small), from 20,000 to 100,000 inhabitants (medium) and with more than 100,000 inhabitants (large), 02. the level of electoral competition isolated; 03. the level of poverty in the municipality isolated; 04. the self-declared ideology of the councilor, and 05. the councilor's political position. In order to devise the boxplots we transformed all the variables into binary ones. We observed that the highest scores in the PCI were found among councilors who said they supported the mayor (Figure 08), are self-declaredly right wing (Figure 09), and represent poor, small municipalities with low electoral competition (Figures 10-12 ).

Figure 08
PCI by the councilor's political position

Figure 09
PCI by self-declared ideology

Figure 10
PCI by municipality size

Figure 11
PCI by poverty

Figure 12
PCI by level of electoral competition

The next step was to analyze the impact of these variables – characteristics of the context and individual attributes of the councilors – over the legislator's predisposition toward clientelism (PCI). Regarding the context, we took the municipality size 'small' as our reference category, and the variable that resulted from the combination of poverty and level of electoral competition in the municipality with 'non-poor municipalities with competitive elections (C_NP)' as our baseline. As personal attributes, we considered the councilor's political position, that is, whether s/he supported the mayor or not, and his or her self-declared ideology9, if s/he considered him/herself to be center-right, right wing or not10. The dependent variable is an ordinal one; the Predisposition toward Clientelism Index (PCI) has seven levels.

Based on what has already been verified by other researchers (NICHTER and PERESS, 2017NICHTER, Simeon and PERESS, Michael (2017), Request fulfilling: when citizens demand clientelist benefits. Comparative Political Studies. Vol. 50, Nº 08, pp. 1086-1117.), we expected that the smaller and poorer the municipality, the greater the councilor’s predisposition toward clientelism would be. Following Nichter and Peress (2017)NICHTER, Simeon and PERESS, Michael (2017), Request fulfilling: when citizens demand clientelist benefits. Comparative Political Studies. Vol. 50, Nº 08, pp. 1086-1117., and many other studies, we assumed that a smaller municipality would facilitate face-to-face interaction, and increase the councilor's willingness to comply, as it would offer better conditions for monitoring voter behavior and obtaining the expected electoral return. The extent of poverty would indicate a larger contingent of people in material deprivation and with difficulty in accessing public goods and services. As already stated, there is no consensus in the literature on how increasing electoral competition affects clientelism. In this paper, we aimed to verify if, as highlighted by Weitz-Shapiro (2012)WEITZ-SHAPIRO, Rebecca (2012), What wins votes: why some politicians opt out. American Journal of Political Science. Vol. 56, Nº 03, pp. 568-583., regarding municipalities in Argentina, the effect of competitiveness is, in fact, mediated by the level of poverty11.

Traditionally, specialized literature has treated political particularism and clientelism as a more widespread behavior among right-wing politicians, while left-wing politicians would be more amenable to programmatic politics. This would be due to the particularities in the trajectories of politicians belonging parties to the Left and to the Right (MARENCO and SERNA, 2007MARENCO, André and SERNA, Miguel (2007), Por que carreiras políticas na esquerda e na direita não são iguais? Recrutamento legislativo em Brasil, Chile e Uruguai. Revista Brasileira de Ciências Sociais. Vol. 22, Nº 64, pp. 93-113.; POWER, 2000POWER, Timoty J. (2000), The political right in postauthoritarian Brazil: elites, institutions, and democratization. University Park: The Pennsylvania State University Press. 304 pp..). As regards political stances, we expected that the predisposition towards clientelism would be greater among councilors who support the mayor and thus enjoy more opportunities to adopt clientelist strategies, by mobilizing, among other mechanisms, informal contact with the mayor and the public bureaucracy to serve their clientele. Opposition councilors, we supposed, would find it more difficult to take advantage of the resources administered by the Executive Branch.

Below, we present the analysis of the multilevel ordinal response data PCI using the proportional odds model, which implies that the effect of any explanatory variable remains constant regardless of the particular split to the data being considered. The analysis was carried out with the R package 'ordinal', (CHRISTENSEN, 2019CHRISTENSEN, Rune Haubo Bojesen (2019), Regression models for ordinal data. R package version 2019.4-25. Available at ˂http://www.cran.r-project.org/package=ordinal/˃. Accessed on May, 07, 2019.
http://www.cran.r-project.org/package=or...
; R CORE TEAM, 2019)R CORE TEAM (2019), R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at URL ˂https://www.R-project.org/˃. Accessed on May, 07, 2019.
https://www.R-project.org/˃...
which yields ML estimates using adaptive Gaussian quadrature. The multilevel model for ordinal responses takes into account the ordinal nature of the response (PCI), and the hierarchical structure of the phenomenon. Moreover, a councilor's predisposition toward clientelism depends not only on the characteristics of the municipality but also on the councilor's own traits. Therefore, a comparison among municipalities calls for the calculation of net measures adjusting for individual characteristics. To this end, multilevel modelling is a well-suited technique (AGRESTI and NATARAJAN, 2001AGRESTI, Alan and NATARAJAN, Ranjini (2001), Modeling clustered ordered categorical data: a survey. International Statistical Review. Nº 69, pp. 345-371.; GRILLI and RAMPICHINI, 2009GRILLI, Leonardo and RAMPICHINI, Carla (2009), Multilevel models for the evaluation of educational institutions: a review. In: Statistical methods for the evaluation of educational services and quality of products. Edited by MONARI, Paola; BINI, Matilde; PICCOLO, Domenico and, SALMASO, Luigi. Heidelberg: Physica-Verlag Heidelberg. pp. 61-80.; HEDEKER, 2008)HEDEKER, Donald (2008), Multilevel models for ordinal and nominal variables. In: Handbook of multilevel analysis. Edited by LEEUW, Jan de and MEIJER, Erik. New York: Springer-Verlag. pp. 239–276.. The main goal of this analysis was to estimate the effect of context variables over the councilor's predisposition towards clientelism (PCI) while adjusting for councilor's characteristics. The ordinal response PCI was studied via the random intercept cumulative model using the logit link and C=7 categories:

$\mathrm{log}it\left({Y}_{cij}\right)={\alpha }_{c}-{\alpha }_{c}-\left(\mathrm{X}{\text{'}}_{ij}+\mathrm{Z}{\text{'}}_{\mathrm{j}}+{\mathrm{u}}_{\mathrm{j}}\right)c=1,2,3,4,5,6$

where j= 1,2, ... ,44 is the municipality index and i is the councilor index, while Yci j is the cumulative probability up to the c-th category for councilor i in municipality j. The covariate vector X'ij includes the councilor's characteristics and Z'j the municipality's characteristics, whereas the term uj is a random effect representing unobserved factors at the municipality level interpretable as 'perceived clientelism'.

For the data analysis, we followed the 'step-up' modelling strategy. The number of municipalities evaluated was 44 and the number of councilors 403, while the number of councilors per municipality varied from five to 31. To test whether the cluster variance is important, we compared the models with and without random effects. The results are reported in Table 03. The LRT statistic is 48.869 with 1 df and with a tiny p-value, the null hypothesis is easily rejected. Therefore, there is evidence of unobserved heterogeneity at municipality level. In other words, there is variation among municipalities regarding the PCI distribution, as expected. The null model, that is, the random intercept model without covariates, is a benchmark for subsequent models and provides a cluster variance ${\stackrel{^}{\sigma }}_{u}^{2}=0,74\left({\stackrel{^}{\sigma }}_{u}=0,86\right)$. The amount of municipality-level unobserved heterogeneity is summarized by the Intraclass Correlation Coefficient (ICC) = 0,185: this suggests that about 18% of the total variability in the PCI is at the municipality level.

Table 03
Likelihood ratio tests of cumulative link models: with and without random effects

A series of models was fitted to the data. Table 04 shows some statistics for three models: a random coefficient model without covariates (Model 01), a random coefficient model with individual covariates (Model 02), and the full model including context predictors (Model 03). Variation in the slopes was not considered.

Table 04
Results for three multilevel ordinal models (proportional odds)

Each variable on the councilor level was considered separately, consistent with Raudenbush and Bryk's (2002) suggestions regarding strategies for building complex multilevel models. Based on these preliminary analyses, two were selected that were found to be associated with the PCI in the simpler (univariate) models: ideology and support for the mayor. The fixed effects results for Model 01 are used to provide probability predictions for a councilor being at or below a given level of the PCI. With no explanatory variables in the model, the cumulative logit prediction on average across municipalities for PCI level less than or equal to first level on the ordinal PCI scale is -2.67 (odds ratio = 0.07) and steadily increases across the cutpoints (or thresholds). Model 2 includes the councilor characteristics ideology (P-value = 0.002) and support for the mayor (P-value = 0.067).

Each variable on the councilor level was considered separately, consistent with Raudenbush and Bryk's (2002) suggestions regarding strategies for building complex multilevel models. Based on these preliminary analyses, two were selected that were found to be associated with the PCI in the simpler (univariate) models: ideology and support for the mayor. The fixed effects results for Model 01 are used to provide probability predictions for a councilor being at or below a given level of the PCI. With no explanatory variables in the model, the cumulative logit prediction on average across municipalities for PCI level less than or equal to first level on the ordinal PCI scale is -2.67 (odds ratio = 0.07) and steadily increases across the cutpoints (or thresholds). Model 2 includes the councilor characteristics ideology (P-value = 0.002) and support for the mayor (P-value = 0.067).

Comparing Model 02 with the null model, the municipality-level variance is nearly unchanged: this means that in the linear model for the underlying predisposition toward clientelism, the reduction of the level-2 variance due to the covariates is similar to the reduction of the level-1 variance. The municipality-level variance can be reduced by municipality-level covariates, such as the population size, poverty levels and political competition.

To fit the third model, we included the municipality variables – electoral competition and poverty combined in a four-level variable and the municipality size (three levels based on the number of inhabitants). It is worth noting the decrease in variance of the random effect when context variables were added, which suggests that these characteristics explain much of the variability in the PCI level among municipalities.

In Model 02, in which we included only the individual attributes of legislators, we observed that the councilors who declared themselves right wing are 1.97 times more likely to score at the highest levels of the PCI than otherwise.

Those who declared alignment with the mayor are 1.46 times more likely to score at the highest levels of the PCI compared with those who do not support the mayor. Similar results were found in the more complete Model 03.

The coefficients in Model 03 indicate that being in small and poor municipalities with non-competitive elections, self-declaring right wing and supporting the mayor increase the PCI, i.e., higher level ratings are more likely. A logit of zero corresponds to an odds or odds ratio of 1.0 (no effect); a negative logit corresponds to a greater probability of being at or below that cutpoint (lower PCI level), and a positive logit corresponds to a lower likelihood of being at or below that cutpoint and, therefore, an increased likelihood of being above that cutpoint (higher PCI level). For example, the likelihood of a councilor in a poor municipality with non-competitive elections being classified with higher PCI levels is 2.61 times that of a councilor in a non-poor municipality with competitive elections. Visual inspection of residual Harrell's plot (HARRELL, 2001) to assess proportional odds assumption and a Normal QQ plot to assess normality and magnitude of the random effects do not indicate violations of these assumptions. It is important to note that Harrell's (2001) plot does not allow for the inclusion of random effects that could make the difference in evaluating the assumptions behind the model.

In order to flesh out our analysis we can take some municipalities as examples. Frei Gaspar, a municipality in Jequitinhonha/Mucuri Valley, is one of the poorest regions in Minas Gerais. The town has 5,879 inhabitants (2010) and 30.53% of its population lives in conditions of poverty. The town scored the highest average PCI (5.64), considering the answers of the eight councilors interviewed. From 1996 to 2012 the elected councilors in the municipality accounted for, on average, 41.03% of the votes. The town is one of the cases classified as poor municipalities with non-competitive elections (NC_P). Another case in this group in which electoral competition is low and councilor PCI is high is the municipality of São João do Pacuí, a small town located in the North of the state, another poor region. In the opposed group of 'non-poor municipalities with competitive elections' we find the six largest municipalities of the sample, all with more than 100,000 inhabitants and less than 08% of the population living in poverty: Belo Horizonte (the state capital), Juiz de Fora, Uberaba, Coronel Fabriciano, Pouso Alegre and Ubá. In all of them, the elected councilors concentrated in the 1996-2012 period, on average, less than 31% of the votes. The average PCIs are among the lowest, ranging from 1.51 to 2.58.

Regarding the relationship between competition, poverty and clientelism, Lopez and Almeida (2017)LOPEZ, Felix G. and ALMEIDA, Acir (2017), Legisladores, captadores e assistencialistas: a representação política no nível local. Revista de Sociologia Política. Vol. 25, Nº 62, pp. 157-181., in investigating the behavior of a sample of Minas Gerais councilors came to a different conclusion. In the study, the likelihood of a councilor being classified as paternalist, which in their analysis meant being more prone to clientelism, was higher in small municipalities, as we also observed. However, according to the authors, this likelihood tends to increase with the exacerbation of political competition, which did not have its effect moderated by the extension of the poverty (LOPEZ and ALMEIDA, 2017LOPEZ, Felix G. and ALMEIDA, Acir (2017), Legisladores, captadores e assistencialistas: a representação política no nível local. Revista de Sociologia Política. Vol. 25, Nº 62, pp. 157-181., p. 173)12.

Among the individual attributes, we have verified, as expected, a positive and statistically significant association between being right wing and the willingness towards clientelism. Therefore, the analysis supports the claim that the predisposition towards clientelist exchange varies according to the ideological inclinations of the legislators and that right-wing politicians, less committed to political parties and social movements (MARENCO and SERNA, 2007MARENCO, André and SERNA, Miguel (2007), Por que carreiras políticas na esquerda e na direita não são iguais? Recrutamento legislativo em Brasil, Chile e Uruguai. Revista Brasileira de Ciências Sociais. Vol. 22, Nº 64, pp. 93-113.), tend to seek support based on personal and informal relations with the voters. The analysis also corroborates the hypothesis that councilors who support the mayor are more prone to clientelism.

By way of conclusion, it is important to highlight some of the peculiarities of the Brazilian electoral and party system that could help contextualize our findings and better define the scope of our conclusions. Studies on distributive politics stress the impact of different electoral systems over the propensity of politicians and voters to engage in clientelistic exchanges. In the case of Brazil, in particular, scholars underline the incentives generated by a candidate-centered system with multi-party competition (AMES, 1995AMES, Barry (1995), Electoral rules, constituency pressures, and pork barrel: bases of voting in the Brazilian Congress. Journal of Politics. Vol. 57, Nº 02, pp. 324–343.; MAINWARING, 1991MAINWARING, Scott (1991), Politicians, parties and electoral systems: Brazil in comparative perspective. Comparative Politcs. Vol. 24, Nº 01, pp. 21-43.). According to them, there are strong incentives for individualistic behavior, either during the campaign, when candidates seek to emphasize their personal qualities in detriment of the party program, or once in office. Among other consequences, the open-list electoral system tends to strengthen personalism in politics, which, in theory, could contribute to increased incentives for clientelistic behavior. However, as stated by Hicken (2011)HICKEN, Allen (2011), Clientelism. Annual Review of Political Science. Vol. 14, pp. 289–310., in the current state of the studies it is not possible to establish precisely whether electoral and party rules have an independent effect on the level of clientelism (HICKEN, 2011HICKEN, Allen (2011), Clientelism. Annual Review of Political Science. Vol. 14, pp. 289–310., p. 306). This occurs because to some extent the institutional choices are themselves affected by preferences regarding clientelism, that is, politicians more prone to engage in this type of political exchange will tend to opt for rules that reinforce it. Indeed, we can find very similar exchange patterns to what we described here – characterized by voter requests for private benefits followed by strong efforts on the part of the councilors to fulfill them – under different electoral rules, as observed by De Rozas (2017)DE ROZAS, Victoria Ortiz (2017), Personal, particular y político: la producción del bien público en la práctica política cotidiana. Revista de Sociologia Política. Vol. 25, Nº 63, pp. 53-77. in Argentina.

# Concluding remarks

In this article, we sought to approach the subject of clientelism from a set of data produced with a sample of city councilors from the state of Minas Gerais. We sought to answer two questions: is clientelism disseminated evenly among councilors? If not, what are the factors that explain the variations between them? As a proxy for councilors' predispositions towards clientelism we composed an index based on four questions presented to councilors related to the frequency and place where they are sought by the voters, to what extent they are exposed to voter requests for private benefits and to what extent are they willing to fulfill them through informal means.

A descriptive analysis of the data shows that, for most councilors, some of the necessary, albeit not sufficient, conditions for the configuration of a clientelist relationship are present. Councilors occupy a privileged position vis-à-vis voters (hierarchy and asymmetry regarding access to public goods) and many of them maintain daily contact with the latter (iterated relations) for the submission of requests for private benefits. A large proportion of councilors seek to respond to these requests by means of informal strategies. Therefore, it can be said that a significant share of the councilors interviewed, exhibit the key elements that are present in any definition of clientelism in their daily interactions with voters: hierarchy, personalism, particularism, and informality.

On the other hand, the distribution of the dependent variable suggests that not all councilors are equally prone to engage in clientelism. This reinforces what had already been concluded by other authors who point out that among councilors there are those who engage in other types of relationships with voters: relations more restricted to the confines of the city council, and characterized by the reception of more collective demands and more institutionalized response strategies.

Based on a multilevel analysis, we verified the impact of some individual attributes and some characteristics of the municipalities on the councilors' predispositions towards clientelism. The results showed that predisposition is higher in smaller, poorer municipalities with a less competitive electoral scenario. More specifically, the analysis suggests that low political competition has a positive impact on councilors' predispositions towards clientelism, regardless of poverty levels. Poverty seems to increase the effect of low competition. This suggests, as other studies point out, that municipalities with high levels of poverty and low political competition are more favorable scenarios for clientelism. Although our analyses add one more piece of evidence in favor of a negative relationship between competition and clientelism, our results are only suggestive. We consider that the main contribution of the article is descriptive, which is not small considering the lack of information regarding the daily routine of councilors, and their interactions with voters.

Although in most cases we found associations in the expected direction, some questions remain open. The analysis suggests that less competitive settings, when combined with a largely poor population, are associated with a greater propensity towards clientelism. While this evidence is consistent with some expectations found in literature regarding the relationship between competition, poverty and clientelism, it is still necessary to understand the mechanisms that operate in producing that dynamic. According to Weitz-Shapiro (2012)WEITZ-SHAPIRO, Rebecca (2012), What wins votes: why some politicians opt out. American Journal of Political Science. Vol. 56, Nº 03, pp. 568-583. this happens because the presence of a sizable middle class (which implies reduced levels of poverty) increases the electoral cost of clientelism for politicians. She states that the middle class would be unlikely to support a politician who relies on clientelism. However, that explanation is based on the assumption that the middle class values programmatic policies and spending on public goods, while the poor tend to attach greater marginal utility to small material inducements. Since the author did not present empirical evidence to support these assumptions, we believe that the puzzle persists.

The relationship between competition and clientelism, in this case, may be, in fact, the opposite of what is commonly supposed in the literature. Is it the practice by councilors of delivering private benefits that makes politics less competitive by giving advantages to incumbents who have proven ability to deliver? Alternatively, is it the lower levels of political competition – explained by other variables – that favor the incumbents, making them the only ones who can offer credible promises to voters?

Our approach and the set of evidence we presented allows us to apprehend part of the phenomenon we addressed. One of the limitations lies in the fact that our data report information on only one side of the relationship, the supply side. We do not have information about voter perceptions, the demand side. The proxy used to operationalize clientelism also has limitations, among them, the fact that it does not grasp the aspect of exchange conditionality.

On the other hand, our study offers an important contribution by bringing new empirical evidence on interactions between voters and their representatives at the municipal level, the political sphere that is most tangible and concrete to the average citizens. If we consider that an expressive share of our political leaders begins their carriers as councilors, knowledge about their perceptions, profiles, attitudes, and their relationship patterns with voters and groups is important for a broader understanding of how our political system works.

# References

• AGRESTI, Alan and NATARAJAN, Ranjini (2001), Modeling clustered ordered categorical data: a survey. International Statistical Review Nº 69, pp. 345-371.
• ALMEIDA, Maria Hermínia Tavares and CARNEIRO, Leandro Piquet (2003), Liderança local, democracia e políticas públicas no Brasil. Opinião Pública Vol. 09, Nº 01, pp. 124-147.
• AMES, Barry (1995), Electoral rules, constituency pressures, and pork barrel: bases of voting in the Brazilian Congress. Journal of Politics Vol. 57, Nº 02, pp. 324–343.
• ANDRADE, Régis Stephan de Castro (ed) (1998), Processos decisórios na Câmara dos Vereadores e na Assembleia Legislativa de São Paulo. In: Processo de governo no município e no estado: uma análise a partir de São Paulo. São Paulo: EDUSP/FAPESP. pp. 15-40.
• ARRETCHE, Marta Teresa da Silva (2012), Democracia, federalismo e descentralização no Brasil Rio de Janeiro: Editora FGV: Fiocruz. 232 pp..
• AUYERO, Javier (2001), Poor people's politics: peronist survival networks and the legacy of Evita. Durham: Duke University Press. 272 pp..
• BRUSCO, Valeria; NAZARENO, Marcelo and, STOKES, Susan Carol (2004), Vote buying in Argentina. Latin American Research Review Vol. 39, Nº 02, pp. 66–88.
• CALVO, Ernesto and MURILLO, Maria Victoria (2004), Who delivers? Partisan clients in the Argentine electoral market. American Journal of Political Science Vol. 48, Nº 04, pp. 742–757.
• CHRISTENSEN, Rune Haubo Bojesen (2019), Regression models for ordinal data. R package version 2019.4-25. Available at ˂http://www.cran.r-project.org/package=ordinal/˃ Accessed on May, 07, 2019.
» http://www.cran.r-project.org/package=ordinal/˃
• D'ÁVILA FILHO, Paulo Mesquita; LIMA, Paulo Cesar and JORGE, Vladimyr L. (2014), Comportamento político e produção legislativa: padrões de emissão de indicações pelos vereadores cariocas. In: Eleições, partidos e representação política nos municípios brasileiros Edited by ROCHA, Marta Mendes da and KERBAUY, Maria Teresa Miceli. Juiz de Fora: Editora UFJF. pp. 191-214.
• DE ROZAS, Victoria Ortiz (2017), Personal, particular y político: la producción del bien público en la práctica política cotidiana. Revista de Sociologia Política Vol. 25, Nº 63, pp. 53-77.
• GEDDES, Barbara (1991), A game theoretical model of reform in Latin American democracies. American Political Science Review Vol. 85, Nº 02, pp. 371-392.
• GRILLI, Leonardo and RAMPICHINI, Carla (2009), Multilevel models for the evaluation of educational institutions: a review. In: Statistical methods for the evaluation of educational services and quality of products Edited by MONARI, Paola; BINI, Matilde; PICCOLO, Domenico and, SALMASO, Luigi. Heidelberg: Physica-Verlag Heidelberg. pp. 61-80.
• HARRELL JR., Frank E. (2001), Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. New York: Springer-Verlag. 572 pp..
• HEDEKER, Donald (2008), Multilevel models for ordinal and nominal variables. In: Handbook of multilevel analysis Edited by LEEUW, Jan de and MEIJER, Erik. New York: Springer-Verlag. pp. 239–276.
• HICKEN, Allen (2011), Clientelism. Annual Review of Political Science Vol. 14, pp. 289–310.
• HILGERS, Tina (2011), Clientelism and conceptual stretching: differentiating among concepts and among analytical levels. Theory and Society Vol. 40, Nº 05, pp. 567–588.
• IBGE (2010), Indicadores sociais municipais. Uma análise dos resultados do universo do Censo demográfico 2010. Instituto Brasileiro de Geografia e Estatística. Available at ˂https://biblioteca.ibge.gov.br/visualizacao/livros/liv54598.pdf˃ Accessed on October, 25, 2013.
» https://biblioteca.ibge.gov.br/visualizacao/livros/liv54598.pdf˃
• KERBAUY, Maria Teresa Miceli (2005), As câmaras municipais brasileiras: perfil de carreira e percepção sobre o processo decisório local. Opinião Pública Vol. 11, Nº 02, pp. 337-365.
• KITSCHELT, Herbert and WILKINSON, Steven I. (2007), Patrons, clients, and policies: patterns of democratic accountability and political competition. Cambridge: Cambridge University Press. 377 pp..
• KUSCHNIR, Karina (2000), O cotidiano da política Rio de Janeiro: Relume-Dumará. 162 pp..
• LANDÉ Carl H. (1983), Political clientelism in political studies: retrospect and prospects. International Political Science Review Vol. 04, Nº 04, pp. 435-454.
• LEVITSKY, Steven (2007), From populism to clientelism? The transformation of labor-based party linkages in Latin America. In: Patrons, clients, and policies: patterns of democratic accountability and political competition. Edited by KITSCHEKT, Herbert and WILKINSON, Steven I.. Cambridge: Cambridge University Press. pp. 206-226.
• LOPEZ, Felix G. and ALMEIDA, Acir (2017), Legisladores, captadores e assistencialistas: a representação política no nível local. Revista de Sociologia Política Vol. 25, Nº 62, pp. 157-181.
• LOPEZ, Felix G. (2004), A política cotidiana dos vereadores e as relações entre executivo e legislativo em âmbito municipal: o caso do município de Araruama. Revista de Sociologia Política Nº 22, pp. 153-177.
• MAGALONI, Beatriz; DÍAZ-CAYEROS, Alberto and ESTÉVEZ, Federico (2007), Clientelism and portfolio diversiﬁcation: a model of electoral investment with applications to Mexico. In: Patrons, clients, and policies: patterns of democratic accountability and political competition. Edited by KITSCHEKT, Herbert and WILKINSON, Steven I.. Cambridge: Cambridge University Press. pp. 182-205.
• MAINWARING, Scott (1991), Politicians, parties and electoral systems: Brazil in comparative perspective. Comparative Politcs. Vol. 24, Nº 01, pp. 21-43.
• MARENCO, André and SERNA, Miguel (2007), Por que carreiras políticas na esquerda e na direita não são iguais? Recrutamento legislativo em Brasil, Chile e Uruguai. Revista Brasileira de Ciências Sociais Vol. 22, Nº 64, pp. 93-113.
• NICHTER, Simeon and PERESS, Michael (2017), Request fulfilling: when citizens demand clientelist benefits. Comparative Political Studies Vol. 50, Nº 08, pp. 1086-1117.
• PIATTONI, Simona (ed) (2001), Clientelism, interests, and democratic representation. In: Clientelism, interests, and democratic representation: the European experience in historical and comparative perspective. New York: Cambridge University Press. pp. 193-212.
• PNUD, Programa das Nações Unidas para o Desenvolvimento (2013), Atlas do Desenvolvimento Humano no Brasil. Available at ˂http://www.atlasbrasil.org.br/2013/pt/o_atlas/idhm˃ Accessed on April, 30, 2013.
» http://www.atlasbrasil.org.br/2013/pt/o_atlas/idhm˃
• POWER, Timoty J. (2000), The political right in postauthoritarian Brazil: elites, institutions, and democratization. University Park: The Pennsylvania State University Press. 304 pp..
• R CORE TEAM (2019), R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at URL ˂https://www.R-project.org/˃ Accessed on May, 07, 2019.
» https://www.R-project.org/˃
• RAUDENBUSH, Stephen W. and BRYK, Anthony S. (2002), Hierarchical linear models: applications and data analysis methods. Thousand Oaks: Sage Publications. 512 pp..
• SCOTT, James C. (1977), Political clientelism: a bibliographical essay. In: Friends, followers and factions: a reader in political clientelism. Edited by SCHMIDT, Steffen W.; SCOTT, James C.; LANDÉ, Carl H., and GUASTI, Laura. Berkeley: University of California Press. pp. 483-505.
• SNYDER, Richard (2001), Scaling down: the subnational comparative method. Studies in Comparative International Development Vol. 36, Nº 01, pp. 93–110.
• STOKES, Susan; DUNNING, Thad; NAZARENO, Marcelo and, BRUSCO, Valeria (2013), Brokers, voters and clientelism: the puzzle of distributive politics. New York: Cambridge University Press. 316 pp..
• WEITZ-SHAPIRO, Rebecca (2012), What wins votes: why some politicians opt out. American Journal of Political Science Vol. 56, Nº 03, pp. 568-583.

# Publication Dates

• Publication in this collection
10 Jan 2020
• Date of issue
2019