Epstein (2009) Clientelism vs Ideology

CLIENTELISM VERSUS IDEOLOGY

Problems of Party Development in Brazil

Daniel J. Epstein

A B S T R A C T

Although aggregate data on party competition in Brazil seem hopeful,unsettling trends appear in state-level party systems in the years of theconsolidation of Brazilian democracy (the period this article examinesextends through the 2002 elections), such as instability and fragmenta-tion, which exceed the extent of problems apparent at the national level.These hamper the informational role parties can play for voters pickingfrom a large number of candidates. While other possible explanationsshed little light on these problems, a hypothesis about clientelistic party-building strategies may explain the patterns in party competition acrossstates. Such strategies depend on the distribution of selective benefits,such as patronage or vote-buying, to attract candidates, elicit votes andgain office for the party. Clientelistic party-building strategies provide noextra-material incentive for party cohesion, and may stymie the develop-ment of a stable competitive system. Furthermore, it is through clientel-ism that party clans exercise hegemony over the local political systemin some states, promoting an undemocratic monopoly on power.

KEY WORDS � Brazil � clientelism � political parties � regional politics

Introduction

Parties have a key role to play in any democracy: that of providing infor-mational short cuts. The need for parties to play this role is especially criticalin new democracies where politics is in a state of flux that may exhaust voters’capacity to process information about individual politicians. Brazil presentsan extreme case for the indispensability of parties because its Open ListProportional Representation (OLPR) system can present voters with lists of60–600 candidates from which they must choose one.

The party system was one of the most problematic realms of consolidationin the years after the Brazilian military regime left power. Fernando Collor,

PA R T Y P O L I T I C S V O L 1 5 . N o . 3 pp. 335–355

Copyright © 2009 SAGE Publications Los Angeles London New Delhi www.sagepublications.com Singapore Washington DC

1354-0688[DOI: 10.1177/1354068809102250]

at MORA - Parent on November 3, 2015ppq.sagepub.comDownloaded from

http://ppq.sagepub.com/

the first popularly elected president of the current democratic period, createdthe briefly successful Partido Reconstrução Nacional (PRN) as a vehicle forhis election campaign in 1989 (Hagopian, 1996: xvi–xx). However, he wasimpeached in 1992 and the party collapsed almost immediately. In the late1980s, the largest party to emerge from the years of military regime, thePartido do Movimento Democrático Brasileiro (PMDB), split in two, withmany defectors following academic (and later president) Fernando HenriqueCardoso to form a splinter party, the Partido da Social Democracia Brasileira(PSDB). Cardoso and the PSDB went on to win the presidency in 1994, severalgovernorships, including the two largest states, São Paulo and Rio de Janeiro,and nearly a fifth of the seats in the Chamber of Deputies. Nevertheless, theoriginal PMDB was undaunted, and met with success in both gubernatorialand legislative races, maintaining one of the largest congressional delegations.Furthermore, in the next two presidential elections the PMDB supported thePSDB’s candidate, begging the question of how truly separate the parties were.

Over a longer period, one of the other major parties to emerge from themilitary regime has had five different names: Aliança Renovadora Nacional(ARENA), then Partido Democrático Social (PDS), Partido ProgressistaReformador (PPR), Partido Progressista Brasileiro (PPB) and, finally, PartidoProgressista (PP) after 2003. It merged with two other parties along the way,and its current name is in fact the same as an old merger partner (PP website,2008; note that for the graphs in this article this party is labelled by itsacronym at the time of the 2002 elections, PPB). Another source of confusionis the similarity of some party names. The Partido dos Trabalhadores (PT),the Partido Trabalhista Brasileiro (PTB), the Partido Democrático Trabalhista(PDT) and several other minor parties with similar names may muddle aballot for many voters. Moreover, the names of these parties can sometimesbe misleading, in ideological terms; while the PT has a history of ‘anchoringthe left’ (Mainwaring, 1999: 19), the PTB adheres to the centre–right in whatlittle ideology it manifests. Many scholars have lamented the tumult in theBrazilian party system since the end of the military regime (Mainwaring,1999; Power, 2000: 28–30). Explanations of such problems favour the insti-tutional approach (Ames, 1995, 2001; Mainwaring, 1991; Mainwaring andPérez-Liñan, 1997; Mainwaring and Shugart, 1997), especially faulting theOLPR system used to elect both state and national legislatures. Anotherproblem has been the parties’ fluid membership. Mainwaring and Pérez-Liñan (1997) have shown the atrocious record of party unity in the initialpost-transition years. Furthermore, Mainwaring (1999) notes a number ofdisturbing trends in party development, such as high vote-volatility insequential elections, which he claims indicates a weakly institutionalized partysystem. Power (2000: 92–3) brings to light severe identifiability problemsamong parties of the direita envergonhada or ‘embarrassed right’: ideo-logically conservative parties take names that sound leftist, declare them-selves ‘centrist’ and in some cases their candidates even avoid divulging theirparty label during campaigns.

PA RT Y P O L I T I C S 1 5 ( 3 )

336



However, a look at Figures 1 and 2 does not betray a sense of theseproblems. Throughout the late 1990s and into the beginning of the newcentury, four or five parties, which reflect the whole political spectrum,dominate the Brazilian legislature. Though parties like the PMDB and PPhad to give way to rising parties like the PT and PFL over the years, the

E P S T E I N : C L I E N T E L I S M V E R S U S I D E O L O G Y

337

Figure 1. Composition of Brazilian Chamber of Deputies, 1982–2002Source: Instituto Universitário de Pesquisas de Rio de Janeiro website(http://www.iuperj.br/deb/ing/Cap2/Cap2_tab20.htm)

Figure 2. Composition of Brazilian Senate (seats won each election), 1986–2002Source: Instituto Universitário de Pesquisas de Rio de Janeiro website(http://www.iuperj.br/deb/ing/Cap3/Cap3_tab1.htm)

0% 20% 40% 60% 80% 100%

1982

1986

1990

1994

1998

2002

PTPSDBPMDBPFLPPBPDTPTBPLPSBPCB/PPSPSDPC do BPSTPSCPSLPVPMNPRONAPRPPSDCPRN/PTCPDCPTR/PPPRS

0% 20% 40% 60% 80% 100%

1986

1990

1994

1998

2002

PTPSDBPMDBPFLPPBPDTPTBPLPSBPCB/PPSPSDPSTPMNPRN/PTCPDCPTR/PP



system does not show signs of extreme fragmentation, and the major partiesshow ample endurance, with no signs of a collapse of the party system.

The Importance of Healthy Parties

The problems of unstable or ‘weakly institutionalized’ political parties canhave a serious impact on both governability and representation. Difficulty inbuilding legislative coalitions, building public support for new policies andproducing an electoral mandate to oppose vested interests are all supremelydifficult without strong, competitive parties. Representation, however, is asfundamental to democracy as governability (if not more so), and here iswhere parties have an absolutely indispensable role to play.

Downs (1957: 96–9) explains that rational voters need ideologies to sum-marize parties’ policy platforms in order to vote according to their interests,because they have neither the time nor the resources to find and sufficientlystudy all parties’ platforms to make an informed decision. Implicit in thislogic, and more fundamental to democracy, is that voters need party labelsto summarize the policy stances of individual candidates. Without parties toprovide informational ‘short cuts’, most voters (especially those with limitedwealth and education) will hardly be able to vote their preferences, whichmay turn voting into a random act, from a representational perspective.

Three major factors contribute to a party’s ability to provide necessaryinformation:

• Continuity – whether the party is present as a major player in electionafter election.

• Distinguishability – whether the party is distinguishable as a major one,or just another among a profusion of tiny, futile parties.

• Identifiability – whether the statements and activities of politicians clarifyor obscure their membership in the party and what the party stands for.

Without a continued presence in a given political system – or even if a partyflips back and forth from being a serious competitor to presenting nobodiesas candidates – the information that voters acquire about a party is uselessby the next election. This frustrates the rational voter’s attempts to assesswhich parties’ ideologies are worth expending scarce informational resourcesto explore. Continuity results from factors inside a party (how seriously ittries to compete in a given election) and outside (how successful otherparties are at beating it out). Distinguishability depends on similar factors,because even when a party makes a serious effort to compete, if enoughother parties do the same, then they may each be left with a tiny slice of thevote that renders legislating the party’s platform futile. Causal factors foridentifiability, however, are internal to political parties (e.g. the example ofthe ‘embarrassed right’).

PA RT Y P O L I T I C S 1 5 ( 3 )

338



These three factors are critical, especially in a multiparty system as inBrazil, in order to guarantee an essential minimum of democratic represen-tation: the capacity to ‘throw the rascals out’. Unless parties are clearlyenough understood for people to know who to vote for if they want theincumbents out, ‘the rascals’ may survive in office even in the face of amajority willing that they leave. If a system is unstable, then a minority ofvoters might opt for opposition parties that did well in the last election,while others might choose promising new opposition parties, and the ‘rascals’could remain in power. Similarly, if there are so many parties that an oppo-sitionist voter cannot tell who the serious competitors are, incumbentsmight be able to salvage enough votes to maintain their seats because anti-incumbents’ votes are split among too many opposition parties.

The State Level

Although the national data presented above do not suggest that Brazil’sparties are hampered in playing their critical informational role, it is im-portant to consider political development at the sub-national level as well,especially in such a large and heterogeneous country. Brazil is a federalsystem made up of 26 states and one federal (capital) district. Congressionaldeputies, senators and governors are elected in state-wide districts whichconstitute the main arenas of party competition and ‘play central parts inthe drama of national politics’ (Ames, 2001: 98). Moreover, the state-levelorganizations of Brazil’s political parties are fairly autonomous and theirdecisions are the most important, and often national party leaders cannotcontrol them (de Moraes, 1997: 35). Thus, for political parties, state-levelphenomena are more likely to have a causal effect on national politics thanthe other way round.

Problems stemming from continuity and distinguishability can easily beseen at the state level in electoral results. The state of Rondônia (Figure 3)presents a good example of the pattern of a competitive, but unstable, partysystem. In any given election, only a couple of parties win more than a fifthof the seats, which ought to help voters assess the parties, except that bythe next election it is a different set of parties. The PTB disappeared entirelyafter a big victory in 1990 and then returned in 1998 as only a minor party.The PFL, by contrast, seems to have had an aberration in 1990 (the begin-ning of stability for most other parties), disappearing entirely. The PDT hada great showing in 1994, but has won no seats since then. And one of thetwo most successful parties in 2002, the PT, made its first appearance inthe state’s delegation. Other states that show patterns resembling that ofRondônia are Roraima, Paraná and, to a lesser degree, Amazonas.

The problem of distinguishability is most salient in a fragmented system.If many parties win small proportions of the seats, then parties can playonly a limited informational role, because it is difficult to distinguish major


339



parties1 from the rest. A profusion of parties that win a tiny percentage ofthe vote make it hard for voters to learn about them all. Rio de Janeiro(Figure 4) provides the most striking example of this fragmented competi-tion. Major parties over the years have been the PDT (which is based in thestate) in the early 1990s, the PSDB and PFL in the late 1990s, the PMDBat various points, and in 2002 the PT and PSB, but in almost every case theparties emerging strong from the previous election drop to minor-party status

PA RT Y P O L I T I C S 1 5 ( 3 )

340

Figure 3. Party shares Rondônia delegation to Chamber of Deputies, 1982–2002Source: Instituto Universitário de Pesquisas de Rio de Janeiro website(http://www.iuperj.br/deb/ing/Cap3/Cap3_tab1.htm)

Figure 4. Party shares Rio de Janeiro delegation to Chamber of Deputies,1982–2002

Source: Instituto Universitário de Pesquisas de Rio de Janeiro website(http://www.iuperj.br/deb/ing/Cap3/Cap3_tab1.htm)

0% 20% 40% 60% 80% 100%

1982

1986

1990

1994

1998

2002

PT

PFL

PMDB

PSDB

PPB

PL

PTB

PDT

PCB/PPS

PTR/PTR/PP

0% 20% 40% 60% 80% 100%

1982

1986

1990

1994

1998

2002

PT

PFL

PMDB

PSDB

PPB

PL

PTB

PSB

PDT

PCdoB

PV

PSD

PSC

PSDC

PDC

PRN

PTR/PP



in the next, making it difficult for voters to keep track of all the potentialplayers. Many Brazilian states are afflicted by a very fragmented party system,including the large states of Minas Gerais and (to a lesser degree) São Paulo,as well as a number of small states, such as Mato Grosso, Mato Grosso doSul, Espirito Santo, Alagoas and Sergipe.

Santa Catarina (Figure 5) is perhaps the best example of the few Brazilianstates in which parties’ stability and distinguishability promote a healthysystem of party competition. Throughout the late 1980s and 1990s, thePFL, PMDB and PP constituted most of the state’s delegation, and so voterscould easily learn which were major parties and avoid ‘wasting’ their votes.The gradual evolution of PT into a competitive party (and the waning ofPFL) also shows that the system is not so ossified or oligopolistic that itexcludes new points of view. From the perspective of democratic choice andrepresentation, this stable competitive pattern approaches the ideal partysystem, because there are not too many parties (though still enough toaccommodate most ideological viewpoints), and it is clear how to voteagainst a given party. The other states that fit this pattern are Rio Grandedo Sul, Piauí, Rio Grande do Norte and, to a degree, Goiás.

The fourth important pattern of competition found in Brazilian states isa hegemonic party system. It presents a threat to the very concept of demo-cratic choice, because ultimately the system is not one of competition, butof domination (or co-optation). This is comparable to the PRI hegemony inMexico, where for several decades opposition parties were allowed to winonly a handful of seats in the Mexican congress, and never any presidentialor gubernatorial elections. The state of Bahia (Figure 6) is the best exampleof this type of competition, or lack of competition, as the case may be.


341

Figure 5. Party shares Santa Catarina delegation to Chamber of Deputies,1982–2002

Source: Instituto Universitário de Pesquisas de Rio de Janeiro website(http://www.iuperj.br/deb/ing/Cap3/Cap3_tab1.htm)

0% 20% 40% 60% 80% 100%

1982

1986

1990

1994

1998

2002

PT

PFL

PMDB

PSDB

PPB

PL

PDT



After its creation in the mid-1980s, the PFL quickly came to dominate Bahianelections, as one of its leaders, former governor and senator Antônio CarlosMagalhães (aka ACM), was Bahia’s principal political boss. A hegemonicparty system does not necessarily imply that one party always wins all theseats, but rather that it totally dominates the other parties, as well as theoutcomes of the first-past-the-post gubernatorial and senatorial elections. Ifa party consistently wins all or most of these elections, it is exercisinghegemony in the state, even though its use of allies in the proportional repre-sentation races dilutes its numerical dominance of the congressional delega-tions. In Bahia, the PFL won every gubernatorial and senatorial contest from1994 until its hold was broken in 2006.2 Tocantins, Maranhão and Cearáalso show the signs of a hegemonic system. Signs of hegemony were alsopresent in Paraíba and Pernambuco throughout the 1990s, although the2002 elections marked the end of them.

Thus, the four patterns of party competition visible in Brazil’s states are:

(1) Fragmented: many parties win a tiny share of seats in each election,making it difficult for rational voters to gather information usefully, asthere are too many parties and it is unclear which might be majorcompetitors.

(2) Unstable Competitive: in any given election, only a few parties win seats,but the set of major competitive parties changes from election to election,making it difficult for rational voters to gather information since theknowledge they gather in one campaign proves useless in the next.

(3) Stable Competitive: in election after election, the same few parties winseats, facilitating the usefulness of party labels as informational shortcuts for rational voters.

PA RT Y P O L I T I C S 1 5 ( 3 )

342

Figure 6. Party shares Bahia delegation to the Chamber of Deputies, 1982–2002Source: Instituto Universitário de Pesquisas de Rio de Janeiro website(http://www.iuperj.br/deb/ing/Cap3/Cap3_tab1.htm)

0% 20% 40% 60% 80% 100%

1982

1986

1990

1994

1998

2002

PT

PFL

PMDB

PSDB

PPB

PL

PTB

PSB

PDT

PCB/PPS

PCdoB

PV

PDC

PRN

PTR/PP



(4) Hegemonic: one party wins the lion’s share of seats in every election,including majoritarian polls, which marginalizes competition, eliminatingtrue electoral choice for rational voters through unfair, if not illegal,means.

This typology stems from informal analysis of electoral results. A moresystematic, quantitative method to disaggregate the factors of electoraltrends that contribute to these patterns, and sort the states accordingly, willmake it more useful. The two factors are concentration (or fragmentation)and stability (or volatility) in the partisan make-up of the congressionaldelegations.

Measuring Types and Quantitative Categorization

One accepted measurement of fragmentation is the effective number of parties(Laakso and Taagepera, 1979). However, it is difficult to compare the ENPbetween states that have differently sized delegations. Instead, I have deviseda perhaps more crude measure, but one that is more effective for distin-guishing among types. It uses a threshold number of seats above which anyparty is considered a ‘major party’ and below a ‘minor party’. The thresholdnumber varies by state but is around a fifth of the delegation. This thresholdis based on the idea that voters could probably keep track of up to four partiesrelatively easily, but no more (perhaps a moderate and extreme party on eitherside of a left–right spectrum). A one-fifth threshold allows for four majorparties, if we expect minor parties to take up a small portion as well. A thresh-old also provides a useful alternative for thinking about the continuity thatdifferentiates between stable and unstable competitive systems. It is againstatistically crude: simply to count the number of parties that cross the thresh-old (either rising above or dropping below) from one election to the next. Asopposed to traditional measures of volatility that capture all parties, thiscaptures changes in the set of major parties (those competitive enough to bedistinguishable) without injecting ‘noise’ from the turnover of minor ones,or from shuffling of the order within the group of those considered as major.

I calculated my measures using results from three recent elections toBrazil’s Chamber of Deputies (1994, 1998 and 2002) for the 26 states andone federal district (Table 1). The concentration measure is the fractionexpressing the proportion of seats in the chamber won by all parties thatexceeded the state’s threshold. The closer it is to 1, the fewer the numberof seats held by minor parties, the more distinguishable are the partiescompeting and the less fragmentation. The stability measure divides thenumber of threshold crossings by the maximum observed (five in Paraná in2002) and subtracts the result from 1. Thus, the closer to 1, the more stablethe set of major parties.3 The following scatterplot (Figure 7) helps us seehow the various competition types cluster.


343



The five states not shaded were those which, from raw electoral results,were too close to call, and, not surprisingly, they are at the intersection ofall the clusters. Besides these, the Fragmented states have low concentrationand medium stability; Unstable Competitive states have medium concen-tration and low stability; and then the cluster of Stable Competitive states,which show the highest ratings on both measures, overlaps with Hegem-onic states, whose cluster is centred at slightly lower levels of concentrationand stability. The Hegemonic cluster is problematic, as Pernambuco (PE) isa bit far from the cluster’s centre, perhaps due to an apparent recent breakof the PFL’s hegemonic grip (the two states where hegemony looks to havebeen broken in 2002 are italicized). Tocantins (TO) is also a bit of an outlier,but the hold of its local PFL on the governor’s seat and the senate seats,unbroken since 1994, gives confidence that it is Hegemonic – its location on

PA RT Y P O L I T I C S 1 5 ( 3 )

344

Table 1. Average stability and concentration of parties in Chamber of Deputiesdelegations, 1994–2002

State Stability Concentration

Acre (AC) 0.6 0.708333Alagoas (AG) 0.53333 0.37037Amapá (AP) 0.53333 0.583333Amazonas (AM) 0.53333 0.666667Bahia (BA) 0.73333 0.632479Ceará (CE) 0.93333 0.69697Distrito Federal (DF) 0.6 0.583333Espirito Santo (ES) 0.6 0.566667Rondônia (RO) 0.26667 0.625Goiás (GO) 0.66667 0.588235Maranhão (MA) 0.6 0.62963Mato Grosso (MT) 0.6 0.333333Mato Grosso do Sul (MS) 0.46667 0.5Minas Gerais (MG) 0.6 0.496855Pará (PA) 0.53333 0.72549Paraíba (PB) 0.66667 0.694444Paraná (PR) 0.13333 0.566667Pernambuco (PE) 0.93333 0.52Piauí (PI) 0.86667 0.833333Rio de Janeiro (RJ) 0.46667 0.355072Rio Grande do Norte (RN) 1 0.791667Rio Grande do Sul (RS) 0.86667 0.698925Roraima (RR) 0.46667 0.625Santa Catarina (SC) 0.86667 0.75São Paulo (SP) 0.73333 0.538095Sergipe (SE) 0.53333 0.416667Tocantins (TO) 0.46667 0.833333All States 0.62222 0.604811037



the scatterplot may indicate that in congressional elections the PFL switchesallies quite frequently.

Potential Explanations

Much of the recent literature focusing on party problems in Brazil explainsthis by institutional rules such as the OLPR electoral formula, the electoralcalendar, the run-off presidential system and other idiosyncrasies, such asthe recently retired candidato nato rule that guaranteed any sitting deputya place on the party list (Ames, 1995, 2001; Mainwaring, 1991, 1999;Mainwaring and Shugart, 1997). However, while these can explain generaltumult at the candidate level, and especially problems of party unity, theydo not provide as much purchase on problems of continuity or identifiabil-ity. Moreover, they cannot explain variation in party behaviour across states,or variation across parties at the national or state level.

Another potential explanation could be differences in economic develop-ment. For example, poor voters in less-developed states may have no hopethat politics will improve their lot, and thus put little of their scarce timeand resources into researching their choices at the polls, so that if they dovote at all it is nearly a random choice. This could result in high instabilityand low concentration. By the same assumptions, higher-income voters inwealthy states may be much more engaged in politics, leading to morestable, patterned voting and a Stable Competitive system. However, data onper capita income (Instituto Brasileiro de Geografia e Estatística, 2001) of


345

Figure 7. Relative concentration and stability

RN

CE PE

PI SC RS

PB BA GO SP

AC MA DF ES MG MT

PA AM AP SE AG

TO RR MS RJ

RO

PR

Competition type Fragmented Stable Competitive Unstable competitive Hegemonic

Low stability

High concentration Medium concentration Low concentration

High stability

Medium stability



states do not seem to align with any of the four competition types, with theexception of Hegemonic states, which were all below R$4000 per capitaannual income in 2001 (aggregate for Brazil was about R$7000).4 Multi-nomial logistic regressions do show a statistically significant negative relation-ship between income and hegemonic character of the system,5 but none withany other category. There are also six non-hegemonic states that fell withinthe same income range as the Hegemonic states, which suggests that it isnot simply poverty that drives hegemony. Moreover, the fact that the wealth-iest of all, the Distrito Federal (with a per capita income nearly double thatof any other state), fell in the very middle of both the concentration andstability spectra leads us to believe that income does not drive trends inparty competition.

Another possible explanation would draw on the work of Lipset andRokkan (1967) on social cleavages and party formation. However, data(IBGE, 2000 Census) on the most obvious social cleavage in Brazil – race –give small purchase on the divergence of party system competition types.Multinomial logistic regressions show that the only relationships that standout are both related to Stable Competitive systems: a positive one with theproportion of white population6 and a negative relationship with the pro-portion of population reported as neither white nor black (parda – brown,amarela – yellow or indigena).7 Perhaps this would suggest that the more‘middle’ skin tones in a state, the less likely that whites and blacks at eitherend of the pigmentary spectrum will coalesce into durable, competitiveparties. However, the fact that there is no relationship between StableCompetitive systems and proportion of black population casts doubt onthis. Another disconfirmation of race-based party development is that of theStable Competitive states; two have white populations near 90 percent (RSand SC) with tiny black populations. While race could be involved here, it iscertainly not in the form of partisan cleavage. Analysis of another importantcleavage, urban–rural, yields no relationships at all, either with proportionsrural or urban, or with how closely the divide approaches parity.

If wealth or cleavage variations do not explain variation in party systemcompetition types, perhaps parties themselves do. In a multiparty federalsystem like Brazil’s, where many parties do not aggregate across states(Chhibber and Kollman, 1998), electoral arenas in different states will bemore populated with some parties than with others over time. Perhaps someparties are simply more stable, while others experience peaks and valleysover time because of opportunism or lack of discipline or organization. Inorder to test this, it is important to look for the presence of parties overtime in a given party system, and not just their presence in the most recentelection (which, in Unstable Competitive states, is unrepresentative by defin-ition). However, since the competition type can change over time, and weare most interested in the prevailing type as measured by the data presentedearlier from the three elections of 1994–2002, it is useful to weight the morerecent elections more heavily, while still allowing for older elections to have

PA RT Y P O L I T I C S 1 5 ( 3 )

346



had some effect on producing the prevailing competition type. Thus, I havedevised a measure that averages each party’s share of seats in a given stateover the past five elections discounting by 20 percent per cycle the weight ofseats won in earlier elections.8 Again using multinomial logistic regressionsto test the effect of the presence of each of the major parties on competitiontype, very little can be found. The only statistically significant relationshipsare that the PT tends to be absent from Hegemonic systems, and that thePTB tends to be present in Unstable Competitive states.9 The former is un-surprising, given that all of the Hegemonic states are in the Northeast regionof Brazil, where the PT has historically been the weakest. However, we mightthen expect an opposite relationship between the PFL and Hegemonicsystems, as it tends to be quite strong in the Northeast, and the hegemonicparty in several of the Hegemonic systems. However, only at a substantiallylower level of statistical significance can a positive relationship between thePFL and Hegemonic systems be found,10 and adding any control variablesmakes for an even weaker relationship.

A New Hypothesis: Linkage Mechanisms as anExplanation

Another possible explanation for the variation in competition types lies in thestrategies that parties use to connect to voters and win elections. Differenttypes of ‘linkage mechanisms’ (Kitschelt, 2000) that connect parties andthose who vote for them have implications not just for the votes cast onelection day, but also for candidates, who self-select into parties that, in theirestimation, offer the best linkage mechanism. ‘Best’ means ‘most effectivefor winning votes’, but will vary depending on the candidate’s conceptionof himself as a candidate, the type of voters he thinks he will appeal to, andthe nature of the district in which he is running. Moreover, the differenttypes of linkage mechanism imply different kinds of voters and candidatesfor parties that employ different mechanisms, as well as different behaviourover time. In classic Downsian party theory, only one linkage mechanism iscontemplated: ideology. However, a more nuanced view of parties showshow Kitschelt’s idea of tripartite typology may actually be more realistic,and may well help to explain Brazil’s variation in competition type.

Schlesinger (1984: 380) points out that a key characteristic of parties isthat they operate on market-based principles: they must provide a ‘product’that attracts enough ‘customers’ to support their continued existence. Theyoffer policies and candidates, collective goods from which all the electoratederives benefits if they are elected, no matter who voted for them (Schlesinger,1984: 381). Voters then choose the party product they like best, and the elec-toral market rewards with office parties that have made a desirable ‘product’and punishes the others, simultaneously determining what government policyshall be. Downs’s (1957) concept of ideology as an informational short cut


347



was classically the only product parties would offer to voters, who wouldthen decide which ideological ‘flavour’ they preferred.

However, Kitschelt (2000: 849) suggests, in effect, that parties may offersubstitute ‘products’ to voters: personalistic linkages or clientelistic linkages.Personalistic linkages involve the exposition (usually through the media)of appealing personal characteristics of a party’s leaders, and clientelisticlinkages employ an organization designed to distribute selective benefits toindividual voters or groups of voters who support the party. Linkage mech-anism is determined by the kind of organizational infrastructure the partydecides to invest in. If it develops both an organization capable of reachingvoters in person to ‘get out the vote’, and also a means to aggregate theinterests of party members into a programme (or ideology) used to convincevoters, it can employ a programmatic linkage. If it invests in neither organiz-ation nor interest aggregation, then it has little but the leaders’ personalappeal to draw voters: a personalistic linkage. If the party invests only inorganization, but does not aggregate interests of its supporters in order tooffer them a programme, then it must give clientelistic benefits in exchangefor support. If voters are willing to ‘purchase’ these substitute products, thena party need not be concerned with behaviours that reduce its usefulness asan informational tool, and, in fact, maintaining that usefulness might evenimpede the flexibility necessary to maximize access to clientelistic resources.Such substitute products, especially Brazil’s ubiquitous clientelism, disruptthe functioning of parties according to the classic theories.

Clientelism directly undermines the critical role that parties must play indemocracies by removing policy as a reason to vote for one party overanother. In fact, legislators who have employed clientelistic linkage mech-anisms to get elected are actually insulated from the policy interests of theirconstituents because they have substituted non-policy, selective benefits forthe collective, policy-based benefits that traditional Downsian parties use tomarket themselves. Moreover, the party loyalty of both politicians and voterswho have chosen a party based on clientelistic linkages can be problematicfor parties as well. For example, if the supply of clientelistic benefits a partyis able to provide drops, then it is likely that both its voters and candidateswill switch to a party with a greater capacity to supply them. This could occurbecause of a change in government at the state or federal level, an economicdownturn, or even just a shuffling of alliances among wealthy elites. Candi-dates from more programmatic parties, however, presumably joined the givenparty because they agree with its ideology, and believe that there is a sectorof the electorate that also agrees, from whom they can elicit votes. And, iftheir choice of party was predicated on the value of information that associ-ation with the party can convey to voters, then switching parties or changingthe party’s name are events that they would prefer to avoid, because of thedamage they would do to their capacity to convey information. Even if theparty did poorly in a given election, such candidates would be more likelyto stick with it than those of clientelistic parties, making it less liable to total

PA RT Y P O L I T I C S 1 5 ( 3 )

348



collapse. Voters for such parties, also, are less likely to switch or abandonthe party if they picked the party initially based on its ideology. They haveinvested scarce resources in collecting information about it, and are presum-ably not interested in the varying levels of selective benefits that differentclientelistic parties may be offering. And while they might prefer anotherparty of the same ideology, there may not be many out there to select from,and, after all, even that switch would require the gathering of additionalinformation. Clientelistic parties, however, have no hold over their candi-dates or voters besides the very short-term benefits they provide, so if thesechange, their partisans have no reason to stick with them.

Thus, the more the parties in a given state employ clientelistic linkagemechanisms, the more instability is likely. In any given election, only a fewparties may be able to collect the resources to offer selective benefits broadlyenough to be elected. But any number of events could shift the balance ofsuch resources, making it unlikely that it would be the same set of partiesfrom one election to the next: an Unstable Competitive system. By compari-son, considering a state with only programmatic parties – parties that aretruly oriented towards policy programmes – will likely coalesce around afew ideological viewpoints (perhaps an extreme and a moderate point oneither side of a left–right spectrum). Furthermore, party politicians andvoters are likely to stick with their party from one election to the next asthe best way to see their preferred ideology carried into policy (and alsobecause they share opinions with their co-partisans, potentially increasinginternal bonds of the party through legislative cooperation). This is likelyto produce a Stable Competitive system. It is also possible that both typesof parties are capable of existing in the same system. If some parties canhand out enough selective benefits to win seats, and other parties can appealto points on an ideological spectrum to win seats, this may add up to manyparties capable of winning a few seats each: a Fragmented system. Finally,consider if a single party were able to monopolize the resources necessaryfor providing selective benefits, and were thus able to provide such exten-sive selective benefits that most voters chose this party instead of those thatcould only offer the collective benefits of programmatic linkage mechanisms.If the party were able to control state resources over time because of electoraldominance, this would create a Hegemonic system. Importantly, this ismuch more likely in states where private sources of clientelistic benefits arenot extensive enough to compete with state resources, allowing electoraldominance to translate directly into the dominance of clientelistic resources.Moreover, the other opposition parties in such a system would be left withonly programmatic linkages with which to attract votes. Thus, like a Frag-mented system, a Hegemonic system is also theorized to contain a few pro-grammatic parties, though they will likely only show up as minor parties.


349



Measuring Clientelism – Testing the Explanation

Though there is no index of clientelism that can tell us how clientelistic orprogrammatic parties are, we can still make useful observations and evendraw some tentative conclusions from available data. While we found littlerelationship between any national party and competition type, the indepen-dence of state-level party organizations gives us reason to expect variationacross states in the type of linkage mechanism employed by a party. Onepossible proxy for clientelism is suggested by the work of Desposato (2001),who shows that societal variables condition vote-buying and other clientel-istic behaviour (such as corruption); specifically, that greater poverty, in-equality and generally lower social indicators coincide with a far higherincidence of clientelistic behaviour. The survey work of Transparência Brasilon the 2000 and 2002 elections corroborates this, showing that the poorerNortheast, North and Centre–West regions report higher levels of vote-buyingthan the South and Southeast (Transparência Brasil, 2003). Unfortunately, thedata are aggregated to the level of region, which precludes using it to analysestate-level competition types. Nevertheless, descriptive statistics provide someconfidence for clientelism as a cause: all the Hegemonic systems occurred inthe Northeast or North and Centre–West, which reported the highest levelsof corruption and vote-buying, while no Hegemonic or Unstable Competitivesystems are located in the Southeast, which had the lowest scores. However,since three of the five Stable Competitive systems (RN, PI and GO) are foundin the Northeast, North or Centre–West, where vote-buying is most preva-lent, while only two of the seven states in the South and Southeast evince aStable Competitive type, these regionally aggregated data give only weaksupport to the hypothesis that clientelism explains the variation.

Without an index of clientelism or state-level data on vote-buying thatcould be used as a dependable proxy for clientelistic linkages, we must seekother data at the state level that could be used in that capacity, based on ourconception of clientelism and the Brazilian electoral system. One importantaspect of the system is the Open List. Voters in congressional elections votefor one candidate out of dozens or hundreds on a list, each of whom has aparty label. However, because the number of seats allotted to each partydepends on the pooled number of votes on that party’s list, parties are incen-tivized to fill out their lists with candidates, since even those not likely towin enough votes to be elected may still bring in a few votes that will benefitthe overall list. At the same time, a list including a few extremely popularcandidates will often receive enough votes to have seats allocated even forits less popular members. Thus, if some parties depend on a few clientelisticcandidates to sweep a large slate into office, the top vote-getters may receivemany more votes than the least vote-getters, creating a huge variance in thenumber of votes received. By contrast, a programmatic party whose candi-dates all equally represent the same ideology should, theoretically, attractan even number of votes across the list: voters should be indifferent to which

PA RT Y P O L I T I C S 1 5 ( 3 )

350



of the party’s candidates they vote for. Of course, factors such as incum-bency, popularity or ability to convey the party’s message may still skew aprogrammatic party’s vote totals across the list, but it should theoreticallyexhibit a lower variance than that of clientelistic parties. Only perhaps in ahegemonic party would clientelistic benefits be handed out more evenly, ifparty leaders wished to spread them out, and popular candidates could notdefect to another clientelistic party’s list.

I collected data on the variances within candidate lists for parties in 19of Brazil’s 27 federal units for the 1998 and 2002 congressional elections,an unfortunately low N, especially for regression analysis. In order to makedata comparable across parties and states, I calculated the coefficient ofvariance for each party list, and then in order to amalgamate the values tothe state level for analysis I weighted them by the fraction of seats each partywon. This provides an average measure of how skewed or even candidatevote totals are within each state.

Figure 8 shows the 16 states from the four system types (I collected datafrom three too-close-to-call states for the regression analysis, but they areleft out of the chart). From the left, we have four Hegemonic states (BA,CE, MA and PE), three Unstable Competitive states (PR, RO and RR), fiveFragmented states (RJ, SE, SP, MG and MT) and four Stable Competitivestates (RN, RS, SC and PI). The Stable Competitive systems should have thelowest values, as they are expected to have the fewest clientelistic parties,while the Unstable Competitive systems should have the highest values. Hege-monic systems and Fragmented systems should be somewhere in between. Amultinomial logistic regression of the two years combined yielded only anordinal ranking, and that at a low level of confidence: Hegemonic systemsevince the lowest coefficients of variance, then Stable Competitive Systems,then Unstable Competitive and the highest are seen in Fragmented systems.Although the confidence level of these rankings is weak, and the Fragmentedsystems are certainly out of order, the relative ranks of Stable and Unstable


351

Figure 8. Weighted average coefficients of variance for parties in 16 Brazilian statesNote: Coefficients of variance give a standardized measure of how much intra-party vote totals

diverge from the mean for the party

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

BA CE MA PE PR RO RR RJ SE SP MG MT RN RS SC PI

Coe

ffici

ent o

f var

ianc

e

1998

2002



Competitive systems do at least point in the right direction. Intra-party votetotals are more skewed in Unstable Competitive systems than in StableCompetitive systems.

The main mechanism by which clientelism may cause instability is throughparty-switching (see above). This has been studied extensively amongdeputies elected to the Brazilian Congress (Figueireido and Limongi, 2000;Mainwaring and Pérez-Liñan, 1997), but only among successful candidates.Concrete data on all candidates are difficult to come by, and the only alterna-tive is manual coding by checking lists of candidate names. I coded repeatcandidates for 12 states (3 of each type from the 19 above) for the 1998 and2002 elections, calculating the ratio of party-changing to party-loyal candi-dates, which should give a good comparative index (Figure 9) of whethermore candidates abandon their parties in some states than in others, animplication of the theory that clientelistic parties should have a harder timeholding on to their partisans than programmatic parties have.

Most states had fewer candidates switching parties than staying in the sameparty.11 In general, the graph comports with the expectations of the theory:the Stable Competitive states suffered the least party-switching, and UnstableCompetitive states the most. Parties in Hegemonic states, as we might expect,were also pretty effective at maintaining their candidates (as candidates moti-vated by the clientelism of Hegemonic parties would have no other party togo to), though not as effective as those in Stable Competitive states. Partiesin the Fragmented states, on average, fared slightly worse than those inHegemonic states, but better than those in Unstable Competitive states,which we might expect if roughly equal proportions of clientelistic and pro-grammatic parties populate Fragmented states’ party systems. Multinomiallogistic regressions confirmed this ordering, though again at mediocre levelsof confidence (most not better than 60 percent, and in fact much less forthe distinction between Hegemonic and Fragmented systems), most likelybecause of the very low N (12) and the overwhelming frequency of party-changing in Roraima.

PA RT Y P O L I T I C S 1 5 ( 3 )

352

Figure 9. Ratios of party-changing candidates to party-loyal candidiates, 1998–2002Note: A higher value indicates more party-switching and thus perhaps more clientelistic parties

21.81.61.41.2

10.80.60.40.2

0BA CE MA PR RO RR RJ SE SP RN RS SC



The data presented here suggest that the choice by state-level party organiz-ations of clientelistic linkage mechanisms may indeed be the key to explain-ing the variation in patterns of competition in state-level party systems inBrazil. Though statistical analysis does not produce results at high enoughlevels of confidence to be certain, it seems to favour this hypothesis over othersof economic development level, racial or urban–rural cleavages, or nationalparty trends. This notwithstanding, the last of these alternative explanationscontains a germ of the clientelism hypothesis, if we suspect that while moststate-level party organizations choose whether they will pursue clientelisticor programmatic strategies, some parties, e.g. the PTB and PT, may be soimbued with clientelistic or programmatic linkage mechanisms that they douse the same strategy everywhere. Thus, linkage mechanisms may explain thefindings of PTB presence in Unstable Competitive systems and PT absence inHegemonic systems. Meanwhile, we could still expect other parties to exercisemore local autonomy, perhaps even mimicking the linkage mechanisms ofother parties in their state, instead of adhering to a national strategy.

Conclusions

This piece has argued that while Brazil’s national party system looks to be‘settling down’ nicely, divergent trends in state-level party systems, wheremost voters actually interact with parties, are potentially quite damaging toparties’ capacity to play their key informational role in democracy. Afterconsidering some potential explanations, the hypothesis was advanced thatstate-level party organizations’ choice of clientelistic versus programmaticlinkages could be the best explanation. Proxy data on clientelism were pre-sented, as well as some analysis thereof, which gives some confirmation tothe hypothesis, although at rather low confidence levels. These are probablyprimarily the result of a maximum N of 27 (the number of Brazil’s federalunits), and in several cases even fewer because of data constraints. Improve-ment could be made by collecting full datasets for all states, and also by usingdifferent kinds of data that might serve as better proxies for clientelism,such as surveys of corruption and its association with specific parties inspecific states, or state-level measures of ‘pork’ legislation or public employ-ment. Another improvement would be better specified measures of thecompetition type, as some of the states could not be classified at all, perhapsincluding electoral results from state-legislative elections as well, and findinga formula by which to amalgamate these with the congressional, senatorialand gubernatorial electoral results.

Another important set of improvements would stem from field researchallowing for a more qualitative component to this analysis, which mightclarify some of the vaguer statistical patterns and perhaps bring to lightunseen trends (such as how parties cooperate over time, which might changeour assessments of fragmentation and hegemony, if stable coalitions play


353



the informational role of parties in some states). It might also suggest someother potential explanations for the divergence of state party systems.

Nevertheless, Mainwaring’s quote from the governor of Paraíba state maywell be the most apt in assessing the growing pains of Brazilian democracy:‘The biggest party in Brazil is the PCB, Brazilian Clientelistic Party. If wecan’t get rid of this party, we’ll never be able to solve the problems of thecountry’ (Folha de São Paulo, 1 November 1987, in Mainwaring, 1999: 175).

Notes

1 ‘Major’ parties are those that win enough seats to have a real chance to legislate,as opposed to tiny, futile parties, which rarely win more than a couple of seats,if any, and thus are manifestly unable to translate the preferences of theirsupporters into legislation comporting with their ideologies.

2 1994 was also the last time that a PFL candidate had to go a second round towin the governor’s chair.

3 A stability value of 1 means no parties rose above or fell below the thresholdfrom one election to the next. A value of 0.4, for example, means that rising andfalling parties totalled three (two rising and one falling, or two falling and onerising, and so on).

4 $1 � R$2.7; €1 � R$3.5.5 Z-value = 1.98, p > | z | = 0.048. All others showed p-values at 0.47 or higher.6 Z-value = 2.09, p > | z | = 0.037.7 Z-value = –2.16, p > | z | = 0.031.8 2002 seat-shares count 100 percent; 1998, 80 percent; 1994, 60 percent; 1990,

40 percent; 1986, 20 percent.9 PT with Hegemonic: Z-value = –2.01, p > | z | = 0.045. PTB with Unstable

Competitive: Z-value = 2.01, p > | z | = 0.045.10 Z-value = 1.30, p > | z | = 0.192.11 The exception, Roraima, actually had a ratio of 7:1, but the graph has been cut

off at 2 for the sake of comparison.

References

Ames, Barry (1995) ‘Electoral Strategy under Open-List Proportional Representation’,American Journal of Political Science 39: 406–33.

Ames, Barry (2001) The Deadlock of Democracy in Brazil. Ann Arbor, MI: Universityof Michigan Press.

Chhibber, Pradeep and Ken Kollman (1998) ‘Party Aggregation and the Number ofParties in India and the United States’, APSR 92: 329–42.

de Moraes, José Filomeno, Filho (1997) ‘Ceará: O subsistema partidário e o retornoao multipartidismo’, in Olavo Brasil de Lima Júnior (ed.) O Sistema PartidárioBrasileiro: Diversidade e tendencies 1982–94, pp. 35–72. Rio de Janeiro: EditoraFundação Getulio Vargas.

Desposato, Scott (2001) Institutional Theories, Societal Realities, and Party Politicsin Brazil (PhD Dissertation), pdf, UCLA.

PA RT Y P O L I T I C S 1 5 ( 3 )

354



Downs, Anthony (1957) An Economic Theory of Democracy. New York: Harper.Figueiredo, A. C. and Limongi, F. (2000) ‘Presidential Power, Legislative Organiz-

ation, and Party Behavior in Brazil’, Comparative Politics 32: 151–70.Hagopian, Frances (1996) Traditional Politics and Regime Change in Brazil.

Cambridge: Cambridge University Press.Instituto Brasileiro da Geografia e Estatística (2001). Available at: http://www.ibge.

com.br/home/estatistica/economia/contasregionais/tabela03.pdf.Instituto Universitário de Pesquisas de Rio de Janeiro. Available at: http://www.

iuperj.br/deb/ing/Cap3/Cap3_tab1.htm.Kitschelt, Herbert (2000) ‘Linkages between Citizens and Politicians in Democratic

Polities’, Comparative Political Studies 33: 845–79.Laakso, Markku and Rein Taagepera (1979) ‘Effective Number of Parties: A Measure

with Application to West Europe’, Comparative Political Studies 12: 3–27.Lipset, Seymour Martin and Stein Rokkan (1967) ‘Cleavage Structures, Party Systems,

and Voter Alignments: An Introduction’, in Seymour Martin Lipset and SteinRokkan (eds) Party Systems and Voter Alignments: Cross-National Perspectives,pp. 1–64. New York: Free Press.

Mainwaring, Scott (1991) ‘Politicians, Parties, and Electoral Systems: Brazil inComparative Perspective’, Comparative Politics 24: 21–43.

Mainwaring, Scott (1999) Rethinking Party Systems in the Third Wave of Democ-ratization. Stanford, CA: Stanford University Press.

Mainwaring, Scott and Aníbal Pérez-Liñan (1997) ‘Party Discipline in the BrazilianConstitutional Congress’, Legislative Studies Quarterly 22: 453–83.

Mainwaring, Scott and Matthew Soberg Shugart (1997) ‘Conclusion: Presidentialismand the Party System’, in Scott Mainwaring and Matthew Soberg Shugart (eds)Presidentialism and Democracy in Latin America, pp. 394–439. Cambridge andNew York: Cambridge University Press.

Partido Progressista ‘Conheça o PP > Nossa História’. Available at: http://www.pp.org.br/.

Power, Timothy (2000) The Political Right in Post-Authoritarian Brazil. UniversityPark, PA: Pennsylvania State University Press.

Schlesinger, Joseph (1984) ‘On the Theory of Party Organization’, Journal of Politics46: 369–400.

Transparência Brasil (2003) ‘Second Survey on Vote Buying in Brazilian Elections’.Available at: http://www.transparencia.org.br/english.html.

Tribunal Regional Eleitoral do Ceará, Mato Grosso do Sul, Santa Catarina. Availableat: http://www.tre-ce.gov.br, http://www.tre-ms.gov.br, http://www.tre-sc.gov.br.

Tribunal Superior Eleitoral (2002) ‘Divulgação dos Dados dos Candidatos’. Availableat: http://www.tse.gov.br/eleicoes/eleicoes2002/index.html.

DANIEL J. EPSTEIN received his PhD in Political Science in 2008 from the Depart-ment of Government at Harvard University in Cambridge, Massachusetts. He iscurrently a Post-Doctoral Fellow at the Skalny Center for Polish and Central Euro-pean Studies at the University of Rochester, in Rochester, New York. [email: [email protected]]

Paper submitted 28 August 2008; accepted for publication 27 November 2008.


355



Epstein (2009) Clientelism vs Ideology

Documents

largest party

clientelistic party

splinter party

party cohesion

party competition acrossstates

statelevel party systems

party clans exercise

informational role parties