The third wave of studies on clientelism in Latin America: a critical reading

Abstract The article presents a critical reading of the third wave of studies on clientelism in Latin America produced in Political Science since the 2000s. Through a dialogue with the most influential authors and works on the issue, the objective is to show that despite significant advances, this literature still faces difficulties in defining clientelism. The central argument is that these difficulties are due to the adoption of universal and rigid assumptions about politicians’ and voters’ behavior and the strong normative tone of the analyses, almost always grounded on premises of how democracy should work and how clientelism displaces this ideal. To demonstrate this argument, the concept of clientelism was broken down into four central aspects: the object of the exchanges, the participants, the method used, and the temporality. The analysis of these dimensions is articulated with the literature on concept formation, especially the criteria of differentiation, depth, internal coherence, and theoretical utility. The article concludes by proposing ways to overcome the highlighted problems, among them the re-evaluation of the premises that inform the analysis, the valorization of description, and the investment in multi-method research designs.


Introduction 1
The last two decades have witnessed a renewed interest in the subject of clientelism in different parts of the world.Clientelism is usually understood as part of a broader set of distributive politics that concentrate benefits while dispersing costs (Stokes et al., 2013) and as a relationship in which candidates, parties, and politicians exchange private goods for political support.
Contrary to expectations that the practice would be undermined due to democratization and social-economic development, clientelism has adapted and thrived in different social, economic, cultural, political, and institutional contexts (Kitschelt and Wilkinson, 2007).Hence, one of the central puzzles that the third wave of studies on clientelism seeks to solve concerns why and how clientelism persists in competitive regimes characterized by free, regular, clean, and secret elections (Nichter, 2018).
This article presents a critical reading of recent studies on clientelism in Latin America produced in Political Science, focusing on how literature defines clientelism.My objective is not to review the literature exhaustively; 2 instead, I intend to highlight some advances and shortcomings of this body of work regarding the definition of clientelism, using as examples influential studies on the subject published since the early 2000s.The selection results from my ten-year immersion in the literature on clientelism produced in the last decades, especially books and articles published in English and Spanish.The works selected as examples and illustrations of the arguments shaped the field in this period, providing the main theoretical, conceptual, methodological, and interpretative framework for the investigation of clientelism in comparative politics.
I analyzed four aspects relevant to the comprehension of clientelism: (i) the object of the exchange (what), (ii) the participants in the relationship (who), (iii) the method that the exchange relationships assume (how); and (iv) the temporality of the exchanges (when).In addition, I identify some elements that have limited research efforts in the field and suggest ways to overcome these limitations.
1 I thank Lucas Gelape, Ryan Lloyd and Javier Auyero for their careful reading and comments on earlier versions of this paper.I thank the anonymous reviewers of RBCS for their comments and suggestions that contributed significantly to improving the article.This article is one of the products of the research "Distributivismo, clientelismo e brokers na política local brasileira" which is supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and "Representação, dinâmicas de intermediação de interesses e brokers na política local brasileira" supported by the Fundação de Amparo à Pesquisa de Minas Gerais (Fapemig).
3 Clientelism is a central topic in Brazilian political thought and an important key to interpreting the relationship between the state and society in the country.For studies that address clientelism at different moments in Brazilian political history, see Graham (1997), Leal (1997), Nunes (1997), and Gay (1998).For more recent studies produced in the context of the third wave, see D'Ávila Filho et al. (2014), Lopez and Almeida (2017), Rocha et al. (2019), Hoyler and Marques (2023), among others.
Undoubtedly, this literature has widened the overall comprehension of the workings of clientelism in the region and represents an admirable effort to deepen our knowledge about the topic.However, scholarship still struggles with clientelism's definition, with direct consequences on the choices scholars make to operationalize the concept.It is no accident that fundamental conceptual issues often return to the forefront of debates (Hilgers, 2008;Nichter, 2014;Hicken and Nathan, 2020).
As stated by Hicken and Nathan (2020), scholars often describe practices that do not fit the current definitions of clientelism, especially regarding contingency and iteration, two elements central to almost all conceptualizations.According to the authors, an examination of recent literature gives the impression that real clientelism -"with a clearly contingent, iterated quid pro quo" -is actually very rare (Hicken and Nathan, 2020, p. 280).Moreover, one often observes the difficulty of distinguishing clientelism from related, yet distinct, practices and types of exchange.
In this paper, I propose that a disaggregated analysis of some elements central to studies on the subject can help understand and, therefore, overcome some difficulties in defining and measuring clientelism.I argue that though one can observe an increased tendency to define clientelism as a method of distribution, a closer look reveals some difficulties in dissociating the mode (how) from other aspects related to the object (what) and the participants (who).Additionally, with few exceptions, the literature seems not to take seriously the dimension of time (when), posing the same assumptions and questions to exchanges that take place during elections and those that go beyond the electoral period.
I propose that the adoption of rigid assumptions about the behavior of politicians and voters by a considerable number of scholars explains part of the difficulties they face in defining clientelism.These assumptions have hindered the effective integration between qualitative and quantitative approaches despite the admirable efforts of combining different methods and research techniques.Signs of this problem can be found in the persistence of ambiguities surrounding the concept, thereby revealing the limits of existing theories to encompass the diversity of Latin American politics.
The article does not offer a systematic analysis of the concept of clientelism based on the literature on concept formation, with its rules and criteria, although it dialogues with it at some points.Since the influential work of Sartori (1970Sartori ( , 1984)), the scholarship on concept formation has helped us face more systematically the difficult task of defining complex and controversial concepts in the Social Sciences.Such an analysis would certainly be of great use in illuminating and deepening some of the issues raised in this article.In this piece, however, I have chosen another analytical strategy focused on the four most important attributes in most definitions of clientelism.As we shall see, in different ways, they crosscut the discussion on concept formation, especially concerning the criteria of differentiation, depth, internal coherence, and theoretical utility (Collier and Mahon, 1993;Gerring, 1999;Goertz, 2005).
I agree with Gerring (1999) that semantic confusion and linguistic disputes are, to some extent, inevitable in the social sciences, as are contradictory and overlapping definitions and imprecise operationalization.However, in the same spirit as the author, I argue that when these inaccuracies accumulate, they begin to work against the advancement of knowledge.They begin to undermine research efforts simply because we do not pay enough attention to the flaws in the concepts we use and the origin and nature of those flaws.This article acknowledges the advances of the third wave of studies on clientelism, but its central aim is to draw attention to some persistent difficulties that continue to generate contradictions and obstacles in the field from a different angle, by directing our gaze to some of the assumptions that inform research on the topic.My goal is not to exhaust the subject but to stimulate discussion about how we can address these shortcomings by dealing more reflectively and transparently with our choices regarding conceptualization, the biases they carry, their implications for the research process and the conclusions we reach.Especially when dealing with topics and objects surrounded by multiple normative concerns, it is necessary to redouble the attention and vigilance regarding our assumptions.

The third wave of studies on clientelism
The third wave of studies on clientelism emerged in the 2000s (Stokes, 2009).While many studies approach clientelism in Africa, Europe, and Asia, I limit my analyses to Political Science studies focused on Latin America.Some aspects distinguish this third wave from the previous ones: There is a clear inclination to approach clientelism as a market transaction, a voluntary relationship with mutual benefits (Piattoni, 2001;Stokes, 2005).According to scholars, this approach would be better suited to the current context of competitive regimes characterized by secret elections.In this new scenario, clientelism does not work as it did in the past (Hilgers, 2011).The fear of being excluded from the list of beneficiaries keeps voters' commitment and sustains exchange relationships (Brusco et al., 2004), rather than feelings of obligation and affection, bonds of reciprocity, or fear of physical violence.Additionally, the increase in electoral competition and the implementation of electoral reforms driven to secure vote integrity have given voters more power to negotiate with bosses and more options to exit the relationship (Nichter, 2018).
In this utilitarian approach, the central premises of Rational Choice Theory -strategic behavior of individuals who act rationally pursuing their own interests -drive, explicitly or implicitly, a significant part of research on the topic.Therefore, one sees broad mobilization of game theory jargon and formal models with predictive objectives (Medina andStokes, 2002, 2007;Dunning and Stokes, 2007;Stokes, 2005;Calvo and Murillo, 2004;Keefer and Vlaicu, 2008;Robinson and Verdier, 2013;Nichter, 2018).
A related characteristic is the tendency to address participants' behavior by drawing on universal assumptions.Often, the preferences of the mentioned actors are taken as given and fixed.It is supposed that voters want to maximize their access to private benefits and that candidates and parties are driven to maximize electoral support.In line with this tendency is the shift of focus from structural variables to the micro-foundations of clientelism (Stokes et al., 2013).In many studies, clientelism is investigated based on abstract, hypothetical situations in which the behavior of individuals and the dynamics of relationships are inferred from a set of incentives modelled by analysts. 4n addition to a strong empirical and comparative orientation, there is a tendency to adopt minimal definitions of clientelism, draining the concept of the elements that used to characterize it in previous waves (Hilgers, 2008;2011).This seems to be a strategy to develop more accurate ways to conceptualize and measure clientelism and distinguish it from other related or similar practices.
A similarity between recent and previous investigations is the normative debate around clientelism.Many scholars list the damages clientelism inflicts on democracy and economic performance, claiming that it violates the principles of political equality and freedom of choice (Piattoni, 2001;Stokes, 2009;Stokes et al., 2013) and undermines the vote as a mechanism of vocal preferences and as a means by which representatives can be held to account (Stokes, 2005;Gans-Morse et al., 2014).Other scholars claim that clientelism hurdles economic growth since it encourages politicians to keep citizens poor and dependent (Stokes, 2009) and leads to inefficiencies in the supply of public goods (Robinson and Verdier, 2013).Clientelism is also associated with bad governance, bureaucracy politicization, less transparency, and more corruption or perception of corruption (Hicken, 2011;Weitz-Shapiro, 2012).
Despite the pervasiveness of these judgments, only a few studies in the third wave have empirically examined the causal links between clientelism and its presumable consequences.Rather than offering unequivocal answers about the consequences of clientelism, their findings ended up raising new questions (Mignozzettiy and Cepaluni, 2017;Brun and Diamond, 2014).Moreover, there is evidence that clientelism can thrive in parallel with more autonomous forms of collective organization and political participation (Kitschelt and Wilkinson, 2007;Holland and Palmer-Rubin, 2015).For some authors, it should be seen as a legitimate way for citizens to access public goods and services and citizenship rights (Hilgers, 2012;Borges, 2019).
Finally, it is essential to mention that not all studies produced in Political Science in the last two decades on clientelism in Latin America share all the characteristics highlighted above.Some scholars deviate by addressing clientelism differently, as with Levitsky's (2003) research on Argentina, Hilgers' (2008) studies on Mexico, and Borges' (2019) investigation on Brazil.Moreover, there is always something arbitrary about the cutoffs established to analyze the scholarship on a given topic.While there is no doubt about a revival around clientelism in recent decades, it is difficult to establish with precision when the third wave begins and, if it is the case, when it ends.

Conceptualizing clientelism
Although clientelism is a concept often mobilized in Political Science and in Latin American studies, it is usually employed vaguely, suffering from ambiguities and lack of a reliable foundation for measurement.Frequently it is used as synonymous with patronage and pork barrel; at other times, its use goes beyond the boundaries of licit actions and practices, when clientelism is mixed with vote buying and corruption (Hilgers, 2011;Nichter, 2014).Especially in Political Science, a field in which the direct observation typical of ethnographic studies is not very common, the problem of conceptualizing clientelism and defining exactly how it materializes is by no means trivial.
Recent studies feature the general idea that clientelism is a relationship of conditional exchange, that is, that receiving a benefit is conditional to political support in the form of the vote, involvement in electoral campaigns, attendance in rallies, advertisement, and the like (Weitz-Shapiro, 2012;Stokes et al., 2013;Gans-Morse et al., 2014;Nichter, 2018).Additionally, clientelism is considered an asymmetric relationship between individuals with unequal access to goods and services either in terms of their socioeconomic status or their position in the state structure.It is also labeled as particularistic because the benefits target certain individuals and groups.
Although many studies consider clientelism as an iterate relationship, sustained over time, rather than a one-shot exchange (Stokes, 2005), few scholars really explore the dimension of time.In the context of competitive democracies, studies in the third wave also tend to approach clientelism as a voluntary relationship, although they recognize that the costs of exiting the relationship vary from context to context (Medina and Stokes, 2002;Kitschelt, 2000).
In this section, I disaggregate the concept of clientelism, considering four different sets of questions the literature usually emphasizes: Which aspect of the relationship should be highlighted to define clientelism?What is exchanged and/or distributed?Who participates?How does the exchange happen?Or when does it happen?
These four dimensions, alone or in interaction, allow discussing different aspects related to concept formation, including the most appropriate structure for delimitation (dichotomous or continuous, with necessary and/or sufficient conditions), the importance of theory in concept formation (either as a starting or ending point), the relationship between conceptualization and operationalization, and the crucial task of specifying which attributes are central and which are secondary.
According to Goertz (2005), concepts should allow us to distinguish processes, events, and phenomena in the social world.While there is some agreement that good concepts should be robust, not ambiguous, and provide a secure basis for operationalization and theory building, there is less agreement on how to achieve this goal.If the problems pointed out by Sartori more than two decades ago -conceptual travelling and stretching -still deserve our attention, the solution offered by the author based on his "ladder of generality" model (Collier and Mahon, 1993, p. 843) is today the target of criticism.It is considered by some authors to be excessively rigid and demanding, and therefore inadequate to deal with the dynamism of concept formation in the Social Sciences (Collier and Mahon, 1993;Gerring, 1999).
The disaggregation of the concept of clientelism, as addressed in the most influential studies of the third wave, will allow us to address some of the aspects mentioned above while making clear how the construction and reformulation of concepts take the form of a permanent game of trade-offs and adjustments (Gerring, 1999).

What? The object of exchanges
Accounting for the object of exchanges, there seems to exist a consensus that clientelism involves the distribution of a specific type of good: the so-called excludable benefits, which, differently from public goods, allows one to discriminate between those who should benefit from those who should not.Studies on the topic display a long list of goods requested by voters and offered by politicians, including food, clothing, medicine, mattresses, sheets, coffins, school materials, water, female sterilization, appliances, bricks, alcohol, marijuana, and cash (Brusco et al., 2004;Caetano and Potter, 2004).
Among the benefits, employment, in the form of a position in public administration or any other paid work, would be one of the most valued (Calvo and Murillo, 2004;Zarazaga, 2014;Nichter, 2018;Oliveros, 2021).To designate the distribution of jobs in public administration to supporters, some scholars prefer the term patronage (Medina and Stokes, 2007;Calvo and Murillo, 2004).Stokes et al. (2013) see patronage as a subtype of clientelism, a strategy through which elected politicians allocate public positions to party members (Stokes et al., 2013, p. 7).According to this definition, there would be a combination of two criteria: what (public employment) and who (party members). 5he term clientelism is also mobilized to designate the efforts made by politicians and candidates to fulfill voters' private requests related to education, health assistance, water supply (Nichter, 2018), and all kinds of favors provided by politicians to voters (Oliveros, 2016).It is also used to designate politicians' assistance to their constituents and local brokers to help them solve problems and obtain resources and favorable decisions from higher-level state agencies (Oliveros, 2016;Rocha et al., 2019;Rocha and Gelape, 2023).
If the object of the exchange does not seem enough to distinguish clientelism from other types of distribution, it has substantial repercussions for the study of the topic.Kiewiet de Jonge (2015), for instance, shows that the nature of the benefit distributed influences a voter's probability of admitting having received a gift from a candidate.The evidence he presents suggests that, from voters' perspective, it is socially more acceptable to receive certain kinds of goods -food and medicine, for instance -than others like money and durable goods.That is why clientelism underreporting rates are "smallest in countries in which the gifts consist largely of innocuous campaign materials and items such as clothing and food" (Kiewiet de Jonge, 2015, p. 2).
The type of benefit can also affect the strength of bonds between clients and patrons.There is evidence that candidates and voters consider the value of the benefit to adjust the extension of the support and obligations it entails (Zarazaga, 2014).Along these lines, public sector jobs would likely be the benefit most capable of producing loyalty (Zarazaga, 2014) not only because voters value it more, but because continued access to them depends on the victory of the candidate or party (Hicken and Nathan, 2020;Oliveros, 2021).
A definition of clientelism centered on the object of the exchanges maximizes the criterion of familiarity because, to some extent, it echoes established usage in everyday life and within a community of experts.In the media and among politicians in several Latin American countries, the offering and delivery of handouts are commonly recognizedbelittled and condemned -as clientelism, although it is also common for other terms such as vote buying to be employed to designate the practice.Moreover, it is a definition that facilitates direct observation.Outside the electoral period or in contexts where vote buying is not criminalized, the delivery of material benefits occurs in a very visible way, as highlighted by the literature (Brusco et al., 2004).Still, despite maximizing familiarity, favoring observation and empirical verification, and appearing in virtually all influential thirdwave definitions of clientelism, the dimension related to the object of exchanges ("what") does not appear as a central attribute or sufficient condition in definitions of clientelism.

How? Clientelism as a method of distribution
When the emphasis is on what is exchanged, the main underlying issue is the distinction between programmatic and particularistic discretionary policies.According to the literature of mainstream Political Science, clientelism would undoubtedly be among the second (Stokes et al., 2013).Benefits are private and target specific individuals and groups.This is not to say that every particularistic and targeted policy is clientelist.Particularistic policies can also be programmatic when they are consistent with the programmatic objectives of parties and operate based on clear eligibility criteria, with little or no room for politicians to influence, from discretionary criteria, the form of distribution (Stokes et al., 2013).Consequently, the decisive issue for distinguishing clientelism from other forms of distribution would not be the object of exchanges (what) or the participants and beneficiaries (who) but rather "how goods are delivered" (Calvo and Murillo, 2013, p. 875).
From this perspective, clientelism is seen as a specific method of distribution and electoral mobilization distinguished by the discretionary character of the exchanges and by conditionality, that is, by the fact that voters' political support is a condition for access to the benefit (Weitz-Shapiro, 2012;Stokes et al., 2013;Medina and Stokes, 2007;Magaloni et al., 2007, p. 184).Therefore, the most precise way to distinguish clientelism from other forms of distribution would be to focus on the discretion of those who deliver the benefits and verify whether political support is a condition to access the benefit.By highlighting the element of conditionality, scholars shed light on monitoring and brokers.The conditional delivery of private benefits would depend on politicians, parties, and candidates being able to distinguish supporters from non-supporters or make voters believe that they can do it and will sanction non-supporters accordingly.
The literature usually defines brokers as people involved in face-to-face relationships in their communities who know voters' needs and political preferences.They act to fulfill requests and solve voters' problems and, in return, ask for political support for a party or a candidate (Auyero, 2001;Stokes, 2005;Stokes et al., 2013;Holland and Palmer-Rubin, 2015;Zarazaga, 2014;Finan and Schechter, 2012;Gingerich and Medina, 2013;Muñoz, 2014;Novaes, 2018;Eduardo, 2016).Although scholars often portray brokers as key actors through which political machines monitor voter behavior, there is no consensus on this.Zarazaga (2014) and Muñoz (2014), in their studies on Argentina and Peru, respectively, are skeptical about the ability of brokers to monitor voter behavior in order to ensure the return expected by politicians.In a recent work, Hicken and Nathan (2020) show that only a minority of studies present evidence of voter monitoring at the individual level.The unwillingness and low ability of politicians to monitor voters' behavior also casts doubt on the conditional nature of the distribution.
The way most studies measure the extent of clientelism -by employing a question in a questionnaire about giving, distributing, or receiving offers of a gift in the last election -fails to grasp the dimension of discretion and conditionality.6Moreover, using indirect forms of measurement with questions involving hypothetical situations may be adequate to reduce underreporting, but it has limits for capturing people's experience with clientelism.Thus, distinguishing clientelism as a method of distribution is only an apparent solution to conceptual problems.This alternative actually multiplies measurement problems: it would be necessary to show that there is intent and willingness on both sides, ability to distribute on a discretionary basis, and ability and willingness to sanction partners appropriately.For the most part, the literature only offers -if at all -evidence for one of these elements and makes assumptions about the others.For this reason, data obtained in surveys should be considered approximations that do not allow us to confidently establish the conditional character of the exchanges, which, among other things, opens space for questioning the effectiveness of clientelism as a form of electoral mobilization.
Moreover, the literature is full of mixed and contrasting evidence suggesting that very different motivations drive clientelist relations in different contexts.Authors who stress the elements of conditionality and discretion as keys to distinguishing clientelism face the challenge of proving the counterfactual -that is, demonstrating that the benefit would not have been accessed without the counterpart of political support or that the voter would not have voted for a candidate if not for clientelist reasons.Since surveys do not directly ask voters whether they would have preferred to vote for another candidate, scholars need to rely on assumptions about voter behavior in order to close the cycle.
An alternative strategy to address clientelism that seeks to circumvent surveys' limitations is a study focused on how public authorities manage the distribution of benefits from social programs.The objective is to identify whether distribution follows clear and impersonal eligibility criteria or whether there is room for discretionary action and political manipulation.These works, however, also tend to rely on widely accepted assumptions to strengthen the value of the evidence.Weitz-Shapiro (2012), in a study on clientelism in Argentina, shows evidence of discretionary action by identifying the mayor's influence in drawing up the list of beneficiaries of a food distribution program.However, she cannot grasp the dimension of conditionality and has to assume -based on utilitarian assumptions -that if mayors are in a position to distribute private benefits based on political and partisan criteria, they will do so.The premise that politicization is synonymous with clientelism and bureaucratic insulation is synonymous with programmatic and impersonal distribution plays a central role here.
A similar example from Magaloni, Diaz-Cayeros, and Estevez ( 2007) is a study about Pronasol, a program to fight poverty launched in Mexico in 1989.The authors conclude that within Pronasol, clientelist projects accounted for 25% in 1989 and 35% in 1994.However, the only criterion used to label the projects as clientelist was their targeted nature (in this case, individuals and organized groups of producers, indigenous people, and women).Although they adopt a definition of clientelism based on the discretionary and conditional nature of the distribution, ultimately, the authors measure clientelism based on the type and scope of benefits, not on "how" they were distributed.Nevertheless, the fact that they are targeted benefits is not the same as stating that their transfer occurred in a discretionary manner.
This example suggests that it is not possible to conclusively detach the object of exchanges from the method; after all, not all types of benefits lend themselves equally to a conditional and discretionary distribution.These are better combined with so-called excludable benefits and less with more universal public policies.
By focusing on the "how" dimension to define clientelism, part of the third-wave literature opts for the more demanding concept formation framework that provides a necessary condition in the logic of "if and only if" (Collier and Mahon, 1993;Goertz, 2005).The authors who adhere to this strategy seem to be seeking to maximize the differentiation criterion, that is, to determine the boundaries of a concept and the attributes that distinguish it from others, in the game of trade-offs discussed by Gerring (1999).This concern is understandable given that clientelism competes for space with neighboring categories such as pork barrel and vote buying.They also seem to seek a parsimonious definition since, as noted above, this literature has tended to drain clientelism of other elements that used to be in previous definitions.The problem is that even though conceptualization and operationalization are different tasks, they cannot be thought of separately.A differentiated concept must also be operationalizable (Gerring, 1999, p. 379).What happens, however, when the attributes considered most important for differentiation are also the most difficult to observe and the least susceptible to empirical verification?This concept will be less useful and we will have a category without a referent (Gerring, 1999, p. 383).
Hicken and Nathan (2020) drew attention to this very problem by arguing that clientelism cannot be defined through the criterion of conditionality because one would either have to assume that it is much less present than one imagines or that one is not equipped to produce evidence of the quid pro quo.The scarce support for the monitoring thesis at the individual level reinforces this point.Therefore, the question that emerges is why candidates and parties engage in the distribution of private goods if, for the most part, they cannot monitor voter behavior and face a high risk of not getting the expected return.
Different answers to this question seem to prove most productive for the study of clientelism.In these studies, conditionality ceases to be a distinctive element.Muñoz (2014), for example, addresses clientelism as an indirect strategy used by candidates to attract supporters and voters to campaign events and exhibit the viability of a candidate.In this sense, target distribution and vote buying play an informational and heuristic role, central for voters in environments where they lack informational shortcuts (due, for example, to the high number of parties and low rates of party identification) (Lloyd, 2016).Borges (2019), in a recent study on clientelism in the backlands of Bahia, offers another possible answer for why candidates and parties engage in the distribution of private goods with no guarantee of return.She shows that given the voters' belief that "only those who have money can win elections" (p.3), politicians distribute resources not based on a logic of economic efficiency.They distribute indiscriminately (p. 2), following a logic of largesse to embody notions of strength (p.30).Auyero (2001, p. 14), on the other hand, addresses these relationships as "ongoing informal problem-solving networks", crucial for ensuring poor people's material survival.This perspective does not assume the existence of an exchange of favors for votes as a premise.In addition, it questions the supposed efficiency of clientelism as a strategy to win elections, putting the cultural representations driving political behavior at the center of its concerns.
In the eyes of people involved in these relationships, they are seen as "politics as usual" and considered legitimate forms of access to goods, services, and citizenship rights.From this perspective, it is impossible to state beforehand the forms the relationships will take and their consequences.They can assume more or less democratic and horizontal forms and imply exploitative or emancipatory dynamics.They can foster apathy or merge with autonomous forms of collective organization or contentious politics.Along these lines, the crucial thing would be to analyze the context in which these relationships occur.This theoretical framework seems more appropriate to accommodate the empirical evidence gathered in Political Science studies on clientelism in Latin America.
When?The timing of exchanges A somewhat neglected aspect in studies on clientelism concerns when, that is, the temporality of exchanges.The literature often treats clientelism and vote buying as interchangeable terms, measuring them based on a question about having received a gift or a gift offer in the last election.After all, does it matter when a voter receives a private good or assistance from a politician?For the most part, studies on the subject either focus on the campaign period or the period between elections.Studies that focus on the electoral period and campaign events and strategies, such as those by Muñoz (2014) and Szwarcberg (2012b), contribute to illuminating the distinctive aspects of exchanges that take place in this specific period with new and interesting insights.However, taken in isolation, they end up having little to say about what happens between elections.
Simeon Nichter (2018) proposes a distinction between electoral clientelism, characterized by more episodic exchanges concentrated in the campaign period when voters receive benefits before voting and which could therefore be treated as synonymous with buying votes, and relational clientelism.The latter is used to designate the conditional exchanges of benefits for political support beyond the period of electoral campaigns.In his study on Brazil, Nichter (2018) proposes a link between the two moments with evidence that the benefits received by voters during campaigns are often responses to requests they presented to politicians before elections.
It is reasonable to expect that relations between voters and politicians will take different forms depending on the moment.During the campaigns, what financial and organizational resources can candidates count on?What is the potential payoff from engaging in vote buying?What are the risks?Once in office, in their daily relationship with citizens, freed from the pressures of the electoral process, politicians have to deal with other questions.These concern the amount of time, energy, and resources they are willing to put into delivering private benefits, fulfilling voters' demands, and building a lasting relationship with them.
The same difference is manifested by voters when deciding to accept a benefit in exchange for the vote during the electoral campaign, and to ask or accept the help of a representative two years after the last or before the next election.This difference would be even more striking in the presence of legislation that criminalizes vote buying.Accordingly, clientelism and vote buying can only be treated as synonyms at the expense of too much simplification.
Nichter's distinction between electoral clientelism and relational clientelism seeks to overcome the hitherto common conceptual misunderstandings between clientelism and vote buying.However, it creates new difficulties, starting with the term "relational" to designate clientelism based on long-term relationships.By definition, it follows that non-relational clientelism can exist, which for some scholars on the subject may sound absurd.Moreover, the subtypes proposed by the author end up sounding too artificial, like a facile solution to a more complex issue (Dosek, 2023).
Perhaps a more promising way to deal with the issue is to operate with the categories of clientelism and vote buying understanding that the former is based on long-lasting and sustained relationships between politicians, parties, and voters.In other words, it is always relational and goes far beyond elections.On the other hand, vote buying can be a culmination of clientelism, a phenomenon associated with it, but restricted to a specific point in time (Dosek, 2023), or a phenomenon detached from clientelism (Muñoz, 2014).More important than crafting new terms is to deepen our understanding of the mechanisms underlying both clientelism and vote buying and the different ways in which they can (or cannot) merge.

Who? The participants of the exchanges
No third-wave study of clientelism is explicitly based on the socioeconomic status of the clients to define what is and what is not clientelism.Despite this, it is very common for clientelism to be associated with the poor and for poverty to appear as the main factor that predisposes individuals to clientelistic exchanges (Brusco et al., 2004;Magaloni et al., 2007;Weitz-Shapiro, 2012;Stokes et al., 2013;Nichter, 2018).
Stokes even states that "the affinity between poverty (inequality) and clientelism is settled fact" (Stokes, 2009, p. 623) and that "it is impossible to survey the qualitative literature on political clientelism without concluding that it is a feature disproportionately of poor countries" (p.617).Weitz-Shapiro (2012, p. 3), in a study on clientelism in Argentina, does not hesitate to refer to the poor and clients interchangeably, claiming that this equivalence "is a fairly good approximation of reality in the Argentine case".
The third-wave studies of clientelism in Latin America provide a wealth of empirical evidence that the poor are more exposed to clientelism and that clientelism is more widespread in developing and poor countries, where a considerable proportion of the population lives in poverty. 7The most accepted explanation is that the poor are more prone to clientelist exchanges because of the greater marginal utility they assign to private benefits (Dixit and Londregan 1996;Stokes et al., 2013).This means that by focusing on the poor, politicians and governments facing budget constraints can invest less with greater certainty of electoral returns.The very nature of the distributed benefits (what), generally of low value, would be further evidence that the poor are the priority targets of clientelist strategies.Other scholars argue that the poor, as they have more urgent needs, would be more immediate in their electoral calculations and more risk-averse, preferring a private benefit now to an uncertain programmatic benefit in the long run (Desposato, 2006, p. 59;Stokes et al., 2013;Kitschelt and Wilkinson, 2007).
The conclusion that poverty is a strong predictor of clientelism is associated with the postulate that the expansion of income levels and income distribution would be crucial factors in mitigating vote buying.For some authors, these variables -along with institutional changes aimed at curbing the practice -would explain some national experiences of transition from a clientelist policy to a type of unconditional or programmatic distribution (Stokes et al., 2013).The argument here echoes very directly the theories of modernization.The basic idea is that economic development and the transformations it entails would reduce the number of voters willing to engage in clientelist exchanges.Consequently, one would expect an expansion in the proportion of issue-oriented voters (presumably, the middle class).Socioeconomic development would also contribute to undermining clientelism since the concentration of population in large cities increases the costs of maintaining dense political machines and brokers with penetration within the territory.
A full appreciation of this argument would take more space than is available here.However, it is important to highlight some aspects which urge greater care when associating clientelism and poverty.Firstly, even the most enthusiastic proponents of this thesis recognize there is still no clarity about causal mechanisms.Stokes et al. (2013), for example, present evidence in favor of the hypothesis of a greater marginal utility for clientelism among the poor, but do not find the same support for the hypothesis that the poor are more riskaverse and would tend to perceive programmatic policies as more risky.
In addition to being unclear about the causal mechanisms, mixed and contrary evidence is not always duly considered.Lloyd (2016) and Borges and Lloyd (2018) present evidence that the increase in income does not necessarily result in a reduction of clientelist vote if voters do not have at their disposal alternative heuristics -such as high party identification rates -to guide their vote.Nichter (2018) likewise shows that economic development and poverty reduction are not enough to undermine clientelism if vulnerabilities and insecurities persist.Holland and Palmer-Rubin (2015), based on data from the AmericasBarometer, show that, compared with socioeconomic status, membership in associations is a more consistent predictor of exposure to clientelism than income in Latin American countries (Holland andPalmer-Rubin, 2015, p. 1189).Eduardo (2016), in an experimental survey with citizens of municipalities in Brazil, also found no support for the hypothesis that poorer voters are more sensitive to individual incentives compared to those with higher incomes.Borges' study (2019) in the backlands of Bahia provides evidence that politicians distribute benefits indiscriminately and that the role of client is not restricted to poor voters.
To measure the socioeconomic status of citizens and voters, researchers usually use poverty indicators and secondary data provided by government agencies that may consider income exclusively or include information on unsatisfied basic needs (Magaloni et al., 2007;Weitz-Shapiro, 2012).In the case of surveys, poverty is operationalized by answering a question in the questionnaire. 8There is evidence, however, that the parameters estimated using this technique are impacted by the "social desirability bias" that occurs when respondents "fear expressing socially undesirable attitudes or admitting illegal or stigmatized behavior" (Gonzalez-Ocantos et al., 2011, p. 204).Furthermore, these biases are not randomly distributed: there are significant differences in the probability of a respondent reporting vote buying by sex, age, education, income, and place of residence.With regard to income, Gonzalez-Ocantos et al. (2011, p. 215) claim that "poorer voters may be more inclined to regard vote buying as 'politics as usual', as an inherent part of their social and political problem-solving policies" and "less inclined to lie about having received gifts or favors during campaigns".
A similar conclusion is presented by Kiewiet de Jonge (2015).According to the author, as scholars tend to exclude cases of non-response in the question about income from their analysis, and as respondents with higher income tend, more often, not to answer this question, many studies may be generating biased estimates with regard to the relationship between poverty and vote buying (Kiewiet de Jonge, 2015, p. 17).Furthermore, the author argues that if the stigmatization of clientelism is stronger among the more educated, this does not necessarily express itself in an aversion to the practice and a refusal to participate in it, but only in the refusal to admit that they practice it (Kiewiet de Jonge, 2015, p. 17).
Other criticisms and objections can be made to how the literature treats the relationship between clientelism and poverty.In this article, my primary interest is to discuss its implications for the very conceptualization of clientelism.My argument is that if not for certain assumptions about the political and electoral behavior of the poor and for entrenched and widespread prejudices disseminated in the media, among citizens, and (why not?) among academics, the empirical evidence of the association between clientelism and poverty would not sound so persuasive.The widely accepted idea that clientelism is a phenomenon of poor people leads the vast majority, if not all, researchers interested in conducting qualitative research to places where poor people live.Thus, we must continue to rely on surveys to investigate the association between clientelism and socioeconomic status, albeit imperfectly due to underreporting and biases.
Because of those assumptions, the greater part of the literature is disinclined to explore mixed evidence or rival hypotheses that challenge the association between clientelism and poverty.To what extent can public statements of support from poor voters be seen as activism or free publicity?When voting for a candidate who has offered or distributed private goods, are poor voters acting against their real preferences?The literature does not explore these issues in depth, even when it states that voters' choices are multifaceted (Nichter, 2018, p. 79).Therefore, there are reasons to believe that, in many studies on the subject, who matters more than scholars are willing to admit.

Discussion
In this section, I highlight the elements I believe are closely associated with the difficulties political scientists face in defining clientelism and which have limited their efforts in the context of the third wave of Latin American studies.The first point concerns an element highlighted by Auyero twenty years ago, i.e., the fact that political clientelism "has been one of the strongest and most recurrent images in the study of the political practices of the poor -urban and rural alike -in Latin America" (Auyero, 2001, p. 20).The established narrative of clientelism as one of the main pathologies of democracies in the region cautions scholars to be extra careful in the evaluation of empirical evidence as well as to not turn into premises or proven facts what should be research questions.While most social research is guided by explicit or implicit premises, this is even more striking in studies on clientelism influenced by rational choice theory (Borges, 2019).One has to question the value and contribution of studies in the field of using formal models for predictive purposes, since they need to anchor themselves in rigid and universal assumptions about the behavior of politicians and voters and resort to an oversimplification of complex and quite context-dependent processes and dynamics.
Nowadays, the influence of rational choice theory and formal models in the study of clientelism is not as strong as it used to be.This can be seen in the growing number of scholars trying to broaden their outlook to go beyond material motivations and incorporate symbolic issues -norms, feelings, emotions, and beliefs -along with a demand for more qualitative studies.If one considers that the economic approach to clientelism is one of the most distinctive features of the third wave, perhaps it is possible to speak of a transition or a new trend in this subfield.
As studies on clientelism in Latin America build up, it becomes increasingly clear that despite many characteristics shared by countries in the region, there is enormous diversity in how clientelism operates in each context.Also evident are the limits of studies that -not always explicitly -seek to present theories and general propositions for understanding clientelism, its determinants, and consequences.I am not arguing against the effort to build comprehensive theories, but rather that this effort should not be anchored in rigid assumptions about the behavior of voters and politicians, nor in normative assumptions about how democracy works and should work.As Collier and Mahon (1993, p. 845) argue, the controversies surrounding the process of concept formation is "the perpetual quest for generalization", i.e., the difficulties that comparativists face when trying to apply their concepts and test their hypotheses and models to new contexts.The path suggested by Sartori (1970Sartori ( , 1984) ) five decades ago is still valid: it is a matter of finding the best balance between incorporating context into analyses without abandoning broad comparison (Collier and Mahon, 1993).My argument is that the adoption of rigid assumptions, typical of analyses inspired by rational choice theory, takes us away from this balance and reduces the potential of our comparative efforts.The challenge lies in advancing the theoretical building embedded in comparative analyses and rich empirical data, incorporating actors, arenas, relations, and processes that operate at different levels and intersect, overlap, contradict, and complement each other.
In seeking to maximize differentiation and parsimony in delimiting clientelism, the thirdwave literature appears to have lost significantly in internal coherence and depth, two other central criteria for evaluating a good concept (Gerring, 1999).There seems to be internal coherence between the "what" and "who" dimensions.It is assumed that poor voters are the ones who most respond to this type of mobilization and distribution strategy, and therefore the benefits are of low value.Despite their low value, they have great marginal utility for voters since they match their unsatisfied basic needs.The reasoning could be different: due to budget constraints, clientelism is only profitable for candidates if they can mobilize a large number of voters.And this goal can only be achieved with low-value benefits (Brusco et al., 2004;Magaloni et al., 2007;Weitz-Shapiro, 2012;Stokes et al., 2013).
However, the other two dimensions -"when" and "how" -do not accommodate as consistently and unambiguously to the reasoning.According to Gerring (1999, p. 373), internal coherence, one of the eight criteria guiding concept formation, concerns how well "the attributes that define the concept, as well as the characteristics that characterize the phenomena in question, 'belong' to one another".As I have argued, the evidence for conditionality and monitoring is flimsy, and the dimension of temporality is under-theorized.
The loss in depth, in turn, seems to be associated with the choice for the more demanding ("if and only if") structure of concept delimitation and the under-theorization of sufficient conditions.An exhaustive and systematic survey of definitions of clientelism in the third-wave literature would reveal other attributes, less central to the definition, such as the asymmetric and voluntary character of the exchanges.It would also reveal different combinations of attributes in the conformation of the concepts.In emphasizing this point, I am not arguing for the adoption of a single definition.My goal is, once again, to show how certain choices regarding conceptualization -not always transparent and clearly explained -seem to work against the advancement of knowledge in the field.
In search of a deeper understanding of the mechanisms driving clientelist relations, political scientists have increased the use of multi-method research attempting to integrate quantitative and qualitative approaches and techniques.These efforts, however, tend to be limited if used only to confirm findings obtained through one method, in the form of simple data triangulation, or to adjudicate between rival hypotheses, without ever leading to the questioning and problematization of premises.Political scientists interested in the topic of clientelism should use multi-method research in the way proposed by Seawright, as an alternative to "test assumptions that are generally untested in single-method research, thereby transforming key issues of descriptive and causal inference from matters of speculative assertion into points of empirical debate" (Seawright 2016, p. 1).
Another notable constraint of this scholarship concerns the adoption and reproduction of dichotomies -clientelist versus programmatic politics, particularism versus universalism, and atomization and acquiescence versus mobilization -that find little support in empirical findings and function as a kind of straitjacket.Many scholars claim to be aware that politicians, parties, and voters rarely face binary choices between extremes, but few actually explore the complexity of these choices.
The adoption of rigid assumptions frequently conditions the way evidence is interpreted, leaving little room to truly explore the effect of contextual factors, as well as to address rival hypotheses, mixed evidence, and unexpected results.It is the case with the instrumentalist framework underlying much of the literature in this wave.As stated by Lawson and Greene (2014, p. 70), this framework does not accommodate many empirical findings regarding how clientelism works.The evidence which does not fit in it is treated as isolated exceptions or aberrations (p.73).
Finally, it is important to highlight an element that may help to understand the aforementioned obstacles concerning a broader feature of contemporary Political Science that is not limited to studies on clientelism but which finds a remarkable expression in them.I am referring to the importance of causal questions and studies engaged in causal or predictive questions, vis-à-vis the lower status of those devoted to description.According to Gerring (2012), description in Political Science was downgraded to an inferior status, admitted and justified only as a means or a necessary step to achieve the end, which is the explanation.One consequence is that the analyst focuses only on the elements that presumably play a role in causation, with everything else being dismissed or undervalued.
A similar question arises in debates about concept formation and the relationship between conceptualization and theorization.For Goertz (2005), indeed, the central attributes of a concept are those that have "causal powers" when the object relates to the environment.Gerring (1999, p. 364) also perceives a close relationship between conceptualization and theorizing, considering concepts as "the building blocks of all inferences" and including theoretical utility as one of the criteria for evaluating a good concept.However, even if he admits that many concepts are theory-driven, Gerring argues that the work of conceptualization should not be reduced to theorization.In his own words: "Concept formation and theory formation are intimately conjoined; the former is not reducible to the latter" (p.365).
Viewing conceptualization as more than a mere means to theorization implies considering the task of describing differently.Conversely, changing the status of description also has implications for conceptualization and measurement.As a latent concept, clientelism can rarely be directly observed.One consequence has been to measure it based on untested assumptions or on questionable or weak evidence that draws its strength from prior adherence to widespread premises.The disregard for description coupled with the lust for causation makes it possible for political scientists to state causal relations between x and y without being sure about how to conceptualize and operationalize x and/or y.Indeed, there is more consensus on the existence of a causal relationship between clientelism and poverty than on the meaning of clientelism itself.Following Gerring, I argue that political scientists must face description as a complex and fundamental task in studies on clientelism.Valuing description goes hand in hand with greater integration between quantitative and qualitative approaches, another promising way of overcoming the persistent conceptual problems on the topic.
Overcoming the limits in studies on clientelism in Latin America requires a more critical and reflexive perspective from political scientists on the premises that inform their research designs and which end up conditioning each step taken in the field as well as the interpretation of each piece of evidence.It requires transforming many untested assumptions into empirical questions, and adding strategies for real integration between qualitative and quantitative approaches to political scientists' admirable disposition toward empirical research.

Concluding Remarks
After two decades since Piattoni's book (2001), it is possible to take stock of what has been produced in Political Science on clientelism.In this article, I have restricted my analysis to Latin American studies for practical reasons.It is very likely that many of the issues raised here also apply to research on clientelism in Asia, Africa, and Europe.Nevertheless, it is also likely that broadening the scope to these regions would imply the need to incorporate other dimensions to the critique.
By breaking down the analysis of the concept of clientelism into four dimensions, my aim was to point out some dilemmas and problems concerning the definition of clientelism which frequently return to the center of debates.I argued that while no study defines clientelism based on the participants or on the object of exchanges, prevailing understandings of the practice tend to restrict it to exchanges of political support for votes involving poor voters and the so-called excludable benefits or private goods.I also pointed out the recurring tendency of this literature to blur vote-buying and clientelism.I highlighted the enormous difficulties encountered by the literature in its attempt to define clientelism as a method of distribution whose distinguishing features would be discretion and contingency.
The article does not ignore significant advances made by contemporary literature, among them the greater diversification of studies, the empirical orientation of this literature, and the increasing willingness of researchers to combine research approaches and techniques.However, I have argued that political scientists need to overcome some limitations that have hindered their efforts: (i) the adoption of universal and rigid assumptions about politicians and voters' behavior; (ii) the strong normative tone of the analyses, almost always based on assumptions about how democracy should work and how clientelism represents a displacement of this ideal; (iii) little reflexivity and criticism regarding premises mentioned above; (iv) the limited way in which scholars have been trying to combine research approaches and techniques.
I have proposed that some ways to overcome the limitations in the field are: (i) restoring description as a crucial task, not only as a means for producing causal inference, but as a way to enhance the internal coherence and depth of concepts; (ii) reevaluating the choice for very rigid frameworks of concept delimitation, centered on necessary conditions, which present high costs in terms of depth and operationalization; (iii) invest in multi-method research designs with effective integration of approaches rather than mere triangulation; and, most importantly, (iv) deal more critically and reflexively with the assumptions that inform conceptualization efforts.
It is important to make clear that at no point have I proposed abandoning the category of clientelism because of its polysemic nature, the disputes surrounding it, and the difficulties of operationalizing it.I agree with Gerring (1999) that if we were to restrict our interest to phenomena that are free of controversy or that can be directly observed and counted, "we would have very little of importance to say" (p.360).Nor have I argued for the need to construct or operate with a definitive concept, invulnerable to criticism.It is expected that the increase and pluralization of studies on the subject and the attempt to apply concepts, hypotheses, and theories generated in one context to other settings will be followed by conceptual disputes.For this reason, comparative studies of clientelism would have much to gain from a more systematic dialogue with the literature on concept formation.Not as a kind of manual, but as useful guides against which to evaluate real choices regarding delimitation.
Finally, it is important to make clear that in this article I have not proposed a new theoretical approach to the investigation of clientelism, new criteria for its conceptualization, or a unified model capable of integrating the criticisms and suggestions and answering all the problems pointed out.I understand that this cannot be an individual task, but rather a collective endeavor in which all scholars interested in the subject should engage.My goal was to draw attention to the deficiencies and biases that have marked the third wave of studies on clientelism, in a different way from those done so far, as well as to contribute to the renewal of the field.