A Combined Utility Theory of Political Linkage
How can political parties simultaneously compete in ideological and
clientelistic electoral markets? The demand side perspective of political competition
suggests that public electoral markets require programmatic parties, while private
electoral markets call for clientelistic parties (Desposato 2001). However, variation
among and within political systems shows that ideology and clientelism are
encompassing features of both developed and developing societies (Kitschelt 2000,
Brinkerhoff and Goldsmith 2002), which indicates that parties necessitate strategies
to overcome this seemingly ambig uous situation in order to win national-wide
competitive elections.
In this respect, developing democracies have experienced the
advancements of programmatic parties (Hagopian, Gervasoni, and Moraes 2009,
Rosas 2005) although that does not preclude the existence of their clientelistic
counterparts (Montero 2010). The puzzle is: how can these antagonistic forces
coexist in the same political system? The case of Brazil is an interesting example of
this development. Since 1994, two programmatic parties (Workers’ Party – PT and
Brazilian Social Democratic Party – PSDB) have polarized national elections and
leftist parties have made inroads in traditional clientelistic bastions once held by the
now called Democrats – D EM (formerly Liberal Front Party – PFL). Yet, these
clientelistic parties remain as active regional political forces and partners of more
programmatic parties nationally.
At the subnational level, two states stand out as intriguing, Bahia and Rio
Grande do Norte, for having undergone shifts in the pattern of political competition
once dominated by traditional parties on the right. The incursion of the left in Rio
Grande do Norte started earlier in 2002 and lasted two consecutive terms until the
right regained power in 2010, which they have held since. In Bahia, on the other
hand, the left only gained power in 2006 and was re-elected in 2010 and 2014.
These two states are representative of a wave that brought to power more
programmatic, mostly leftist, parties in a region that has historically been
characterized by the dominance of oligarchic groups linked to clientelistic parties,
usually on the right of the political spectrum (Montero 2010, Herrmann 2014). Since
1988, virtually all states in this region have had more programatic parties replacing
clientelistic ones. For instance in Maranha o, the PDT won the 2006 election beating
the Sarney machine for the first time since 1960. However, this election was later
judicially overturned on the basis of political power abuse (Callucci 2009, April 1). In
other examples, the PT won in Piauí in 2002 and Sergipe in 2006; PSB won in
Alagoas in 1998 and in Pernambuco in 2006, both replacing the PMDB; PSB also won
in Ceara this same year defeating PSDB; and finally, PSDB won in Paraí ba in 2006
replacing the PMDB.
Scholars of Brazilian politics are divided between those who claim that
clientelism has declined as a consequence of systemic change (Borges 2011,
Hagopian et. al. 2009, Hunter and Power 2007) or institutional factors (Borges
2011), and those who suggest that clientelism is still a very encompassing element
in politics (Avelino Filho 1994) and rightist parties, in particular (Montero 2010).
Hunter and Power (2007), for instance, credit the success of the incumbent
president Lula in the 2006 election to systemic changes in the Brazilian environment
based on three interrelated explanations. Complementarily, Borges (2011) looks into
the interaction between presidential and state elections to explain the decay of the
local political bosses and, for that matter, the emergence of leftist governors in the
least developed states. By the same token, Hagopian et al. (2009) claim that
structural changes in Brazil have forced parties to switch their electoral
strategies from patronage to programmatic linkages.
Conversely, Montero (2010) refutes the systemic approach by arguing that
leftist victories in the Northeast are a consequence of localized strategies of
partybuilding and state-level elite alliance-formation. In this sense, Montero (2010)
claims that governors can still take advantage of the extant clientelistic network.
Research to date, however, does not explain how parties strategically
operate on a two-dimensional political spectrum when competing electorally. Rather,
ideology and clientelism are confounded in a one-dimensional political linkage. This
dissertation aims to integrate the two levels of political competition. It does so by
treating ideology and clientelism as two interdependent forces interacting through
electoral coalitions among parties in the process of shaping
electoral competition.
Moreover, the extant literature on electoral coalitions deals with cabinet
formation under parliamentarism (Martin and Stevenson 2001); presidentialism
(Amorim Neto 2002, 2006, Cheibub 2002, Geddes 1994); comparative studies of
presidential and parliamentary regimes (Cheibub, Przeworski, and Saiegh 2004);
and electoral coalitions under presidentialism in developed countries (Golder 2005)
in the Americas (Amorim Neto 2006), and in Brazil (Amorim Neto 2002, Machado
2009, Mignozzetti, Galdino, and Bernabel 2011, Nicolau 1996). The
theoreticallyderived hypotheses are drawn from considerations about size, ideology,
and institutional incentives and constraints (Martin and Stevenson 2001). However,
clientelism is not included as an independent explanatory variable leading to
coalition formation. I argue that political parties coordinate their access to public
and private electoral markets by forming electoral coalitions that allow them to
garner votes in clientelistic markets while preserving their ideological party brand.
This mechanism is described below by the combined utility theory of political
linkage.
Formal democratic institutions have been the focus of comparative research
following the most recent wave of democratic transitions. This institutionalist
approach, however, is limited to the extent that socio-economic factors and elites’
strategic operation also contribute to changes in patterns of electoral competition.
The recent rise of more programmatic parties in developing countries indicates
important shifts in national voting patterns and has reignited a longstanding debate
between the emergence of the ideological coherence of parties and the electorate
versus the persistence of clientelistic relations. The objective of this dissertation is to
investigate changes in party competition in developing democracies based on the
political dimensions of ideology (left-rig ht) and clientelism (programmatic-
patronage). I define traditional political linkages as those where electors exchange
votes for individual material benefit (Kitschelt & Wilkinson 2007) and use Brazil as a
case study.
Previous studies tend to consider ideology and clientelism as constituting
two dimensions at opposite poles (e.g. Epstein 2009, Hagopian et. al. 2009, Montero
2010), suggesting that the expansion of one dimension determines the suppression
of the other. This approach however does not explain why more programmatic
parties appear to advance in clientelistic or private electoral markets. To what extent
do inroads by programmatic parties reflect an advan ce against clientelistic bastions,
in their own right? Or do programmatic parties now utilize clientelism to advance
electorally? This research investigates whether shifts in the pattern of political
competition result from structural changes in the electoral market (e.g.
modernization), elite strategies, or perhaps a combination of both. In other words,
has modernization changed the pattern of electoral competition or is patronage still
pervasive in Brazil? This project adjudicates between those two views of electoral
competition by developing a theoretical framework that allows for the concomitant
test of both ideological and clientelistic strategies.
The dissertation develops and tests a combined utility theory, which
proposes that polarized competitive elections in modernizing national electoral
markets compel programmatic parties to coalesce with clientelistic parties to gain
access to regional private electoral markets. Socio-economic changes that bring
about modernization create a public electoral market, which demands
programmatic-oriented representation where patterns of political competition are
based on ideological linkages. However, uneven social-economic change can
perpetuate private electoral markets, which demands patronage-oriented
representation based on patterns of political competition derived by clientelism.
The implication of this theory is that programmatic parties will compete
mainly in public markets, whereas clientelistic parties will compete mainly in private
markets. The relative size of each market determines the dominant pattern of
political competition. If the public market is larger than the private one, than
programmatic parties will polarize electoral competition. Conversely, if the private
market is larger than the public one, then clientelistic parties will polarize electoral
competitions. This further implies that successful patterns of electoral competition
in polarized competitive democratic systems will result from a party’s ability to
recruit additional votes from the dimension on which it is weaker.
Thus if the electoral market is predominantly public, programmatic parties
will try to recruit additional votes in the private electoral market. Since
programmatically dominant parties that engage in a direct clientelistic strategy risk
alienating ideological voters, they will develop strategic electoral coalitions with
clientelistic parties as an indirect strategy to increase their vote share. In exchange,
clientelistic parties are provided access to resources, which can preserve their
existence as the weaker broker. The combined utility theory which considers
structural changes (modernization) and elites’ strategic use of both formal
(coalition) and informal (personal vote) institutions could explain why
programmatic parties are advancing in areas once considered party machine
domains.
This dissertation contributes to the literature on party systems and
clientelism in developing democracies. Political parties are essential components of
democratic regimes: they organize participation, by creating shortcuts to
information according to policy preferences, and structure inclusion, by mediating
those preferences and turning them into policy outcomes. Political parties regularize
democratic practices: they establish the “rules of the game,” making it possible to
transform political conflict in policy outcomes. This research also helps fill the gap in
current scholarship by integrating the di mension of clientelism and ideology as an
elite-centered strategic approach. The advantage of this approach is that elites tend
adapt to existing conditions to achieve an optimum outcome. When faced with
structural changes and constrained by formal institutional rules, elites will use
informal institutions to advance their interests, i.e., gaining and maintaining office. In
this sense, patterns of political party competition should result from the
interaction of economic development, formal rules, and societal characteristics.
Unequal cross-national economic development has historically linked
parties to constituents according to developmental levels in specific regions
(Kitschelt and Wilkinson 2007, Nyblade and Reed 2008, Piattoni 2001);
programmatic parties tend to thrive in more developed regions, while party
machines fare better in bailiwicks (Desposato 2001). In order to explain patterns of
political competition, programmatic parties are defined electorally as those whose
votes are mainly prospected from public goods markets in which the electorate is
more sensitive to ideological appeals. Clientelistic parties are defined by their ability
to prospect their votes from markets in which voters prefer private goods and,
therefore, are less sensitive to ideological appeals. This definition underscores the
electoral dimension of party behavior in contrast to understanding programmatic
parties by their legislative performance. One implication of voters’ preferences for
private or public goods is that politicians have incentives either to cultivate personal
votes through clientelistic linkages or to engender party labels through ideological
appeals. Similarly, an impli cation of changing utility theory is that structural changes
make discretionary resources scarce, creating incentives for programmatic politics.
Neither approach explains why programmatic parties have been able to score
significant victories in less developed areas of Brazil typically associated with
private goods markets.
Methodologically, this dissertation builds on well-established literature on
mixed-method approaches by combining subnational quantitative and comparative
case study analyses to test the combined utility theory on political linkages
strategies adopted by political parties in developing democracies that rely on both
ideological and clientelistic relations to compete in elections.
Brazil represents a crucial test case for this theory (Eckstein 1975, George
and Bennett 2005, Gerring 2007). The country has historically had unbalanced
socio-economic development among the different sub-regions and has had both
programmatic and clientelistic parties (s ee figure 1.1). At the national level, two
programmatic parties have polarized elections since 1994 that have shaped political
competition at the subnational level. For instance, Brazil has experienced a shift in
electoral competition that has undermined oligarchic forces across the country.
Particularly notable are the decline of regional political bosses and the growth of
programmatic parties across the political spectrum, including in the less developed
Northeast region that will be the particular focus of this cross-regional analysis.
Figure 1. 1: Dimensions of Political Competition
The quantitative analysis aims to test the validity of the argument of
ideological (Chapter 2) and clientelistic (Chapter 3) political linkages, looking at
each independently as well as looking at their interactions by means of coalitions
(Chapter 4). In Chapter 2, I use an existing measure of party ideological placement to
test party differentiation based on spatial distribution. In Chapter 3, I devise and test
the validity of a new quantitative measure of clientelism, based on geographic
distribution of votes depicting electoral market characteristics. This measure is the
result of sophisticated computational iterations that run for months at a time. This
approach also contributes to the analytical and methodological refinement of the
concept of clientelism as a systematic political linkage, which has remained elusive
to measurement in spite of its broad application in the fields of anthropology,
sociology, education, public policy, and political science. In chapter 4 I test the effect
of both ideology and clientelism on coalition formation.
Field research in Brazil’s Northeast helps uncover the ways in which
parties and political elites form strategic coalitions with both ideology and
clientelism in mind, in order to increase their electoral resources. For instance, the
case studies considered in this project (Chapter 5) highlight transitions from
dominant political machines to more programmatic party leadership. In the state of
Bahia, Anto nio Carlos Magalha es (ACM) led a political machine that, with the
exception of the 1986 election, dominated politics from when he was appointed its
governor in early 1970s by the military regime until 2006 when the Workers’ Party
won the election outright in the first round. Similarly, in the state of Rio Grande do
Norte, the Alves and Maia families’ political machines took turns in power until their
defeat by the Brazilian Socialist Party in 2002. Case studies contrast with statistical
modeling in that they provide: conceptual validity due to the difficulty in measuring
some important concepts; development of a new hypothesis due to the open-ended
nature of the qualitative question; exploration of causal mechanisms to
accommodate the in-depth investigation of contextual factors; and allow for causal
complexity (George and Bennett 2005). Moreover, while within case studies control
for an array of institutional variables, they still constrain the variance of societal
variables that could be captured in cross-national designs (Desposato 2001).
However, variations indeed happen within the country and to capture them we
resort to a most similar cases approach (Przeworski and Teune 1979).
To preview this dissertation’s main findings, the analysis suggests that both
ideology and clientelism are relevant independent dimensions of political
competition in developing competitive democracies. In this sense, the validity of
ideology is confirmed by gubernatorial runoff election models, which predict the
migration of party votes between the first and second round of election according to
the principle of party ideological proximity. Similarly, the validity of clientelism as a
second dimension of political linkage is partially confirmed by the models that 1)
measure the clientelistic nature of the political parties based on the geographical
dispersion (clientelistic) and clustering (programmatic) of the votes they received
and 2) compare it with the nature of the electoral market in terms of the size of the
local public sector (the larger, the more clientelistic) as well as the private sector
(the larger, the more programmatic). The interaction of these two dimensions
through coalitions, however, is more complex. In this sense, the model suggests that
ideology appear to be more important a dimension than clientelism. However, it
seems that even ideology has become less decisive a factor in determining electoral
coalitions over time. This conclusion is supported by qualitative interviews with
party elites. The findings about clientelism are not conclusive as there is no
discernible pattern of electoral coalition formation at this level, which indicates that
parties form electoral coalitions haphazardly.
This dissertation is organized as follows. Chapter 2 evaluates the validity of
the general argument that party systems in developing democracies are inchoate
and show low ideological distinction as a function of the low level of party identity
against the competing argument that parties are becoming more programmatic. I
apply the general theory of spatial voting distribution, which assumes that voters
choose parties according to ideological proximity, to adjudicate issues related to
ideological differentiation. Further, I propose two approaches that draw both on
aggregate and individual-level data at the subnational level. The first assesses the
ideological distribution of votes between the first and second round in gubernatorial
runoff elections. The second adds another level of validity by analyzing individu al-
level preferences. The analysis clarifies the character of ideological differentiation
and declining clientelism in Brazil.
Chapter 3 deals with clientelistic politics as a second level of political
competition. To unveil the electoral strategies employed by parties when competing
in private electoral marketplaces, this chapter proposes a model of geospatial
dominance and dispers ion of party votes for the Brazilian Chamber of Deputies by
applying the Herfindahl Index, a general measure of market concentration by
companies, as a proxy for clientelism and to assess the evolution of competition
among parties. Further, I refine this assessment by using geospatial measures of
clustering and dispersion of party vote distribution.
Chapter 4 integrates ideology and clientelism to assess the logic of coalition
formation. It applies the combined utility theory developed in this study to explain
party electoral competition dynamics through oversized ideologically heterogeneous
electoral coalitions and, consequently, party development and electoral outcomes in
multiparty presidential regimes in developing democracies.
Chapter 5 fleshes out the methodological findings and provides conceptual
validity of the study. The analysis draws on case studies of two states in Northeast
Brazil, Rio Grande do Norte and Bahia, which have experienced political changes that
highlight national shifts in electoral competitions. Rio Grande do Norte had been the
political domain of two oligarchies (Maia and Alves), which most often competed
against each other but also cooperated with one another to control the state
apparatus. Bahia was the bastion of the Anto nio Carlos Magalha es (ACM), who more
singlehandedly controlled state power. Power shifts that brought more
programmatic parties to power in these states justify their selection as case studies.
Chapter 6 concludes by highlighting the independent effects of ideology and
clientelism on party competition and the interaction of these two levels by means of
electoral coalitions. It also offers suggestions for future research
CHAPTER 2 Literature Review
In this section I discuss previous findings related to the operation of ideology,
clientelism, and electoral coalitions on party competition. Aiming to answer the
puzzle of how parties strategically and concomitantly access public and private
markets in order to win competitive polarized elections in developing democracies, I
propose a combined utility theory, which suggests that programmatic parties will
form electoral coalitions with clientelistic parties to garner votes in private markets
while preserving their ideological brand. In the following paragraphs I review the
literature, present the theory, and propose the independent and combined effects of
ideology and clientelism on party competition by means of electoral coalitions.
Characteristics of the electoral market determine salient dimensions under
which parties will evolve and, as a consequence, define the pattern of party
competition. Whether and what type of societal cleavages emerge varies by country
and region, including distinctions between the left versus right (e.g. Power and
Zucco 2009), secular versus Christian (e.g. Coppedge 1997), xenophobic
nationalistethnic versus regional separatist (e.g. Horowitz and Browne 2005), and
liberal versus conservative (e.g. Huber and Inglehart, 1995). More recent studies on
Brazilian politics have considered the difference between programmatic and
clientelistic politics as an important element in defining party development
(Desposato 2001, Epstein 2009, Hagopian et al. 2009) as well as change in party
competition (Montero 2010, Zucco 2008).
One limitation found in much of the literature on electoral markets and
partisan strategies is that scholars tend to conflate programmatic parties with left
parties and clientelistic parties with right parties. A better understanding of the
pattern of party competition requires a conceptual distinction between ideology and
programmatic/clientelistic dimensions, as parties may respond strategically to the
independent demands imposed by the relative salience of each. In this sense, I will
discuss their independent effects and analyze how this interaction explains patterns
of political competition in Brazil. I break down this discussion into three sections as
follows: the first section deals with ideology, the second discusses clientelism, and
the third will consider the concomitant operation of ideology and clientelism in
electoral coalition formation.
Ideology
Political parties’ brands provide inform al short cuts for voters on the
policies they represent. Parties differentiate themselves based on the policies they
pursue between elections; during the election they hope to be rewarded with votes
from constituents who believe in those same policies. Thus, policies link parties and
voters conventionally positioned on a left-right political spectrum, which defines the
horizontal axis of political competition. Electoral ideological competition, however,
suggests a sufficiently differentiated party system. In this sense, it is necessary to
understand the extent to which Brazilian parties are distinguished among
themselves. This information is crucial in determining this aspect of political
competition and, therefore, the ideological distribution of votes among parties.
Ideology is the most significant framework based on which electoral competition
takes place. This is so because ideology reduces the informational cost of both
parties, which invest resources to publicize their programs, and of voters, who invest
time to learn about the parties’ programs (Downs 1957). Conversely, personal vote
corresponds with a secondary dimension that determines the marginal distribution
of votes, conventionally placed on a vertical axis of political competition. This arena
of clientelistic competition is the main focus of this study. However, since parties, to a
lesser or greater extent, compete in both dimensions it is necessary to understand
these axes equally, as these strategies are not isolated
from each other but can be complementary.
The literature on ideology mainly offers two competing perspectives that
dispute the ideological strength or weakness of the Brazilian party system. These
perspectives can be further divided along party- and voter-centered approaches. On
one hand, party-centered scholars believe Brazilian parties are ideologically
different along the political spectrum and show some degree of ideological
coherence (Power and Zucco 2009). Scholars argue that the party system has been
increasingly stable and have found reduced electoral volatility (Braga 2006, 2007,
Lyne 2005, Santos 2006). Non-leftist parties are also thought to have set down
deeper roots in society and have evolved away from clientelistic and largely
nonideological roots (Hagopian et. al. 2009). Conversely, other party-centered
studies contend that Brazil shows a low degree of ideological organization (Rosas
2005) and, more specifically, that Brazilian parties are divided between the Workers’
Party versus the rest and that the remaining parties do not show ideological
differentiation, which has rendered the system relatively more inchoate (Lucas and
Samuels 2010).
The voter-centered literature tends to agree with the latter perspective (i.e.
low degree of ideological differentiation). Scholars argue that only leftist voting
results from strong ideological commitment (Holzhacker and Balbachevsky 2007)
and that the overall system shows high levels of personalism (Almeida 2007, Ames
2001, Kinzo 2005, Mainwaring 1999, Nicolau 2006) and generally low-levels of
party sympathy (Almeida 2007, Ames 2001, Ames and Smith 2010, Kinzo 2005,
Mainwaring 1999, Mettenheim 1995, Nicolau 2006, Renno 2000, Rua 1997).
These approaches, nonetheless, focus on either elite or public opinion data
based on a unidirectional relationship between voters and parties. In doing so,
scholars have not considered the strategic interplay between parties and voters that
individual-level data may overlook. For instance, public opinion data generally
surveys broad geographic areas and as a result may not account for local effects of
party organization (Ames 1994). This study posits that parties matter electorally
and their brands bear some strategic value both to the leaders and to the voters.
Hence, an understanding of the ideological linkage between party and voters
remains crucial to assess party competition in Brazil.
Clientelism
In contrast to politics that are dominated by ideological differentiation that
prevails among well-developed democracies, clientelism is often viewed either as a
characteristic of traditional societies or as a political strategy in settings where some
level of democratic advancement is observed (Avelino Filho 1994, Kitshelt and
Wilkinson 2007, Nichter 2010). Often framed as a constitutive element of traditional
societies, clientelistic politics is marked by the asymmetric, stable, and directly
personal relationship between patron and client (Scott 1972); the party system is
constituted by cadre parties and the economic resources used in the exchange are
provided privately by the patron (Graziano 1973, Scott 1969).
In contrast, the political strategic component of clientelism is manifest in
democratic institutional settings where the symmetric, intermittent,
instrumentalrational and broker-mediated exchange relationship prevails (Scott
1969, 1972, Weingrod 1968). The electoral nature of clientelism represents both a
passage from a traditional to a transitional and, eventually, developed society
(Avelino Filho 1994) and an evolution from the strictly local to a national level of
hierarchical political machines (Scott 1968). Thus, electoral clientelism is defined as
exchanges of votes from citizens for private goods from politicians (Kitshelt and
Wilkinson 2007). The degree to which clientelism is prevalent influences the
strategy of a party that can even choose whether or not to adopt clientelism
(Kitschelt 2000). Some parties, such as the Argentine Peronist, indeed chose this
path by shifting from labor-based to clientelistic politics in order to proceed with
economic reforms without alienating the working and lower classes (Levitsky 2003).
Recent studies on Brazilian politics have divided scholars between those
who view clientelism as being on the decline or as persisting. Those who view a
decline tend to identify systemic programmatic policy (Borges 2011, Hagopian et. al.
2009, Hunter and Power 2007) or institutional constraints (Borges 2011) as reasons
for this trend. Others argue that clienteli sm is still an encompassing element of
Brazilian politics, in general, (Avelino Filho 1994) and within rightist parties, in
particular (Montero 2010).
Hunter and Power (2007) credit the success of the incumbent president
Lula in the 2006 election to systemic changes in the Brazilian environment based on
three interrelated explanations: first, an economic retrospective vote based on the
positive economic performance that especially benefited the lower classes; second,
Lula’s coattail effect due to his popularity as a consequence of his social and
economic policies; and third, the very effect of his conditional cash transfer program.
Complementarily, Borges (2011) looks into the interaction between presidential and
state elections to explain the decay of the local political bosses and, for that matter,
the emergence of leftist governors in the least developed states. While Hunter and
Power (2007) find no correlation between the growth of Lula’s vote share in less
developed areas and the growth of his own party’s vote share for the Chamber of
Deputies, Borges (2011) finds a presidential coattail effect on gubernatorial
elections, which support the claim of interdependence between
national and state elections.
By the same token, Hagopian et al. (2009) claim that structural changes in
Brazil have forced parties to switch their electoral strategies from patronage to
programmatic linkages. According to Hagopian and colleagues, economic reforms
have limited access to federal and state government discretionary and
patronagebased spending thus encouraging politicians to become more party-
oriented as a strategy to develop their party brands and organize their electoral
competition. Their study innovates methodologically by moving away from the
conventional assessment of party behavior based on floor voting records and party
switching. They do so by incorporating the basis of the deputies’ campaign strategies
and attitudes that reflect the empowerment of party leaders to the detriment of
their own individual power (Hagopian et. at. 2009).
Among the authors who see the decline of clientelism, there is general
agreement that this is occurring while party machines are also on the decline. Where
the analysis differs is with respect to why and how this has happened. Hunter and
Power (2007) and Borges (2011) approach the problem from the demand side of the
electoral market and consider a left-rig ht dimension of interaction between the
actors. Hagopian et al. (2009), in turn, approach the problem from the supply side of
party development that operates on a programmatic-clientelistic dimension. The
electoral demand side appears to offer a better framework to explain the
development of parties, as it seems unlikely that parties would unilaterally adopt an
evolving strategy that does not reflect the changes in the electoral market itself. On
the other hand, the left-right dimension may not be the most appropriate approach
to explain competition on clientelistic market, as it would require an assumption
that clientelistic politics is determined by party’s position on the political spectrum.
Contrary to those who argue clientelism is on the decline, Montero (2010)
argues that leftist inroads into the poorest areas of the Northeast, the grotões, are
the result of localized strategies of party-building and state-level elite
alliancefor mation. Montero’s (2010) rationale is threefold: one, that the
federalization of social policies and the corresponding municipalization of its
implementation do not prevent governors from cultivating actual clientelistic
networks; two, that anyone can claim credit for the success of economic and social
policies due to the incapacity of the less educated population to differentiate the
source of these policies; and, three, that the systemic explanation fails to indicate
why the decline of clientelism would benefit the left at the subnational level. B ased
on that, Montero (2010) claims that governors can still take advantage of the extant
clientelistic network even when
the economic and social policies are not intermediated by them.
The problem with Monetro’s (2010) rationale is that he assumes the
electorate is not sophisticated enough to identify the originator of these policies,
thus allowing incumbent mayors and governors of any political persuasion to
creditclaim their success. In fact, Montero (2010) cites a study conducted in Recife,
Pernambuco by Figueiredo and Hidalgo (2009) who indicate that beneficiaries of the
Bolsa Família link this program to Lula, to whom they show a positive attitude, but
they do not extend their support to Lula’s co-partisans or party. This seems to
contradict his view that the level of differentiation of the less educated voters, who
tend to be the target of cash transfer programs, is low. Moreover, focus group
interviews conducted in Bahia and Pernambuco suggest that the poor are indeed
sophisticated in their knowledge of the origins of social programs; recipients of the
Bolsa Família associate this policy to Lula or to the federal government , not local
mayors and governors (Sugiyama and Hunter 2013).
Montero (2010) is correct in his observation that systemic modernization
explanations fail to indicate the mechanisms through which the decline of
clientelism would benefit the leftists at the subnational level, but not for the reasons
he indicates. In this sense, his view assumes that competition takes place on a
leftright dimension, thus ignoring the potential that all parties can engage in
clientelism. If parties’ competition is organized on an ideological (left-right) scale,
then how are they competing in clientelistic electoral markets? Moreover, Montero
(2010) rightly indicates that we should look into party-building and state-level elite
alliance-formation to understand the shift in politics in the Northeast. However, the
ideological dimension of party formation is insufficient to explain allianceformation,
which also may respond to national hierarchical forces. In fact, this is equivalent to
Borges’ (2011) institutional approach to the problem, except that he also fails to
establish the causal mechanisms that integrate the subnational and national levels of
government. In sum, both the systemic and the strategic approaches fall short of
integrating the national and state levels of Brazilian politics.
Understanding clientelistic politics is essential to explaining party building
strategies and electoral outcomes. It is unlikely that clientelism alone however could
account for patterns of all political competition in modernizing societies. In thi s
sense, for Avelino Filho (1994), the understanding of the passage from a clientelistic
political system to a more institutionalized one requires moving away from a
dichotomous analysis of these two systems to an analysis that is focused on their
points of intersection, i.e. the ability of the patrons to manipulate client’s hope for
future benefits.
I agree with those premises, but while the mechanisms by which this
intersection occurs to form a conti nuum between clientelistic and universalisti c
practices is useful to study transformations at the system’s level, it is problematic for
the understanding of party development and party electoral advancement. The
contention resides in the assumption that clientelism perpetuates itself by
eliminating competitiveness through its capacity to manipulate citizens’ hope for
future benefits (Chubb 1981 cited by Avelino Filho 1994). In order to capitalize on
future benefits, clientelistic networks have to rely on personal trust between patron
and client. However, this kind of relationship is the main characteristic of traditional
clientelism as opposed to the more mercantile characteristic of modern clientelism
expressed in democratic institutional settings.
Accordingly, democratic electoral competition improves the client’s
bargain power (Piattoni 2001) by offering options at the clientelistic level as well as
providing the client with an exit option (Kitshelt and Wilkinson 2007) by voting
along ideological preferences. Thus, it appears that some authors confound practices
of the old (relational) and new (electoral) forms of clientelism, which they claim to
distinguish. By this token, recent studies on clientelism have helped to advance the
literature by explaining how clientelistic politics operate.
In this view, Nichter (2010) differentiates the strategies associated with
clientelism as relational and electoral. By focusing on the latter, he theorizes about
the dual commitment problem, which derives from the interaction between elites
and clients. Nichter (2009) further distinguish between prospective and
retrospective clientelism. Prospective clientelism is related to the credibility of the
promises made by both elites and clients in terms of their likelihood to be fulfilled
after the elections, thus the dual commitment problem. Retrospective clientelism, in
turn, refers to the credibility of the citizen, but not of the elite, because the citizen
has already received the benefit before the elections. This approach is an advance on
Avelino Filho’s (1994) study, which misses the retrospective aspect of electoral
clientelism. On the other hand, Nichter (2009) misses the point that the co-existence
of prospective clientelism, with more developed democratic electoral institutions, is
only possible under a system of moral control that is proper to the relational kind of
clientelism. In other words, those authors appear to overstate the importance of the
traditional form of clientelism over the more modern manifestation of this
phenomenon. The extent to which promises of future benefit influence behavior
implies that the promise itself becomes a private good. For the purpose of this study,
it is more important to establish the degree to which clientelism is prevalent than to
consider the nuances of the operation of the clientelistic relations.
Furthermore, Nichter (2010) identifies the different types of electoral
clientelism (vote buying or turnout buying during the electoral period) and the
mechanisms through which it works, but he does not address the payoffs resulting
from the adoption of such political strategy. Moreover, by focusing on the strictly
local aspect of clientelism, as does Montero (2010), Nichter fails to explain how
national and sub-national politics are linked strategically.
In addition, because the model assumes the existence of only one political
machine (Nichter 2010), competition among political machines, which has
implications for the relationship between patrons and clients (Caciagli and Belloni
1981), is not considered in his study. There are three important implications of this
competition: one is related to the increasing cost of keeping the clientele; the second
refers to the unsustainability of the patrons-client relationship, even though this
competition is responsible for the change from a strict relational to a state-party
hegemonic clientelistic relationship (Caciagli and Belloni 1981); and the third
implication of the competition between local bosses is related to the decreasing cost
of the central government to enlist the cooperation of local bosses that no longer
monopolize the clientelistic relations in a given bailiwick. An opposite problem is
present in Borges (2011), who acknowledges the competition among political
bosses but does not consider competition at the national level.
A better understanding of the dynamics of political competition requires a
theory that can account for party building and the pattern of political competition
overall. Thus, it is necessary to consider the integration of different levels (national
and state) and dimensions (ideological and clientelistic) in which this competition
takes place. It is possible that the integration of these levels and dimensions occurs
by means of electoral coalition practices. This approach, for instance, would
contribute to the understanding of why some parties sacrificed ideological
principles and historic state interparty disputes to coalesce nationally, and why in
some states the national strategy did not supersede local rivalries.
Coalitions
Scholars can integrate analysis of ideological competition alongside
clientelistic practices by focusing on the practice of coalition formation. The
systematic study of coalitions was originally directed toward the understanding of
parliamentary regimes and cumulated with the subsequent expansion of the concept
to include presidential regimes. The main objective of these studies is to explain
cabinet formation given the constraints and incentives inherent to each one of those
government types. Although not completely ignored by the literature (Duverger
1959), until recently little attention had been put on the pre-electoral variant of
coalitions, defined as associations of parties with the objective to stand elections.
The works dedicated to this subject tend to focus on formal institutions and the size
and ideology of the parties, thus missing informal institutions of which clientelistic
politics is an important component (e.g. Golder 2006, Martin and Stevenson 2001).
The consideration of clientelism as a second dimension operating in the political
system could help to explain oversized ideologically heterogeneous electoral
coalitions and, consequently, party development and electoral outcomes in
multiparty presidential regimes in developing democracies.
Pre-electoral coalitions are thought to be important to the extent that they
are utilized, shape policies, are normatively representative (Golder 2005), and
structure cabinet formation (Laver 1998, Strom and Muller 2000). These studies
have been based on both a priori and empirical approaches, which consider the
office- or policy-seeking motivational assumptions of the politicians (Lever 1998,
Martin and Stevenson 2001). A priori approaches seek to build logic models of
government-formation processes whereas empirical approaches examine the
specific variables that account for government formation (Laver 1 998). Those
models have been successfully used to assess cabinet formation in parliamentary
systems (Martin and Stevenson 2001), presidential systems (Amorim Neto 2002,
2006, Cheibub 2002, Geddes 1994), comparative perspectives on both presidential
and parliamentary regimes (Cheibub et al. 2004), and electoral coalitions in
industrial societies (Golder 2005) in the Americas (Amorim Neto 2006) and in
Brazil (Amorim Neto 2002, Machado 2009, Mignozzetti et. al. 2011, Nicolau 1996).
Theoretically derived hypotheses are drawn from considerations about size,
ideology, and institutional incentives and constrai nts (Martin and Stevenson 2001).
Analysts of coalition formation are highly influenced by the work of Riker
(1962), who used a rationalist approach to argue that party size creates incentives to
form a minimum winning coalition, where actors seek to distribute gains among the
smallest number of participants. Riker’s (1962) assumption is a refinement of t he
concept of minimal winning coalition, which ceases to be winning when any one
member of this coalition is subtracted (Von Neumann and Morgenstern 1944). By
considering a minimum winning coalition as a subset of the minimal winning
coalition, Riker reduces the number of parties to predict coalition formation.
However, he does not consider the role of ideology, which is further specified by the
analogous assumption that it is easier to form a coalition of a small number of
parties than it is to form a coalition of ideologically proximate parties.
The role of ideology in coalition formation is the obvious focus of the
policyseeking approach but it is also retained implicitly in office-seeking approaches,
which consider policy proximity an important element of bargaining. Other scholars
have elaborated in this vein with the concept of a “minimal connected winning
coalition” that is formed along a contiguous ideological dimension but that ceases to
be either winning or connected when a single member leaves the coalition (Laver
and Schofield 1990).
The problem with this approach is that assuming parties as being
officeseeking implies that as parties are invested in the purpose of gaining office,
they will apply the necessary strategies to achieve such an end. In this sense, the
consideration of ideology as the only de terminant dimension limits the possibilities
of combined strategies. In other words, if a party is policy-seeking it will certainly
place a higher value on ideologically contiguous coalitions. However, if a party is
office-seeking it may relax the requirement of ideologically consistent coalitions,
provided they can enter those coalitions without being severely punished by an
ideologically demanding electorate. More generally, assumptions about the goal of
the party as either policy- or office-seeking will certainly fail to account for party
systems transformation over time. For instance, the transformation of a party
system from clientelistic to more programmatic, as Hagopian et al. (2009) claim is
the case in Brazil, implies that parties are placing more value on policies. Conversely,
the transformation of parties from policy- to a more office-seeking strategy, which is
the case of the Brazilian Workers’ Party (Hunter 2007, 2010, Samuels 2004),
suggests that a more balanced approach that considers the influence of both
strategies is more conducive to an understanding of the dynamics involving
coalitions (Strom 1990).
Scholars who embrace an institutional approach, on the other hand,
emphasize the role that institutions play on the process and outcome of the
coalition-formation, as well as the resulting implications on decision-making after
the government has been formed (Martin and Stevenson 2001). The pre-formation
approach is usually associated with the a priori literature based on formal
bargaining while the post-formation approach relies on spatial voting models and
specific rules that affect the decision making of the already established government.
These models are not self-exclusive in the sense that the use of one approach does
not preclude the application of the other. In fact, formal theory has been used to
generate institutional hypotheses on cabinet formation in presidential regimes
(Amorim Neto 2006) and on pre-electoral coalition formation in developed
parliamentary democracies (Golder 2006, Martin and Stevenson 2001).
Institutional explanations have been particularly importa nt in explaining
cabinet formation in presidential systems in the Americas (Amorim Neto 2006) and
pre-electoral coalition in parliamentary regimes with inferences about
semipresidential and presidential systems (Golder 2006). Institutional hypothe ses
also have been applied in the Brazilian case to explain policy outcome as a result of
coalition formation (Amorim Neto 2002), minimum winning electoral coalition
(Machado 2009), and electoral coalition in proportional elections (Mignozzetti et. al.
2011). The main determinants of electoral coalitions as an outcome variable are:
size of the party – when small and constrained by electoral threshold either formal
or given by district magnitude (Nicolau 1996), parties tend to enter electoral
alliances with other parties to overcome this institutional hurdle (Nicolau 1996,
Soares 1964); maximization of seats (Lima Jr. 1983), which aim at reversing the
effects of party fragmentation produced by proportional systems, thus electoral
uncertainty (Golder 2005, Nicolau 1996); strategic interaction among the different
levels of competition (Borges 2011, Lima Jr. 1983); and, ideological consistency or
the alliances (Amorim Neto 2002, Mignozzetti et. al. 2011, Schmitt 2005, Soares
1964), which is central to this study.
Those approaches, however, present some shortcomings as the ideological
dimension alone can only account for the ideological manifestations of the parties’
preferences and strategies depending on the main goal of the party. While it would
be plausible to predict that a policy-seeking party would gravitate toward parties
that are ideologically consistent, the introduction of office-s eeking strategies would
expand the strategies of the parties to include others that are not so consistent with
their policy prospects. In fact, when the two Brazilian democratic periods are
compared (1946-64 versus post-80s), the most recent democratic period shows a
more consistent ideological pattern of electoral coalitions (Schmitt 2005). However,
over time, the post-80s period have shown a decline in the ideological consistency of
electoral alliances (Carreira o 2006). This trend urges an explanation and it is likely
that clientelistic politics may account for some of the variation of those electoral
coalitions.
The incentives for electoral coalition formation are varied and are also
conditioned by coordination strategies superseded by different governmental levels.
The direction of the coattail effect is often disputed. Some authors have
demonstrated coattail effects of local elections on national elections (Ames 1994,
Samuels 2000), while others have argued the opposite (Brambor, Clark, and Golder
2006). Specifically in regards to electoral coalitions, some authors emphasize the
importance of subnational elections (Machado 2009, Montero 2010), others focus on
the national determinant of electoral coalitions (Borges 2011), and yet some others
stress the interaction of the two levels (Power and Mochel 2008). Borges (2011), for
instance, indicates that the decay of the local political bosses in the least developed
states is due to two interrelated constraints that affect the capacity of political elites
to establish and maintain electoral coalitions: vertical competition (i.e. at the federal
and state levels) and the weakness of the parties.
Focusing on presidential and gubernatorial elections as respectively the
national and state dimensions of voting distributions, Borges (2011) shows the
presidential coattail effects on gubernatorial elections, which supports his claim of
the interdependence of national and state elections. Borges (2011) shows evidence
of the decline of the political bosses and, although an electoral coalition seems to be
the right approach to assess the elite’s strategies, he fails to indicate the pattern of
the electoral coalition and its operation at the national and state level. To be fair, the
author does not ignore the interplay of national and regional strategy, but this
interaction needs to be clearly est ablished by explaining the mechanisms under
which electoral coalitions are formed.
Montero (2011), on the other hand, argues that the subnational pattern of
party-building and elite alliance formation at the state-level explain the vote share
between leftist and conservative in selected federal units. The dynamics of those
alliances result from segmentations either on the leftist or on the conservative side,
thus altering the electoral result that could both displace conservative machines and
bring them back to power. Electoral alliance is certainly a key element in
understanding the electoral dynamics in Brazil, but analysis that consider left versus
conservative parties limit this scope as it implicitly assumes that clientelistic politics
segments parties ideologically with the left appearing programmatic and the
conservative as clientelistic.
Political competition in multiparty democratic developing countries occurs in
a very heterogeneous cross-national electoral market. Thus, the national success of a
party depends on its ability to garner votes in both public and private goods
markets. However, a programmatic party cannot overtly pursue clientelistic politics
because it would risk losing public market voters, hence defeating its purpose of
building a party brand as the most effici ent way of electoral linkage in the public
goods market. By the same token, an expanding programmatic party cannot alienate
clientelistic votes, especially if it has exhausted its universe of expansion in the
public goods market and in view of competition with other programmatic parties
within this same market. Given the expansion of the competitive public goods
market as a consequence of structural changes in developing democracies, the
success of programmatic parties depend on their ability to extract additional votes
from private goods markets at the subnational level. The way to do this without
compromising the party brand is to coalesce with political machines already
established in those private goods markets.
Toward a Utility Theory of Partisan Coalitions
Theories of voter-party linkages have focused either on the demand or on the
supply side of politics. The demand side approach derives from the assumption that
the voter’s relative preference for public or private goods creates different
informational demands on the political system (Desposato 2001). The supply side
approach claims that structural shifts resulted in parties’ decreased access to
patronage-based funds, thus creating an incentive for politicians to develop their
party brands (Hagopian et. al. 2009). Both perspectives contribute to the
understanding of political linkages though they have some shortcomings.
The demand side approach focuses on parties’ legislative behavior, which is
assessed by cohesion in roll-call votes, and considers the private and public electoral
markets as dichotomous extremes. This implies that a party i n the legislature is
more responsive to the type of demand that characterizes each electoral market. If
extended to the electoral arena, this approach would predict the election of either
programmatic parties in public goods markets or clientelistic parties in private
goods markets. In this sense, it falls short of explaining programmatic parties’
victories in predominantly private goods markets as well as clientelistic parties’
successes in predominantly private goods markets. A better explanation of parties’
electoral outcomes, as opposed to legislative behavior, would have to relax the
extreme dichotomous assumption in favor of the possibility that
parties can engage both markets.
The supply side, in turn, considers both the legislative and the electoral levels
of party behavior. However, its assumption of the electoral market as a public goods
market would only predict the election of programmatic parties, thus missing the
clientelistic one. Moreover, analysis of partisan electoral behavior is informed by
surveys of elected deputies’ campaign discourses. If the general perception is that
programmatic behavior is more morally accepted than clientelistic behavior, then
surveys may be biased toward positive responses, which would overpredict
programmatic electoral behavior.
In order to address this gap in the literature that tends to treat ideology and
clientelism as two opposite continuous poles, I propose a combined utility theory,
which bridges the concomitant operation of both ideology and clientelism as salient
processes of political linkages. While the analytical distinction of ideology and
clientelism is useful in studying the static manifestation of those political
phenomena at the systemic level, the comprehension of the dynamics of electoral
competition and party development requires the interaction of those two elements,
each one with its own continuum. In this sense, the pattern of political competition
in multiparty developing democracies operates on two levels: ideological (left-right)
and clientelistic (programmatic-patronage). Socio-economic changes that bring
about modernization create a public electoral market that demands
programmaticoriented representation determining patterns of political competition
on the ideological dimension. However, uneven social-economic changes perpetuate
the existence of a private electoral market that demands patronage-oriented
representation determining pattern of political competition on a clientelistic
dimension. See figure 1.1 in Chapter 1 for a visual illustration of this theory, which
posits that parties have features of both ideological commitments as well as
clientelistic practices.
The implication of this theory is that programmatic parties will compete
mainly but not exclusively in public markets, whereas clientelistic parties will
compete mainly but not exclusively in private markets. The relative size of each
market determines the dominant pattern of political competition. If the public
market is larger relative to the private one, then the electoral competition will be
polarized by programmatic parties. Conversely, if the private market is larger
relative to the publi c one, then the electoral competition will be polarized by
clientelistic parties.
This further implies that successful patterns of electoral competition in
polarized competitive democratic systems will result from the ability of a party to
garner additional votes from the dimension in which a certain party is weaker. Thus
if the electoral market is predominantly public, programmatic parties will try to win
additional votes in the private electoral market. Since programmatic dominant
parties engaging in direct clientelistic strategy risk alienating ideological voters, they
will coalesce with clientelistic parties as an indirect strategy to increase their vote
share. In exchange, clientelistic parties are provided access (or the possibility of
access) to resources that preserve their existence as the weaker broker. The
combined utility theory which considers structural changes (modernization) and
elite’s strategic use of both formal (coalition) and informal (personal vote)
institutions could explain why programmatic parties are advancing in areas that
were once considered party machine domains.
In the following section I will lay out the theory-derived hypotheses on
ideological and clientelistic levels of political linkage and how those dimensions
interact in the process of electoral coalition formation in explaining patterns of
electoral competition.
Ideology
Socio-economic changes that engender modernization create a public
electoral market that, in turn, demands ideologically-oriented political parties. In
this sense, I hypothesize that the Brazilian party system is ideologically
differentiated on a left-right political spectrum. The confirmation of this hypothesis
is a necessary condition to further assess the systemic and clientelistic base of the
electoral competition.
Clientelism
Electoral clientelism implies the coexistence of minimally advanced
democratic institutions with archaic societal characteristics that survives in modern
days as a consequence of uneven socio-economic development. Thus, I hypothesize
that private goods markets would favor clientelistic political parties.
Coalition
Electoral coalitions are the coordination among political parties with the
aim to win elections. The composition of coalitions will depend on the objective of
the parties that join those alliances. In this sense, the assumption that parties are
rational office-seeking entities implies that they are utility maximizers; in other
words, they will pursue the best strategy to gain office. Elites realize their strengths
and limitations and operate strategically by forming coalitions that better support
their political survival and/or expansionist ambitions. In a multiparty political
system with nationally polarized elections among programmatic parties, I expect
that coalitions will form with a heterogeneous composition of programmatic and
clientelistic parties.
In the following sections I undertake the analysis of the independent effects
of ideology (Chapter 3) and clientelism (Chapter 4). Further, I will test the
independent effects of both ideology and clientelism and then establish a link
between them according to the operation of coalitions in explaining patterns of
electoral competition.
CHAPTER 3 Party Ideological Differentiation in Brazil
Have parties in emerging democracies become more programmatic? The
conventional view holds that parties in developing democracies are institutionally
weak, undisciplined, and non-programmatic. The feckless nature of these party
systems, scholars contend, undermine ideological linkages ensuing clientelistic
behavior based on personal vote and individualized electoral strategies. Recent
electoral success of more programmatic parties in developing democracies however
has reignited the debate regarding the consistency of the party system versus the
prevalence of clientelistic relations. A burgeoning literature has recently challenged
conventional wisdom by empirically demonstrating the strengthening of these party
systems (e.g. Hagopian et. al. 2009). This is particularly true for the Brazilian case,
where programmatic parties have made inroads in traditionally oligarchic
strongholds. This chapter purports to evaluate the validity of the general argument
that party systems in Brazil are inchoate and shows low ideological distinction as a
function of the low level of party identity.
Party-centered and vote-centered studies offer competing perspectives about
party differentiation in Brazil. These studies are mostly based on national legislative
and executive elections or on party behavior in the legislature. Recently, there has
been some interest in subnational politics related to party behavior in state
legislatures (Desposato 2001) and gubernatorial influence on the national legislative
agenda (Abru cio 1998, Cheibub, Figueiredo, and Limongi 2009, Desposato 2004,
Samuels 2000). However, little attention has been paid to subnational election as an
important level of analysis to understand partisan evolution in developing
democracies. To address this gap, this chapter focuses on gubernatorial electoral
outcomes to assess party ideological differentiation in Brazil. Runoff elections are
designed to confer legitimacy to the political system by forcing the voter to select
from a subset of candidates who compete for the plurality of the votes in a second
round when the first round fails to produce a majority winner (Pe rez -Lin a n 200 6).
Gubernatorial runoff elections are common in Brazil, with about half of the 27
federal units electing executives in the second round.
Governors are crucial political actors that wield influence far beyond their
state’s borders and can shape national politics. The importance of the governors
resides in the consistency in which they can transform local demands in far reaching
policies and influence party behavior. The electoral district in a gubernatorial
election is the entire state, which in mos t cases is large and diverse in socioeconomic
composition. Where socio-economic disparities are more pervasive, clientelism and
patronage (private markets) – the quid-pro-quo exchange for votes that has
historically been prevalent in Brazil – is thought to diminish the
development of programmatic parties.
Accordingly, this chapter focuses on the public market dimension that
explains the ideological choices that demands programmatic party behavior. Brazil
has now experienced over two decades of stable and competitive elections at all
levels, which may have allowed sufficie nt time for some party amalgamation. At the
same time, the country has also experienced dramatic socio-economic and
demographic changes that could also result in higher demands for programmatic
party behavior. In this sense, some regularity in party behavior that would explain
party-voter linkages at the subnational level should be expected. This chapter argues
that programmatic and clientelistic politics are, respectively, a fu nction of the degree
to which an electoral market is sensitive to public- or private-goods
demands.
In order to adjudicate the issue related to ideological differentiation, I
propose two approaches that draw both on aggregate and on individual-level data at
the subnational level. The analysis clarifies the character of ideological
differentiation in Brazil. Specifically, this asks whether ideology matters for electoral
competition in Brazil.
Theoretical Approach
This chapter draws on a general theory of spatial voting distribution to
analyze the evolution of Brazilian political parties. It assumes that voters choose
parties that best reflect their ideological position (Downs 1957, Enelow and Hinich
1984, Green and Shapiro 1994). The main research hypothesis derives from the
principle of spatial voting, which suggests that the probability of voting for a party
increases as the proximity between party and voter in the political spectrum
increases. This hypothesis will be operationalized differently according to the scope
of the aggregate- and individual-level data. The left-right ideological dimension,
which is the focus of this chapter, is the most used and parsimonious spatial voting
model. In this sense, the spatial distribution of the Brazilian parties considered in
this study is illustrated in Figure 3.1 according to Power and Zucco’s (2009) scale.
Figure 3.1: Spatial Distribution of Selected Brazilian Parties
This analysis draws on gubernatorial elections from 1990 to 2010 to answer
the question of whether political parties in Brazil ideologically distinguishable
among themselves. The dichotomous dependent variable for the aggregate-level
model represents the elected party in the runoff elections. For the individual-level
model, the dichotomous dependent variable represents the respondents’ declared
party choice in the first round elections.
The independent variable of interest is ideological proximity. Several
ideological cleavages could be considered: left- right (Power and Zucco 2009);
Christian-secular (Coppedge 1997); xenophobic nationalist-ethnic/regional
separatist (Horowitz and Browne 2005); and liberal-conservative (Huber and
Inglehart 1995). Among those cleavages, the unidimensional left-right scale tends to
be a consensual distribution of the political parties’ ideology in Brazil (Coppedge
1997, Huber and Inglehart 1995, Power and Zucco 2009). I extend this dimension
for the ideological cleavage of the voter as well, based on the theory that the
ideological position of the parties reflects the ideological position of the voter
(Downs 1957). Further, assuming ideological consistency, in the runoff, the voter will
choose the closest party to her ideal point, which is indicated by her choice in the
first round. For this reason, all else being equal, the probability of a party to be
elected increases if the aggregate votes received in the first round by this party and
by its ideological proximate parties are greater than the votes received by all the
other parties.
The variable of interest for the aggregate-level model is ideological proximity,
which is the result of the addition of votes according to the spatial distribution of the
parties in the left-right ideological political spectrum. I use Power and Zucco’s
(2009) index of party ideology, which varies from one (left) to ten (right), to position
each party in the political spectrum. Then, I redistribute the first round votes
according to their ideological proximity, assuming that i n the first round each party’s
voters occupy that party’s ideal point and that in the second round the party chosen
is the ideologically closest to the voter’s single-peaked preference. After
redistributing the votes, I generate a binary variable, which assumes the value one
for the highest aggregation and zero oth erwise.
The control variables for the aggregate-level analysis are incumbency,
position in the first round, president proximate coattail, and president concurrent
coattail. The theory of incumbency advantage states that, for good or bad, the
executive is personally responsible for the performances of the policies adopted.
From an advantageous perspective, incumbent executive officers have access to
public resources and distribute “pork.” They also benefit from name recognition
(Hinckley 1980), can use free and continuous access to the media, and have more
access to campaign funds (Jacobson 1985). While incumbency may be negative
during times of economic and political turmoil, it is mostly advantageous for the
candidate. Given the first round threshold of 50%, the existence of a runoff and the
advantages of incumbency, I expect a positive relationship between a candidate or
party’s incumbency and subsequent party election.
The distinction between candidate and party incumbency is important for
several reasons. First, party and candidate incumbencies do not necessarily
correspond, as party switching is a widespread phenomenon in Brazil (Ames 2001,
Melo 2000). Although party migration is less frequent among governors, as they are
extremely powerful (Samuels 2000) and exert considerable control over the party
decisions, there is no incentive for party defection. Second, reelection for the
executive offices has only been adopted since 1998. By adding party incumbency, the
analysis captures the effects of incumbency thoroughly, without missing the
marginal effect of candidate incumbency before the re-election rule. Finally, party
and candidate interact complementarily rather than exclusively (Hinckley 1980).
Hence, the structure of the incumbent party will necessarily grant some advantage
to his candidate. Both party and candidate incumbency are binary and assume the
value one when the candidate or party is incumbent and zero otherwise.
First round elections that failed to produce an outright winner, thus
provoking a runoff, may have mechanical and behavioral implications. From a
mechanical perspective, the best-positioned candidates in the first round needs to
acquire fewer new voters than the runner-up candidate in order to succeed in the
runoff. The behavioral effect on voters may result from the fact that the
bestpositioned candidate in the first round can claim the result as a preliminary
victory. That may create a sense of inevitability since rational voters do not like to
waste their vote, in the sense that such a vote would not contribute to selection of
the wi nner. Hence, I expect a positive relationship between position in the first
round and party election. The variable first round position refers to whether the
candidate finished the first round in the first or second place. This is a binary
variable in which one represents the first place and zero the second place.
Studies on coattail effects suggest that h igh presidential approval ratings
increase the electoral chances of other party members (Campbell 1986, Ferejohn
and Calvert 1984). The coattail effects of proximate elections have also been related
to legislative fragmentation in cross-national election (Cox 1997, Golder 2006,
Lijphart 1994) as well as subnationally (Ames 1994, Samuels 2000). The result s of
presidential coattail effects on party fragmentation, though, are conflicting . Some
authors argue that presidential elections increase party system fragmentation
(Filippov et. al. 1994), while others disagree (Cox 1997, Shugart and Carey 1992).
Yet, a third view finds no correlation between presidential elections and
congressional party size (Samuels 2000, Coppedge 2002). Although not consensual,
the literature indicates some level of influence of presidential elections on legislative
contests. In Brazil, gubernatorial and presidential elections are concurrent and
campaigns are mounted around the presidential candidate so as to minimize
coordination cost and maximize electoral outcomes. Thus, I predict a positive sign
between elected president and elected party in gubernatorial races.
Finally, I control for the concurrent coattail effect of the incumbent
president’s party on gubernatorial runoff elections and expect a positive sign
between those two variables. T he influence of presidential coattails is modeled by its
proximate and concurrent effects. The proximate effect is coded one when the
previous elected president and currently elected governor belong to the same party
and it is coded zero otherwise. Similarly, the concurrent coattail is coded one when
both president and governor elected simultaneously belong to the same party.
The independent variable of interest in the individual-level model is distance,
which is measured by the difference between voter’s self-pla cement in the political
spectrum and the party’s placement according to the Power and Zucco (2009) scale.
The individual-level analysis controls for demographic, socioeconomic, and
sociocultural variables. Older age is associated with the rightist vote due to the
memories of military regime anti-leftist propaganda (Ames and Smith 2010). The
pattern of the women’s vote in Brazil has also been associated with rightist parties
(Ames and Smith 2010). Higher income has generally been associated with
identification with the leftist Worker’s Party (PT). However, since Lula’s reelection in
2006, there has been a shift in the PT’s electoral base to include lower income
voters.
Personal income is captured as a catego rical variable with eight increments.
A higher level of education is associated with voting for the PT (Zucco 2008). Th e
Bolsa Família (BF) program, a federally funded conditional cash transfer policy that
benefits poor families in excha nge for keeping their children in school and
vaccinated, is also considered. This program has been linked to the success of the
PT’s re-election for the presidency in 2006 (Zucco 2008), but it appears that the
same correlation is non-existent for governors, as voters credit the program to the
central government and Lula (Sugiyama and Hunter 2013). BF is a dichotomous
variable with 1 representing participants in the program and zero otherwise. Finally,
I include clientelism, as its effects in developing democracies have been studied
extensively (Kitschelt 2000, Remmer 2010). This variable is dichotomous and
measures respondent’s attitude toward clientelism with 1 indicating a clientelistic
attitude and zero otherwise.
The nature of the individual-level statistical model does not allow us to relate
the alternative specific variables directly to the dependent variable. However, it is
possible to predict the variables’ direction based on the position of the party in the
political spectrum. As explained above, the expectation is that the higher the income
the higher the probability to vote for the leftist party. We expect the same
relationship for educational levels. Since recent research suggests no true linkage
between BF and governors, no direction is anticipated on this variable. Finally,
following the literature, age and clientelism are expected to be negatively associated
to leftist parties and positively associated to rightist parties.
Aggregate-level Data and Analysis
The data for this dissertation was collected from the Brazilian Superior
Electoral Tribunal (Tribunal Superior Eleitoral), which serves both as electoral
commission and appeals court. The dataset includes 78 gubernatorial runoff
elections for six consecutive elections between 1990 and 2010, totaling 156
observations as there are two candidates for each runoff. States-years with deci sive
first round elections are not included in the dataset. The dichotomous dependent
variable elected party assumes the value one when a party is elected and zero
otherwise. The independent variable of interest is dichotomous ideological
proximity, measured as the aggregate vote of the most ideologically proximate
parties in the first round. I expect that the larger the aggregation of these votes, the
greater the chance of the most proximate party to be elected in the runoff election.
The aggregate-level analysis applies a unidimensional deterministic model of
spatial electoral competition in runoff elections for governors. In a related study in
France, Rosenthal and Sen (1977) test the validity of the different approaches and,
although they conclude that probabilistic models as opposed to deterministic ones,
offer an overall best prediction. They also add complexities that are necessary to
account for when the electoral system allows more than two candidates in the
runoff. Since the Brazilian system yields only two runoff candidates, the
deterministic model is perfectly adequate. Moreover, they also call attention to the
temporal stability separating the voter’s choice between the first and second round
and assert that runoff election can be reasonably predicted by resorting to
one single political dimension (left-right).
This method allows for the concomitant assessment of the distribution of
parties’ votes along the political spectrum and the corresponding redistribution of
votes in the runoff according to the principle of most proximate vote. This model
also allows for the inference of the winner of the runoff election. Finding regularities
in the redistribution of votes between the first round and the runoff elections would
imply some level of party and electorate ideological differentiation at the aggregate
level. Parties matter electorally and their brands bear some strategic value both to
the leaders and to the voters, as the strategies adopted by party elites are not
dissociated from those of the electorate.
Although many external factors concur to explain the electoral outcome, I
focus on endogenous variables leading to the runoff election to infer the ideological
differentiation of the Brazilian party system. In fact, an advantage in studying runoff
models of spatial voting is that the exogenous informatio n is mostly accounted for in
the first round (Rosenthal and Sen 1977).
The spatial voting model offers a predictive model for gubernatorial runoff
elections and sheds light on the issue of ideological differentiation. A predictive
model of a binary outcome only makes sense if it can forecast over 50% of the
variation of the dependent variable, since a random pick would yield a 50% chance
of selecting the winner. Indeed, the mod el overall makes correct prediction of about
75% of the cases. More specifically, the model predicts the wi nner in 79% of the
cases and the loser in 71% of the cases. The baseline model (model 1), as shown on
Table 3.1, indicates that ideological proximity, first round order, and presidential
proximate coattail are statistically significant at p<.05 or lower and the signs are in
the expected direction.
Table 3.1: Determinants of the Winner Party in Runoff Election
DV: Elected Party
Model 1
Model 2
Model 3
Model 4
Model 5
Ideological
Proximity 1.447***
1.615***
1.429***
1.534**
1.153*
(0.39)
(0.40)
(0.39)
(0.58)
(0.56)
Incumbent
Candidate 0.325
0.180
1.370
-0.553
(0.88)
(0.51)
(0.83)
(0.69)
First Round
Order
1.816***
1.806***
1.825***
1.516*
2.256***
(0.39)
(0.40)
(0.40)
(0.60)
(0.56)
Proximate
Coattail
1.315*
1.285*
1.486
1.165
(0.53)
(0.54)
(0.78)
(0.82)
Concurrent
Coattail
3.052**
(1.14)
Incumbent
Party
0.17
(0.45)
Constant
1.927***
-1.927***
-1.897***
-2.049***
-1.748***
-0.38
-0.38
-0.37
-0.56
-0.51
Pseudo R2
0.250
0.273
0.249
0.276
0.265
LR
-81.09
-78.58
-81.24
-37.18
-41.78
N
156
156
156
74
82
* p<0.05, ** p<0.01, *** p<0.001, standard errors in parenthesis
The position in the first round confirms the ideological differentiation
hypothesis and is the single strongest predictor of the winner in gubernatorial
runoff elections. In model 1, each unit change on this coefficient indicates a 514%
increase in the odds of being elected. Moreover, since the literature suggests that the
effect of incumbency and order in the first round may not be an additive one, I test
the interaction between those two terms but find no statistical significance (not
shown in the model).
Models 2 and 3 test the coattail effect while also controlling for variations in
incumbency. Presidential proximate coattail is statistically significant at p<.05. This
is the weakest coefficient but it still indicates that a unit increase in the presidential
proximate coattail increases the odds of being elected by 272%. Most importantly,
however, is the effect of concurrent coattail on the outcome of the dependent
variable. According to model 2, the independent effect of belonging to the same
party as the president in concurrent elections suggests that for each unit increase of
this variable the odds of winning the runoff contest also increase by 2000%. This
finding strongly supports the concurrent coattail hypothesis and indicates, as
Brambor et. al. (2006) suggest, the overall influence of presidential coattails in other
elections.
Ideological proximity also offers a robust support for the outcome of the
gubernatorial runoff election at the p<.001 level across the first three models. When
the ideology proximity variable assumes a value of one for the largest aggregation of
votes in the first round redistributed between the two runoff candidates, a unit
increase change in the odds of predicting the elected party increases about 325%.
This means that on average, the electorate differentiates the parties according to
their ideological position in the political spectrum and will rationally choose parties
in the runoff that are ideologically closer to the one that they chose in the first
round. Figure 3.2 shows the probability of a party to be elected given the aggregate
votes of the most ideologically proximate preference of the elector in the first round
of the election. Thus, when ideological proximity changes from zero to one, the
probability for a party of getting elected increases from about 0.3 to about 0.7.
Figure 3.2: Probability of Party Election Given Ideological Proximity
Interestingly, candidate incumbency is not statistically significant, although it
is in the expected direction. A further test of party incumbency also failed to show
statistical significance (model 3). A possible explanation for the lack of explanatory
power for incumbency may be related to the majoritarian requirements of runoff
elections coupled with competitive elections and multiparty characteristic of the
Brazilian party system. Multiparty competitive elections would make it more
difficult for a candidate to decide the election in the first round. In fact, if
incumbency gives any advantage to a candidate, it would be realized in the first
round. However, run-off elections exist exactly as a consequence of the failure of any
candidate to achieve the majori ty of the votes in the first round. In this case, it could
be argued that the failure of the incumbent candidate (provided there is one in the
race) to win the election outright in the first round may indicate that the aggregate
voters prefer all challengers over the incumbent (Bullock and Johnson, 1985).
Moreover, by the time the run-off campaign starts, the non-incumbent candidate
who won a spot in the run-off ticket has already had sufficient exposure to offset the
name recognition advantage conferred to the incumbent during the first round
election.
In order to refine the analysis, two addit ional models (4 and 5) separate
electoral competition on a left-right basis. It has been said that the differentiation of
the Brazilian party system is observed only on a left-right basis. To test this claim,
model 4 restricts the data to electoral contests that occur between leftist and rightist
parties. Model 5 restri cts the data to electoral competitions either among leftist
parties or among rightist parties. Thus, if Brazilian parties were indeed
differentiated on a left-right basis, model 4 would show a statistically significant
relation between ideological proximity and elected party. Conversely, if Brazilian
parties are only differentiated on a left-right basis, then we should expect no
statistical significance of the variable ideological proximity in model 5. If the
differentiation happens across the political spectrum, we would expect the
coefficient on model 4 to be higher than the coefficient on model 5, because it would
be easier for the electorate to differentiate between a left and a right party than
among left of the center or among right of the center parties.
Indeed, this test provides additional evidence on party differentiation, as the
variable ideological proximity is statistically significant for both models 4 and 5.
Most importantly, the finding indicates the electorate is not only able to differentiate
parties on a left-right dimension but also to make more nuanced differentiation
among parties on the left when leftist parties face competition against one another,
or among parties on the right that compete amongst other rightist parties. Moreover,
as expected, the coefficient in model 4 is not only more robust than that in model 5
but also the level of statistic significance is better at p<.01 versus p<.05. Thus, we
can confidently say that the analysis shows that the Brazilian party system is indeed
ideologically differentiated.
Individual-level Data and Analysis
To shed further light on partisan ideological differentiation, I turn to
individual-level data on the 2006 first-round gubernatorial election. The data
comprises 620 interviews with Brazilians conducted by the Latin American Public
Opinion Project – LAPOP. The dependent variable is the self-declared vote in the first
round gubernatorial election of 2006. The survey was conducted in 18 of the 27
states (including the Federal District) and is representative of all five Brazilian
geographical regions. The independent variable of interest is distance, which is
measured by the difference between voter’s self-placement on the political spectrum
and the party placement according to the Power and Zucco (2009) scale. The
expectation is that the smaller the distance between voters’s self-placement and the
party’s placement, the more likely the individual will vote for the candidate of that
party.
The individual-level analysis resorts to conditional logistic (fixed-effects)
models to assess the distance between party placement and voter’s self-placement
in the political spectrum. Both respondent and party placement are measured on a
one to 10 scale, with one representing the extreme ideological left and 10
representing the extreme ideological right of the political spectrum. Contrary to
multinomial logit models that consider only the position of voters, conditional
logistic models have the advantage of placing voters relative to parties (Alvarez and
Nagler 1998). Thus, it is the appropriate model to assess the ideological
differentiation of voters and parties in a given party system. This is important
because the outcome is the choice among several candidates given the multiparty
nature of the Brazilian system. It also makes it easier to combine case-specific and
alternative-specific variables (Long and Freese 2006).
The conditional (fixed-effects) logistic model includes both case-specific and
alternative-specific variables. This model allows for inferences about variations
across individual observations (case-specific) and about variations within individual
observation (alternative-specific). In this model the case- specific variables are:
gender, age, income, and Bolsa Família status. The alternative specific variables are
the self-reported votes for the 2006 gubernatorial election. The results are shows in
Table 3.2.
The alternative-specific conditional logistic model of ideological vote in Brazil
shows 3,535 observations for 505 cases distributed in seven alternatives per case
(party choice). Since many of the Brazilian parties are regional, region is nested into
the model to limit the individual choices to the effective choice of parties available in
that region, thus avoiding perfect prediction for those parties. However, no
inferences can be made about specific regions, as the model with a region dummy
does not converge statistically. The nested model is overall statistically significant at
p<.05.
The key finding is that the Brazilian political party system shows some degree
of party and voter ideological differentiation. In models 1 and 3, the variable
distance shows the right sign and it is statistically significant. It shows that the
shorter the ideological distance between the voter and the party the higher the
probability for the voter to chose that party, which indicates voters’ and parties’
ideological differentiation. In the full model (shown in Appendix A for the sake of
space) and in model 2 the variable distance is in the predicted direction, but it is not
statistically significant after the exclusion of the Bolsa Família variable.
Table 3.2: Alternative Specific Conditional Logit
Estimation for the 2006 Brazilian
Gubernatorial Election
DV: Party Choice
Base Line: PP
Model 1 Model 2 Model 3
b/se b/se b/se alt
distance -0.020* -0.017 -0.023*
(0.01) (0.01) (0.01)
PDT
income -0.105 -0.100
(0.19) (0.18) bolsa família -
0.361 -0.147
(1.47) (1.44) constant
0.733 0.682 0.317
(0.72) (0.70) (0.30)
PT
income -0.137 -0.148
(0.15) (0.15) bolsa família
0.178 0.666
(1.12) (1.08) constant
2.250*** 2.306*** 1.747***
(0.60) (0.58) (0.25)
PMDB
income -0.136 -0.152
(0.15) (0.15) bolsa família
0.548 0.859
(1.09) (1.06) constant
2.287*** 2.377*** 1.845***
(0.59) (0.57) (0.23)
DEM
income -0.025 -0.063
(0.17) (0.16) bolsa família
1.039 1.507
(1.14) (1.09) constant
0.808 1.035 0.573*
(0.67) (0.63) (0.27)
PSB
income -0.349* -0.400*
(0.17) (0.17) bolsa família
1.339 2.055
(1.10) (1.06) constant
2.081** 2.375*** 1.004***
(0.64) (0.61) (0.27)
PSDB
income 0.004 -0.034
(0.15) (0.14) bolsa família
0.914 1.091
(1.08) (1.05) constant
2.010*** 2.217*** 2.014***
(0.59) (0.56) (0.23)
N 3535 3563 4200
BIC 1878 1852 2156
* p<0.05, ** p<0.01, *** p<0.001
The demographic (age and gender) and socio-economic (income) variables,
while contributing to the fit of the model, are not statistically significant, except for
income when it comes the PSB (Socialist Party of Brazil). This finding is inconsistent
with the literature that usually associates high-income voters with parties on the left
of the political spectrum (e.g. Zucco 2010). In fact, each unit increment in income
decreases the odds of voting for PSB versus voting for PP by factor of 0.71, or 29%.
However, it is somewhat consistent with recent electoral shifts that link low-income
vote to the main Brazilian leftist party (PT).
The relationship between income and party choice (Figure 3.3) shows that
the Brazilian Social Democracy Party (PSDB ) is the most appealing party to higher
income voters. In fact, the slope shows that a one-unit increase in income increases
the odds of voting for PSDB as opposed to voting for PP (the baseline party) by a
factor of 1.004 (0.4%). The opposite is true for its main electoral competitor, the PT.
In this case, a one-unit increase in income decreases the odds of voting for the PT as
opposed to voting for PP (the base line party) by a factor of 0.87 (13%). This fi nding
is consistent with Zucco (2008) and Montero (2010), who find regional divides in
presidential elections.
The direction of the variable Bolsa Família shows that all parties, with the
exception of PDT, benefit from this program. However, none of them are statistically
significant, which aligns with Sugiyama and Hunter’s (2013) analysis that the Bolsa
Família is perceived as being federally provided, indicating that state and local
politicians do not reap any benefit from managing the recruitment of its
beneficiaries.
Conclusion
Scholars of party institutionalization have engaged in a longstanding and
animated debate over the existence of ideological coherence among developing
democracies. Brazil is exemplary of such debate, where scholars have long grappled
with the question of whether parties have developed meaningful ideological frames.
Contrary to some assessments that Brazilian parties display non-programmatic and
clientelistic characteristics, this study shows remarkable nuance in terms of
ideological differentiation. Importantly, parties demonstrate ideological linkages
with voters at the subnational politics level, where personalism and clientelism have
historically prevailed. This chapter draws on an innovative methodological
approach that uses aggregate- and individual-level data to help settle this debate.
At the aggregate level, the unidimensional deterministic model of spatial
electoral competition for gubernatorial runoff elections confirms the hypothesis of
ideological differentiation. This claim is supported by logistic regression models
based on the principle of the most proximate vote, which posits that voters will
chose the closest parties to his own ideological position. To test this hypothesis, the
model aggregates the votes cast to parties in the first round and redistributes them
accordingly to the ideological proximity between the two parties selected for the
runoff election. This method allows for probabilistic inference on the outcome of the
runoff election by assuming ideological transference of votes between the first round
and the runoff election. Further, as a consequence of the predictive regularities, it
also indicates ideological differentiation of the Brazilian party system.
Most importantly, the empirical analysis at the aggregate level confirms the
hypothesis of ideological proximity. This result suggests that in runoffs, the
electorate indeed choses the most proximate party to the one she selected in the first
round of the election. Moreover, this study suggests the Brazilian party system is
differentiated both at the party level and at the electorate level. The evidence at the
electorate level is also strongly confirmed by additional models that restrict the
sample to electoral disputes among left of center parties, as well as competition
among right of center parties.
Applying a new methodological approach based on alternative-specific
conditional logistic regression models, this dissertation further confirms the
ideological differentiation of the Brazilian party system through individual- level
data. Using national survey data from the 2006 LAPOP, it finds a negative and
statistically significant relationship between the case-specific variable distance,
measured as the modular difference between voter’s self-ascribed ideological
placement and party’s external measure of ideological position in the political
spectrum, and voter’s choice.
This chapter makes an important contribution by adding nuanced analysis
about ideological political linkages in subnational politics, where personalism is
more likely to prevail. In this sense, subnational elections provide a high-bar for
testing whether parties have developed meaningful ideological linkages with voters.
The evidence strongly supports the ideological differentiation of the political parties
in Brazil. Most remarkably, even competition between parties within the same
ideological spectrum shows partisan differences across Brazil where subnational
political competition is often deemed prone to patron-client relationships.
Having settled the matter of whether Brazilian parties display ideological
differentiation, we can turn to the issue of how clientelism may relate to the more
stable ideological features of subnational partisanship. In the next chapter I discuss
the clientelism dimension of party linkage. First, I propose a measurement of
clientelism (Herfindahl Index), which is based on the geographic concentration of
votes received by parties and an alternative measurement based on clustering and
dispersion of votes (Cluster Index). After locating the parties along the
programmatic-clientelistic political spectrum, I compare and test the validity of the
measurements proposed with regression models.
CHAPTER 4 Political Competition in Brazil: Measuring Party
Clientelism
Is there a clientelistic dimension to partisan competition in Brazil? This
chapter focuses on the political aspect of clientelism, defined as electoral
transactions between citizens and politicians involving exchanges of votes for
private goods (Kitschelt and Wilkinson 2007). I argue that, in spite of programmatic
advancements, clientelism persist as a crucial element in political parties’
competitions. Competitive elections in multiparty democracies with diverse
electoral markets compel parties to adopt strategies that allow them to garner votes
in both predominantly public and private markets. To unveil the electoral strategies
employed by parties when competing in private or public electoral marketplaces, I
test a model of geospatial dominance of partisan voting for the Brazilian Chamber of
Deputies from 1998 to 2010. Working with two hypotheses: 1) that public goods
markets would generally benefit programmatic parties and, conversely, 2) that
private goods markets would benefit clientelistic parties, this chapter applies the
Herfindahl Index, a general measure of market concentration by companies, as a
proxy for clientelism to assess the dynamics of party competition. I further refine the
methodological approach by proposing a new Cluster Index of geospatial measures
of clustering and concentration of votes.
This chapter is organized as follows: the first section proposes a
measurement of clientelism based on the Herfindahl Index (HI) the second presents the
models a nd findings, the third analyzes the findings for the HI models, the fourth proposes an
alternative measurement based on clustering and dispersion of votes (Cluster Index- C I), the
sixth discusses the findings for the CI, and the last section concludes.
The Herfindahl Index
As developed in Chapter 2, this dissertation examines whether partisan
competition is structured along two dimensions: ideological commitments and
clientelistic/programmatic strategies. This chapter tests the hypothesis associated
with the clientelistic dimension of partisan politics. Specifically, I draw on Ames’
(2001) concept of dominance and clustering, which describes the spatial
distribution of electoral support for a candidate for the Chamber of Deputies and
Montero (2010), who uses geographical information system (GIS) to measure the
degree of dispersion and concentration of parties according to their ideology. In this
vein, dominance is related to the ability of a deputy to receive a high percentage of
vote shares in municipalities that contribute to the majority of his individual vote.
This phenomenon is more common in rural areas in which the mediation of the local
boss is crucial for the deliverance of such high vote share thus making the candidate
less accountable to the electorate. Conversely, clustering is related to strategy in
which deputies prospect their individual votes in contiguous municipalities.
This phenomenon, in turn, does not necessarily require the mediation of local
bosses thus making the candidate more accountable to the electorate. This is
consistent with the theory develop ed h ere, which states that public electoral
markets are more conducive to the success of programmatic parties while private
electoral markets tend to benefit clientelistic parties. Different from the Ames
(2001) and Montero (2010), I borrow the Herfindahl Index (HI) concept, which
measures monopolistic behavior by companies, to estimate the level of competition
among parties. High HI indicates clientelistic behavior while lower HI indicates
programmatic behavior.
The logic is that programmatic parties’ votes are more dispersed, while
clientelistic parties’ votes tend to be more concentrated due to the effect of
dominance. By this token, through the dispensation of private goods, clientelistic
parties are able to secure the support of local political bosses who, in exchange,
deliver bulks of votes in a given municipality. Without accounting for the contiguity
of the municipalities, clientelistic parties’ votes share of these basic electoral
districts would be proportionally larger in comparison to programmatic parties,
which would catch ideological votes dispersed among the different electoral
districts. This aims to answer the question of whether there is a clientelistic
dimension to partisan competition in Brazil.
Data and Measurements
The Brazilian electoral commission provides party election data for the
Chamber Deputies from 1998 to 2010 for all states, except the Federal District,
resulting in 2,639 available observations. Utilizing the data on the Chamber of
Deputies, as opposed to the majoritarian election, makes it possible to compare
parties sub-nationally as they are more likely to run candidates across the different
states and coordinate it as part of a national strategy to maximize their seats. The
majoritarian election restricts the number of parties for the effect of comparison
since only a few parties run candidates for executive positions. The mayoral election
is an exception, but even in this case, the eminently local nature of this kind of
election makes it less comparable nationally.
Dependent Variable
Herfindahl Index
I draw my measurement of clientelistic parties from Ames (2001) according
to the principle of dominance and clustering as described above. However, the
Herfindahl Index (HI) only considers the individual municipality where votes are
prospected. In this sense, the HI does n ot account for the contiguity of the
municipalities, which is equivalent to the clustering strategy. Instead, by looking into
the municipalities individually, the HI considers concentration and dispersion of
votes as two opposite strategies respectively associated with clientelistic and
programmatic party behavior. The concentration and dispersion of the votes is first
calculated for each candidate in each municipality, which also takes into account
population size. Further, the results are integrated at the party level by taking the
average of the candidates per party and for each state and election-year. Finally, the
values of HI are normalized to account for the extremely high distribution of smaller
parties. The HI varies from 0 to .8, where high indices of the variable would indicate
that the party’s votes are more concentrated while low indices of the variable would
indicate that votes obtained by the party are more dispersed. Thus, since each
strategy adopted reflects the maximizing utility of each party according to the
characteristics of the electoral market in which they are competing, this measure
would indicate the degree to which a party is electorally
clientelistic.
Independent Variables
Ideology
Right of center parties are usually associated with clientelistic linkage
strategies (Montero 2010). The test of this variable will be conducted by using the
left-right measures of ideology as devised by Power and Zucco (2009), which
comprises six waves of measurements based on interviews with congressional
deputies and senators. The scale goes from one (extreme left) to 10 (extreme right).
Thus, we would expect a positive sign for this variable. That is, as parties move to the
right of the political spectrum they tend to concentrate more votes.
Private Sector’s Size
This variable measures the ability of the municipality to generate taxes and is
obtained by dividing local tax revenues by current municipal income and averaged
by state. Autonomous municipalities suggest a more developed private sector and
thus a population that is less susceptible to quid pro quo politics resultant from
clientelistic linkages. I would expect a negative relationship with the dependent
variable. That is, the greater the size of the private sector, the less likely a clientelistic
party will prospect votes in that municipality.
Public Administration’s Size
This variable measures the public sector economic contribution as a
percentage of the economy in the year prior to the election (Zucco 2008). This is
obtained by dividing the GDP of the public administration by the municipality’s
overall GDP, and then averaged by the state. A jurisdiction with a relatively high
proportion of its economy reliant on the public sector would indicate a strong
dependence of the municipality on publ ic funds, which can be a source for
clientelism. Social indicators condition vote-buying and other clientelistic behavior:
greater poverty, inequality, and generally lower social indicators coincide with a far
higher incidence of clientelistic behavior (Desposato 2001). Thus, I would expect a
positive relationship with the dependent variable.
GDP per Capita
This variable captures the municipal GDP per capita, which includes all
sectors and uses a national implicit deflator. The GDP per capita is the average of all
municipalities for each state for the four years preceding the election. This variable
is a proxy for modernization, which predicts that as economic development
increases demands for private goods or clientelism decreases. Clientelism is
generally conceived as reminiscent of traditional societies. As such, modernization
would eventually erode the moral structures that mold patron-client relations
(Nichter 2010) paving the way to programmatic linkages (Scott 1969). I would
expect a negative relationship with the dependent variable.
Control Variables
Urbanization
Urbanization is also a proxy for the modernization effect that predicts
increasing demand for policies. This variable measures the ratio of urban to rural
population in the state for each electoral cycle according to the most recent census
conducted by the Brazilian Institute of Geography and Statistics (IBGE). The data
was obtained from the Institute of Appli ed Economic Research (IPEA) website. I
expect a negative relationship with the dependent variable, as clientelistic parties
would fare better in more rural environments.
Population
This is a standard control variable as it has been widely reported that the size
of the population matters in explaining socio-political factors. I use the log of the
population of the states based on the most recent census prior to the electoral cycle.
I predict a negative relationship with the dependent variable, as state population
growth in Brazil is historically associated with the modernization effect of
industrialization and urbanization.
Region
The Northwest, North, and Center-West regions report higher levels of
votebuying than the South and Southeast (Epstein 2009). To test this variable I use a
dummy for the five regions with the Southeast as the baseline. Thus, I would expect a
positive sign for the first two regions and a negative one for the South region.
District Magnitude
District magnitude refers to the number of representatives elected in a single
electoral district, which in Brazil coincides with the state boundaries. There is large
variation in the district magnitude with the smallest state electing eight
representatives and the largest electing the maximum of 70. The size of district
magnitude has been positively associated with personal vote: as the district
magnitude increases in systems in which intraparty competition is high (open list
with single vote) so does the necessity for each candidate to differentiate hersel f
from her copartisans, thus resorting to personal vote strategies (Shugart and Carey
1995). Thus, I expect a positive sign for this variable.
Model and Findings
The statistical model purports to test the validity of the dependent variable,
Herfindahl Index, which is a proxy for clientelism measured by the principle of
concentration of party votes in the electoral district. This original measure helps to
unveil parties’ electoral strategies on a dimension that has proven elusive to
measurements. The continuous nature of the dependent variable requires an OLS
model. Although several variables are correlated, tests show that collinearity is no
reason for concern (see Appendix).
The full model can be expressed econometrically in the equation below:
HI= β0 + β1Ideology + β2Local_Taxes + β3Size_Public_Sector + β4GDP_Capita
+ β5Urbalization + β6Log_Population + β6 + β7North + β8Northeast +
β9South + β10Centerwest + β11District_Magnitude +
β12Local_Taxes*Size_Public_Sector + e
Table 4.1 shows the estimation of five regression models explaining the
clientelistic-programmatic behavior of the Brazilian parties. Model 1 includes the
main explanatory variables, Model 2 adds the socio-demographic variables, Model 3
includes the regional and institutional variables, Model 4 adds an interaction term,
and Model 5 combines the regions into two more homogenous grand regions.
Surprisingly, ideology is not statistically significant in any model tested. It appears
that the common belief that rightist parties tend to be more clientelistic does not
bear out in these models. The variable private sector’s size is statistically significant
(p<.001) from Models 1 to 3 but the effect on the parties’ clientelistic behavior is
contrary to the expected. However, in Model 4 and 5 it is not statistically significant
and the sign inverts its direction in Model 5.
The variable public administration’s size shows a positive sign and it is
statistically significant (p<.001) in Model 1. However, contrary to the prediction,
when the socio-demographic, regional, or institutional variables are added (Models
2-5), the sign inverts and it remains statistically significant (p<.001). The variable
GDP per capita, which testes the modernization effect, i s statistically significant
(p<.001) and in the expected direction in all models. For each unit increase in the
GDP per capita there is an 11% decrease in clientelistic voting (Model 4).
Table 4.1: Concentration and Dispersion of Party Vote DV:
Clientelistic Party (Herfindahl Index)
Model 1 Model 2 Model 3 Model 4 Model 5
b/se b/se b/se b/se b/se
ideology_100 0.026 -0.027 -0.070 -0.067 -0.047
(0.00) (0.00) (0.00) (0.00) (0.00) localtxyr
2.213*** 1.972*** 1.805*** 0.343 -0.151
(0.10) (0.10) (0.10) (0.34) (0.33)
sizepub 0.078*** -0.034 -0.076** -0.217*** -0.231***
(0.02) (0.02) (0.02) (0.04) (0.04) gdpcavg -
0.020*** -0.009*** -0.014*** -0.011*** -0.008***
(0.00) (0.00) (0.00) (0.00) (0.00) urbanization
-0.003*** -0.005*** -0.005*** -0.004***
(0.00) (0.00) (0.00) (0.00) logpop -
0.037*** -0.046*** -0.044*** -0.060***
(0.00) (0.00) (0.00) (0.00) north
-0.030*** -0.031***
(0.01) (0.01) northeast -
0.072*** -0.068***
(0.01) (0.01) south -
0.052*** -0.052***
(0.01) (0.01) centerwest -
0.071*** -0.061***
(0.01) (0.01) magnitude
0.001*** 0.001*** 0.002***
(0.00) (0.00) (0.00) c.localtxyr#c.size~b
3.817*** 4.800***
(0.84) (0.82) grandnorth
-0.025***
(0.01) constant 0.183*** 0.759*** 0.976*** 0.979***
1.172*** (0.01) (0.03) (0.06) (0.06)
(0.05)
R-sqr 0.205 0.342 0.402 0.407 0.386
dfres 2632 2630 2625 2624 2627
BIC -4978.2 -5459.5 -5673.1 -5686.0 -5619.3
* p<0.05, ** p<0.01, *** p<0.001
The results for the socio-demographic variables are consistent with the
predicted direction and they show a statistically significant (p<.01) relationship with
clientelism in all models. As urbanization increases one unit clientelistic votes
decreases by 5%. Likewise, for each 1% increase in population size clientelistic votes
decrease by .00044 units in Model 4. The regional variables slightly contribute to the
overall fit of the model as their introduction increases the R-square from .342 in
Model 2 to .402 in Model 3. However, using the Southeast as a baseline, only the
regions South and Center-West show the predicted sign with a negative and
statistically significant relationship with clientelism (p<.001). The North and
Northeast regions are both statistically significant but with the inverted sign.
The full model (Model 4) tests political, clientelistic-programmatic,
modernization, and demographic-regional variables. The F test indicates a good fit
for the model, with F(12, 2624) = 149.97 and a significant probability level (Prob > F
= .000). The model explains 40.7% of the variance on clientelism (R-squared = .407).
The strongest predictor in this model is public administration’s size (-.217; p<.001).
Finally, the institutional vari able district magnitude shows a positive, weak, and
statistically significant relationship with clientelism. For each unit change in the size
of the district magnitude, an increase of only 1% would be expected in clientelism.
Discussion of The Herfindahl Index
The Herfindahl Index (HI) has mainly been used to assess the monopolistic
behavior of companies. It has also been applied in Political Science to measure party
system fragmentation (Dalton 2008) and concentration (Bardhan and Yang 2004).
To my knowledge, the HI has not been used before to proxy party clientelistic
behavior. I argue that parties devise electoral strategies that are reflected in the
geographic concentration or dispersion of votes. Concentration results from parties’
engagement with local political bosses who tend to control bulks of votes in their
respective municipalities, which is a characteristic of clientelistic politics.
Dispersion, in contrast, derives from parties’ engagement on ideological politics ,
which will find likeminded voters across the regions. The adaptation of the HI to
capture the electoral market concentration of political parties could provide a useful
measurement of clientelistic party behavior.
The models overall offer some support for the validity of the measurement of
clientelism based on the HI but their interpretation is more nuanced. In fact, taken as
states’ average, municipalities in which the public administration’s size is greater
indicates that parties tend to be more clientelistic. However, a change in the sign of
both public administration’s size and public sector’s size suggests an interaction
between the two variables. In order to understand this interaction, Figure 4.1 breaks
down the values of public administration’s size into the 5th, 50th, and 95th percentiles.
It shows that the median effect of private sector’s size on clientelism
increases as the public administration’s size also increases.
Figure 4. 1: Predictive Margins of the Effect of the Private Sector’s Size on Clientelism
(Herfindahl) by Higher, Median, and Lower Values of the Public
Administration’s Size.
The interpretation of this effect should consider the supply and demand side
of clientelistic politics. Municipalities with more capacity to generate local taxes, that
is, a more dynamic private sector, implies the existence of a bigger public market
with more demands for parties with ideological appeals. The consequence is that the
diminished pool of private market voters would increase the price of the clientelistic
vote, thus requiring more publi c resources to buy it. However, this relationship
becomes more complex when the highest (95th percentile) and lowest
(5th percentile) values of the public administration’s size are considered. In fact, on
one hand when the private sector’s size is below .05, states in which the
municipalities have bigger public administration’s size show a smaller impact on
clientelistic party behavior than in states in which the municipalities have smaller
public administration’s size. On the other hand, when the private sector’s size is
above .05, states in which the municipalities have bigger public administration’s size
have more impact on clientelistic party b ehavior than in states in which the
municipalities have smaller public administration’s size.
Following from the previous explanation, given similar supply of clientelistic
votes in municipalities with small private sector’s size, more public resources appear
to have less impact on clientelism than less public resources. It could be the case that
having more public resources would allow for more investments in public goods,
thus offsetting the necessity to resort to clientelistic politics. This situation is
perfectly plausible as having more public resources would allow for the natural
expansion of public goods and services thus increasing the number of people
employed in the public sector through legitimate public entrance exams.
This relationship, however, is more evident to lower and higher values of the
private sector’s size as Figure 4.2 shows.
Figure 4. 2: Plot of Size of Private Sector on Size of Public Sector
As Figure 4.2 suggests, the correlation between medium-to-high values of the
size of both public administration and private sector is non-existent. The scatterplot
also shows an outlier, which refers to the small state of Roraima in the year of 1998
and is explained by the fact that the only data available for that year was the state
capital. Thus, differently from the other states, it does not reflect the average of the
municipalities.
Figure 4. 3: Predictive Margins of the Effect of the Public Administration’s Size on
Clientelism (Herfindahl) by Higher, Median, and Lower Values of the
Private Sector’s Size
While Figure 4.1 considers the effect of private sector’s size on clientelism
moderated by size of public administration, Figure 4.3 inverts it to analyze the effect
of public administration’s size on clientelism moderated by private sector’s size. In
this sense, Figure 4.3 shows that the median effect of the public administration’s size
on clientelism decreases as the private sector’s size increase. However, this
relationship is different for the higher (95th percentile) and lower (5th percentile)
values of the private sector’s size. Indeed, it shows that the effect of the public
administration’s size on clientelism decreases for lower values of private sector’s
size increase, while it increases for higher values of the size of private sector.
In addition, it appears that private markets require more public resources in
the dispensation of clientelistic politics. This analysis is consistent with reports from
politicians who claimed during the field research interview that votes were
becoming more expensive because there were fewer residual clientelistic votes to be
bought (A. Imbassahy, personal communication, August 14, 2013).
On a different note, contrary to expected, clientelism does not appear to be
sensitive to party ideology as it is consistently not statistically significant throughout
the models. It may be the case that parties’ realignment in the last two decades,
which resulted in movements across the political spectrum, may have confounded
any correlation. In fact, Power and Zucco’s (2012) waves of surveys measuring
partisan ideological placement show a substantial variation within the parties.
Additionally, it has been indicated that parties are to some extent engaging in both
clientelistic and ideological politics (Strom 1990). Indeed, this claim is confirmed in
the field research interviews where elected officials cautiously (e.g. J. Neto, personal
communication, July 25, 2013) or openly (e.g. L. Lima, personal communication,
August 27, 2013) suggest such practices.
The Cluster Index
The Herfindahl Index (HI) measure of electoral market concentration shows
some validity as a proxy for clientelism. But while this index captures the
concentration of party votes in an electoral district, which is consistent with party’s
clientelistic behavior, it does not take into account the contiguity of the
municipalities in which candidates prospect their votes, which could be related to
the strategies of more programmatic parties. According to Ames (2001), clustering,
or votes that are obtained in contiguous municipality, tends to generate candidates
who are more accountable to their constituents and less susceptible to local bosses.
In this sense, clustering could be a better measure of programmatic party’ strategy.
Since HI deals with concentration taking into consideration votes received in
municipalities of each state, the assessment of programmatic party behavior could
be improved further. In this sense, the HI approach can only reveal the degree to
which a party is more or less clientelistic (a function of the concentration of their
votes in the municipality), thus missing the programmatic behavior (a function of
clustering) to which this analysis now turns.
In order to refine the measurement of clientelistic party behavior, I propose
a Cluster Index, which is a new approach that takes into consideration the clustering
and dispersion of party votes. This measure consists of first sorting the vote share
for each candidate by municipality ranked according to its population size. Second, a
new binary variable is coded one if the cumulative sum of the vote share per
candidate by municipality is lower than .95 and zero otherwise. Third, a spatial
clustering algorithm is run considering the new binary variable described above and
then the number of clusters are iterated to find the best fit up to a limit of 10. Fourth,
after the optimum cluster solution is found, for each one of clusters it calculates the
proportion of votes the candidate received in that cluster (vp) and the proportion of
the population in each cluster (pp). Finally, the dependent variable is obtained by
calculating the sum of the product of the proportion of the votes received by each
candidate and the proportion of the population [mean product = sum (vp*pp)]. This
measure indicates the increases in dispersion, which is consistent with clientelistic
party behavior. The data is aggregated by state, party, and year.
The new variable, thus, represents a party clientelistic index and is
substituted in the previous discussed econometric model as described below:
CI = β0 + β1Ideology + β2Local_Taxes + β3Size_Public_Sector + β4GDP_Capita +
β5Urbalization + β6Log_Population + β7North + β8Northeast + β9South +
β10Centerwest + β11District_Magnitude +
β12Local_Taxes*Size_Public_Sector + e
Table 4.2 tests the same models as Table 4.1 but with the new variable cluster
index, which shows overall best fit for the models. Ideology shows the expected sign
but it still cannot support the common belief that rightist parties tend to be more
clientelistic. The variable private sector’s size is statistically significant (p<.001) for
all models but the effect on the parties’ clientelistic behavior changes direction with
only Models 1 through 3 showing the unexpected sign. However, in Models 4 and 5 it
is the expected direction and robust. The variable size of public administration’s size
shows similar behavior across the models except that it is not
statistically significant in Model 5.
Table 4.2: Clustering and Dispersion of Party Vote
DV: Clientelistic Party (Cluster Index)
Model 1 Model 2 Model 3 Model 4 Model 5
b/se b/se b/se b/se b/se
ideology_100 0.123 0.050 0.039 0.061 0.028
(0.00) (0.00) (0.00) (0.00) (0.00)
localtxyr 1.910*** 0.790*** 0.343** -5.745*** -4.518***
(0.11) (0.12) (0.12) (0.41) (0.39) sizepub
0.749*** 0.516*** 0.422*** -0.117** -0.032
(0.02) (0.02) (0.03) (0.04) (0.04) gdpcavg
0.009*** 0.013*** 0.022*** 0.032*** 0.023***
(0.00) (0.00) (0.00) (0.00) (0.00) urbanization
0.005*** 0.003*** 0.005*** 0.005***
(0.00) (0.00) (0.00) (0.00) logpop -
0.055*** -0.121*** -0.106*** -0.079***
(0.00) (0.00) (0.00) (0.00) north
-0.005 -0.025*
(0.01) (0.01) northeast
0.065*** 0.057***
(0.01) (0.01) south
0.009 0.022**
(0.01) (0.01) centerwest
0.006 0.036***
(0.01) (0.01) magnitude
0.005*** 0.004*** 0.003***
(0.00) (0.00) (0.00) c.localtxyr#c.size~b
15.758*** 12.862***
(1.03) (1.00) grandnorth
0.006
(0.01) constant -0.018 0.888*** 1.793*** 1.746***
1.383*** (0.01) (0.04) (0.06) (0.06)
(0.06)
R-sqr 0.424 0.523 0.585 0.622 0.591
dfres 2436 2434 2429 2428 2431
BIC -4062.3 -4503.7 -4806.1 -5024.4 -4860.0
* p<0.05, ** p<0.01, *** p<0.001
The variable GDP per capita and urbanization are both statistically significant
(p<.001) but in the unexpected direction in all models. However, the log of the
population is in the expected direction and shows a statistically significant (p<.001)
relationship with clientelism in all models. For each one percent increase in
population size, clientelistic votes decreases by .0011 units in Model 4. The regional
variables only slightly improve to the overall fit of the model as their introduction
increases the R-square from .523 in Mo del 2 to .585 in Model 3 and only the
Northeast region shows the predicted sign with a positive and statistically
significant relationship with clientelism (p<.001). In Model 4 they are all statistically
significant but again only Northeast has the predicted sign.
The full model (Model 4) improves the explanatory power substantially
compared to the same model tested with the Herfindahl Index in Table 4.1. The F
test indicates a good fit for the model, with F (12, 2428) = 332.53 and a significant
probability level (Prob > F = .000). The model explains 62.2% of the variance on
clientelism (R-squared = .622) compared to 40.7% in the equivalent HI model. The
strongest predictor in this model is private sector’s size (-5.745; p<.001). District
magnitude remains a positive, weak, and statistically significant predictor of
clientelism but its coefficient improves slightly from .001 in the equivalent HI model
to .004 in the current Cluster Index model. That is, for each unit change in the
district magnitude, an increase of 4% would be expected in clientelism.
Discussion of the Cluster Index
The Cluster Index improves on the validity of the measurement of clientelism
vis-a -vis the previously tested Herfindahl Index. Here I focus on the interaction
effect of the main predictors: private sector’s size and public administration’s size.
As with the HI models, Figure 4.4 also considers the 5th, 50th, and 95th percentiles of
public administration’s size. Overall, the effect of private sector’s size on clientelism
increases as public administration’s size also increases. However, the magnitude of
this effect changes for the lower and higher percentiles. When the public
administration’s size is below .31 the effect of lower values of the private sector’s
size (5th percentile) is higher on clientelism than the higher values of the private
sector’s size (95th percentile). This relationship inverts when public administration’s
size is above .31 with higher vales of the private sector’s size exerting a more
pronounced effect on clientelism.
Figure 4.4: Predictive Margins of Public Administration’s Size on Clientelism
(Cluster) mediated by Private Sector’s Size
Conversely, Figure 4.5 takes into consideration the effect of private sector’s
size on clientelism moderated by public administration’s size. It shows that the effect
of private sector’s size on clientelism decreases for lower and median values of
public administration’s size. However, this relationship is different for the higher
(95th percentile) values of public administration’s size. In this case, it shows that the
effect of private sector’s size on clientelism increases for higher values of public
administration’s size.
Figure 4.5: Predictive Margins of Private Sector’s Size on Clientelism (Cluster Index)
mediated by Public Administration’s Size
The analysis of Figure 4.5 suggests that private sector’s size is less conducive
to clientelistic politics in places where public administration’s size is at the median
or lower. Conversely, it appears that both private sector’s and public administration’s
sizes work in tandem toward clientelistic politics. This further suggests that the
more advanced the municipality is, the mo re public resources are necessary to feed
the clientelistic demands of the electoral market. This argument is also supported by
the interpretation of Figure 4.4 in which, to a higher or a lesser degree, the effect of
public administration’s size on clientelism mediated by private sector’s size is always
positive for the three percentile levels considered.
Now that we have analyzed the validity of the statistical models explaining
the clientelistic behavior of the political party system, we can turn to the question of
how well these models explain political parties. In order to answer this question, I
plot the distribution of the 32 existing Brazilian political parties during the period of
1998 to 2010 aggregated by state and year and compared them according to the
mean value of the Herfindahl Index and the Cluster Index, ordered by the latter.
Figure 4.6: Clientelistic/Programmatic Positions of the Political Parties by
Herfindahl and Spatial (Cluster)
According to Figure 4.6, the Herfindahl plot does not show a discernible
pattern when compared to the spatial plot (Cluster Index), which is our reference
model. The spatial plot, however, yields interesting findings. It is notable that the
distribution of the political parties according to Cluster Index tends to support the
literature, which claims that parties tend to be more clientelistic as they move to the
right of the political spectrum. In fact, if we consider the clusters of the dot plot
distribution, five distinct clusters can be identified. The first one includes the leftist
parties Brazilian Communist Party (PC do B), the Workers’ Party (PT), and the
Popular Socialist Party (PPS), which are closer in the lower values of the Cluster
Index. The second cluster shows the Brazilian Democratic Movement Party (PMDB),
the Green Party (PV), the Brazilian Socialist Party (PSB), and the Socialism and
Freedom Party (PSOL). The third cluster includes the Brazilian Social Democracy
Party (PSDB), the Democrats (DEM), and the Brazilian Labour Party (PTB). The
fourth cluster shows parties such as the Christian Social Party (PSC), the Party of the
Republic (PR), and the Christian Social Democratic Party (PSDC). The last cluster
includes parties in the higher value of the Cluster Index such as the more rightist
parties the Progressive Party (PP), the Progressive Republican Party (PRP), and the
now extinct the Party of the Reconstruction of the National Order (PRONA). Two left
leaning parties also unexpectedly appear in this last cluster: the Democratic Labour
Party (PDT) and the Workers’ Cause Party (PCO). Overall, this finding offers
additional evidence supporting the vali dity of the clientelistic measure of party
behavior as assessed by the spatial distribution of votes (Cluster Index).
The position of the political parties on a two-dimension plot of political
linkage (left/right and clientelistic/programmatic) in Figure 4.7 shows an
interesting picture. The plot uses the Power and Zucco (2009) scale to position the
parties on an ideological dimension. The clientelistic dimension is derived from the
Cluster Index and re-scaled for better comparison with the ideological one. Although
the parties are located in each of the specific quadrants (1-right/clientelistic,
2left/clientelistic, 3-left-programmatic, and 4-right/programmatic), we cannot tell
what the exact score is in which a party transitions from being programmatic to
clientelistic, for instance. Still, the distribution of the parties yields interesting
findings. Overall, the parties fall into the expected quadrants.
Figure 4.7: Political Parties Position on a Clientelistic/Programmatic and
Left/Right Dimension
Following the literature that associates leftist parties with programmatic
politics and rightist parties with clientelistic politics, we would expect to find most
of the parties falling along the 1st and 3rd quadrants, which is the case. The whole
picture is more complex and some unexpected results also surface. In the
right/clientelistic quadrant (1st), for instance, small and median parties such as the
extinct ultra-right Party of the Reconstruction of the National Order (PRONA), which
merged with the Party of the Republic (PR) in 2006 (also located in the same
quadrant), and the Progressive Party (PP). In the left/programmatic quadrant (3rd)
are parties such as the Brazilian Communist Party (PC do B), the Workers’ Party
(PT), and the Socialist People’s Party (PPS). In the left/clientelistic quadrant (2nd)
there are only three parties: the Workers’ Cause Party (PCO), the Unified Socialist
Workers’ Party (PSTU), the Democratic Labour Party (PDT), which is unexpected
given the historic programmatic appeals of these parties. The case of the PDT is
particularly interesting and it is possible that its position is a result of shifts
undergone in the last 10 years following change in leadership. As a long time PDT
representative puts it:
The political platform of my party is non-existent. They allow state
compositions [alliances or coalitions] with no criteria. Our party is no longer a
reference [of an ideological party] since Brizola has passed away. It is a fact
that PDT was created to fulfill his [Brizola] goal to become president of Brazil,
since he could not gain control of the PTB brand after the redemocratization
period. He had fame as being a “caudilho” but he was democratic and when he
would lose the internal discussion he would concede. We used to discuss
national politics, without prioritizing local or regional interests. (Miro Teixeira,
Personal Interview, August 27, 2013)
In the right/programmatic quadrant (4th) fall parties like the Brazilian Social
Democracy Party (PSDB), the Brazilian Democratic Movement Party (PMDB), and
the Democrats (DEM). While the PSDB is consensually recognized as a programmatic
party, PMDB is known for its lack of ideological appeal or as a fisiológico (tit-for tat)
party in the words of a former party representative (Jose Dias, Personal Interview,
July 31, 2013). The DEM (former PFL) is known for its nonideological approach to
politics; as one DEM party representative suggests, there is not an ideological party
linkage but a personal appeal to voters (Paulo Azi, Personal Interview, July 23,
2013).
Conclusion
Clientelistic political linkages have received considerable attention recently
by scholars who have moved beyond the anthropological or sociological
understanding of the mechanisms that mediate the relationship between patrons
and clients. More specifically, scholars have focused on clientelism as an electoral
strategy of politicians or political parties. However, the challenge remaining is
related the elusiveness of this concept to measurement, as it practices posit ethical
or even illegal concerns related to vote buying, patronage, and deliverance of
personal benefits to the detriment of the collective good. This chapter purported to
devise a new measurement for clientelism based on the distribution of votes in
electoral districts. I argued that competitive elections in multiparty proportional
electoral systems compel parties to seek votes either in public electoral markets,
where voters are more sensitive to ideological appeals, and private electoral
markets, where voters are more inclined to clientelistic linkages.
The demand side of the electoral markets will drive the strategies of the
political parties in their task to gain seats. In this sense, electoral public markets will
demand more clientelistic parties, while private electoral markets demand more
programmatic parties. Parties then compete and the outcome of the election reflects
the strategies adopted according to the programmatic -clientelistic behavior of each
party. In this sense, more clientelistic parties would tend to have more concentrated
votes as they recruit local political bosses who in return deliver bulks of votes to
them in the municipality, while more programmatic parties would show less
concentrated votes.
In order to test this theory I use the Herfindahl Index (HI), a general measure
of concentration usually applied to measure monopolist behavior by companies, and
the Cluster Index, which improves on the previous measure to account for the
clustering of votes according to the contiguity of the municipalities. In this sense, the
HI calculates the share of votes of each candidate in each municipality, and then
integrates it by party year in the electoral district, which coincides in the state in
Brazil. After calculating the HI for each party, I then test for the validity of this
measurement by running regression models with HI as the dependent variable
against four main independent variables (ideology, local taxes, size of public sector,
and GDP per capita) and other control variables. Due to the limitations of the HI
assessment of programmatic party behavior, I then use the same statistical models
to test the validity of a more refined Cluster Index, which aggregates the distribution
of votes in optimum clusters to calculate the dispersion of by party, year, and state.
Party ideology has been associated with clientelistic party behavior. It is
suggested that in Brazil more rightist parties tend to be more clientelistic. However,
the statistical models failed to confirm this hypothesis as the sign changes direction
and it is not statistically significant. The modernization theory, which suggests that
as GDP per capita, urbanization, and population size increases clientelism decreases,
is partially confirmed.
These variables suggest that as municipalities modernize, parties become less
clientelistic, though only the (log) size of population, which are demographic
variables operating in the same logic of the modernization theory, consistently
confirms this throughout the Cluster models. This indicates that as the size of the
population increases clientelism decreas es. The results for the regional variables are
mixed with only the Northeast confirming the prediction for the Cluster models. I
also grouped the most similar regions together dividing them into North
(grandnorth) and South, which is statistically significant and in the predicted
direction. This study also confirms previous findings as the Cluster models predict
that clientelism increases with the size of the district magnitude.
The results concerning the main variables of interest are mixed. In the case of
private sector’s size, the hypothesis that a more robust local private sector tends to
make its citizens less susceptible to clientelistic politics is only partialy confirmed.
In the Cluster models, it is statistically significant throughout the models but only
Models 4 and 5 show the predicted direction.
In the case of the public administration’s size, the hypothesis that its
increased size relative to the size of the private sector creates more opportunities for
clientelistic politics since the local population becomes more dependent on scarce
public resources is partially confirmed in the Cluster models. In this case, it is
statistically significant for Models 1 to 3 and in the predicted direction but in Models
4 and 5 it changes direction and/or loses statistical significance.
However, the most interesting finding is related to the interaction effect of
both public administration’s size and private sector’s size on clientelism. In this
sense, for the Herfindahl models it appears that the effect of the private sector’s size
on clientelism is higher when the public administration’s size is bigger. Conversely,
this effect is lower when the public sector’s size is small but, in this case, lower
values of the public sector’s size has more impact on clientelism than higher values.
From a different perspective, the effect of public administration’s size on clientelism
increases for higher values and decrease for median and lower values of private
sector’s size.
Comparing the Herfindahl and the Cluster models, it is possible to conclude
that the effect of private sector’s size on clientelism in unequivocally positive and
robust for upper values of public administration’s size. However, the same effect for
median and lower values of public administration’s size is either less pronounced for
the predictive margins of the Herfindahl model or negative for the predictive
margins of the Cluster models. This suggests that in larger private markets the
residual pool of clientelistic votes is smaller, thus requiring more public resources
for the dispensation of patronage, while in smaller private markets the supply of
clientelistic votes is higher, thus reducing the among of public resources necessary
for patronage. Likewise, the effect of public administration’s size on clientelism is
unmistakably positive for the upper values of private sector’s size for both
Herfindahl and Cluster models. However, in the first model, the median and lower
values of private sector’s size are negative while positive but less robust for the
second model. In sum, more developed municipalities have a smaller pool of
clientelistic votes that are crucial to win competitive elections mainly contended
among programmatic parties. As a consequence, they require more public resources
to feed the clientelistic machine.
In the next chapter, I will integrate ideology and clientelism as concurrent
dimensions of political competition. The question we will seek to answer is whether
pre-electoral coalitions can be explained as function of both formal (ideology) and
informal (clientelism) institutions.
CHAPTER 5 Ideological and Clientelistic Determinants of Electoral
Coalitions in Brazil
This chapter aims to integrate two concurrent dimensions of political
competition (ideology and clientelism) to unveil the patterns of political competition
in Brazil. The question we seek to answer in this chapter is whether pre-electoral
coalitions can be explained as a function of both ideolog ical and clientelistic
linkages. I define coalitions as associations of political parties with the objective to
stand elections. Previous works have mainly dealt with formal institutions that link
voters to parties through policy. To be sure, recent scholarship has paid
considerable attention to informal institutions of which clientelistic politics is an
important component (e.g. Kitshelt and Wilkinson 2007, Nichter 2010).
However, these studies tend to treat ideology and clientelism independently.
This chapter considers both ideology and clientelism as integral and
simultaneous factors in the electoral competition process. For this purpose, I have
developed the combined utility theory, which argues that polarized competitive
elections in modernizing national electoral markets constrain programmatic parties
to coalesce with clientelistic parties to gain access to regional private electoral
markets. This dichotomy is the result of uneven socio-economic changes that make it
possible for parties to coexist in both public and private electoral market domains.
Programmatic parties may avoid direct engagement in clientelistic politics, as these parties
tend to be more accountable to their voters. Alternatively, they will form electoral coalitions
with clientelistic parties as an indirect strategy to increase their vote share.
In order to test this theory, reliable measures of party ideology and
clientelism placement on the political spectrum are necessary. The ideology measure
is drawn from Power and Zucco (2012) and its reliability is discussed in Chapter 3.
The clientelism measure is described in Chapter 4. In this chapter I will assess the
validity of these measurements while testing the extent to which these two factors
explain the patterns of electoral coalitions.
The working hypothesis is that parties preserve their ideological coherence
by forming electoral coalitions with ideologically proximate parties. In this case, a
negative relationship between explanatory and outcome variables is expected. That
is, as the ideological dispersion of the parties increases the probability of forming
coalitions decreases. On the other hand, i n order to gain access to additional votes
necessary to win competitive elections, programmatic parties will engage
clientelistic ones. In this case, a positive relationship is expected, suggesting that as
the clientelistic dispersion of the parties increase so does the probability of forming
coalitions.
This chapter is organized as follows. The first section presents the data and
measurements. The second section introduces the model and discusses the main
findings. The final section concludes.
Data and Measurements
The test of the hypotheses associated with coalition strategies will be
conducted through data collected from Superior Electoral Tribunal’s (TSE) website
on four electoral cycles between 1998 and 2010 for the Chamber of Deputies for
each of the 26 Brazilian states. The data was collected from Superior Electoral
Tribunal’s (TSE) website. The selection of the electoral period is due to data
availability as the levels of aggregation and organization of the data preceding the
1998 elections are i ncomplete for all states. Despite some limits in coverage, the
dataset is representative of the crucial cases involving two polarizing parties we set
up to analyze. The use of data relative to the Chamber of Deputies is justified
because it allows for better comparison among the various political parties
nationalwide. For instance, not all parties run candidates for president, governor, or
senator and the race for state-level Chamber of Deputies positions may be subject to
very particular local demands. The data is aggregated by state, year, and party and
comprises 5,823 observations.
Dependent Variable
Electoral Coalition
Parties coalesce electorally to win elections and, in seeking this goal, they
devise strategies that lead them to such an objective. Studying parties’ electoral
objectives and strategies, scholars are able to explain in general the factors that
bring parties together in electoral coalitions (e.g Golder 2005, 2006). This
dissertation focuses on polarized competitive elections, which suggests that parties
resort to other dimensions of political linkage in order to obtain the differential
votes for the victory on election day. In this sense, our main concern refers to the
coalition formation around polarizing parties without neglecting the role that the
remainder parties play in the overall electoral coalition process. In this sense, the
dependent variable consists of a group of variables that considers the electoral
coalition for each party independently in any given electoral year. In this case, each
party corresponds to a binary variable that is coded one for every time another
given party forms a coalition with it and zero otherwise. For instance, the variable
PP_coaltition is coded one when other parties form a coalition with the Progressive
Party (PP) for the elections years between 1998 and 2010 and zero otherwise.
Independent Variables
Ideology
The literature suggests parties are attracted to ideologically proximate
parties when forming coalitions (Golder 2006). In order to test whether this is the
case – whether ideology matters in coalition formation – the model tests for the
effect of ideological distance between parties. The variable is based on Power and
Zucco’s (2007) measure of ideological placement over time, which accounts for
variations within parties across time. T he parties are placed on a continuum that
varies from one (most leftist) to 10 (far rightist). Further, I calculate the ideological
distance, which is the absolute difference between a party’s ideology and the
ideology of all other parties, which is a measure of dispersion both to the left and to
the right of the coalition formed by any given party. In this case, it would be expected
that as the ideological distance increases, the probability of forming a coalition
would decrease.
Clientelism
Polarized national electoral competitions compel programmatic parties to
seek additional non-ideological votes in bailiwicks. This suggests that these parties
will resort to clientelistic parties by means of coalitions to gain access to voters who
are non-responsive to programmatic appeals. The literature suggests that leftist
parties tend to be more programmatic while rightist parties are more clientelistically
inclined (Montero 2010). Accordingly, it would be expected that the variable
clientelism would be positively and statistically correlated with rightist parties and
negatively and statistically correlated with leftist parties. However, I make no
assumption about the programmatic nature of the political parties as far as their
position in the political spe ctrum goes. One field research interview suggests though
that parties create coalitions irrespective of ideological commitments (J. Neto,
personal communication, July 25, 2013). If this holds true across the board, I would
expect that as parties’ clientelistic dispersion increases, so does the
probability of forming electoral coalitions.
As discussed in Chapter 3, the measure for clientelism is based on sorting the
vote share for each candidate from biggest to smallest by municipality, weighted by
its population size. Subsequently, a variable is coded one if the cumulative sum of the
vote share per candidate per municipality is lower than .95 and zero otherwise. For
instance, if the sequence of vote share is .3, .25, .15, .1, .1, .05, .02, .02, and .01, the
cumulative sum is .3, .55, .7, .8, .9, .95, .97, .99, and 1. In this case the new variable is
coded 1, 1, 1, 1, 1, 1, 0, 0, 0. After that, a spatial clustering algorithm is run
considering this dummy variable; then, the number of clusters is iterated to find the
best fit up to a limit of 10. After the optimum cluster solution is found, it calculates,
for each cluster, the proportion of votes the candidate received (vp) and the
proportion of the population (pp). Further, the dependent variable is obtained by
calculating the sum of the product of the proportion of the votes received by each
candidate and the proportion of the population [mean product = sum (vp*pp)]. This
measure indicates the increases in dispersion, which is consistent with clientelistic
party behavior. Finally, like the ideology distance, I also calculate the clientelistic
distance, which is the absolute difference between a party’s measure of clientelism
relative to all other parties’ measure of clientelism.
Control Variables
Party Size
In proportional elections, parties first need to overcome the electoral
quotient to seat any representative. The electoral quotient is obtained by dividing
the number of votes by the number of seats. Since smaller parties are less likely to
reach this quotient, you would expect th at small parties would coalesce with larger
parties in order to gain representation in the Chamber of Deputies. Thus, it would be
expected that when the big parties are considered as dependent variables, as the
size of the parties increase so does the probability of forming coalitions. Since we are
arbitrarily picking the parties according to their size, the inverse is true for small
parties; that is, as the size of the parties increases the probability of forming
electoral coalitions also increases. The size of the party is based on two
measurements: party size seat, or the number of seats for each party in the House of
Representatives, and the party size vote, or the number of votes each party receives
in the election.
Presidential Coattail
In order to test the possibility that presidential politics influences
subnational legislative elections (Brambor et. al. 2006), I include a dummy variable
that captures the legislative coalition with the party in power in the previous
election year. The presumption is that parties seek coalitions with the party in
power so as to gain access to federal resources. Thus, I would expect a positive and
significant effect of the variable coalition coattail on the dependent variable party
coalition. Since only two parties (PT and PSDB) have governed during the period
considered, the test of this variable is only plausible in models that consider the
electoral coalitions for both the PT and the PSDB parties.
District Magnitude
In the open list with single vote ballot systems, the district magnitude size has
been associated with appeals for personal votes (Shugart and Carey 1995). These
systems increase intraparty competition, thus requiring candidates to separate
themselves from other parties within the same party. Thus, inclusion of this variable
tests the hypothesis that the larger the district, the more likely the dispute for
clientelistic votes. Since the party brand is a constant for candidates, the only way
they can distinguish themselves is by means of personal vote appeal (Shugart and
Carey 1995). In this sense, the magnitude of the district is a proxy for
competitiveness, in which case it would suggest that the higher the district
magnitude the higher the competitiveness and therefore the necessity to form
electoral coalitions.
Model and Findings
In order to test the hypotheses that ideology and clientelism are both at work
when forming political coalitions, I use a logistic model that treats each party
coalition individually. In this case, each mo del consists of a dependent variable that
specifically tests all possible coalitions formed by one party with another party. For
instance, PP_coalition will be coded one every time that any other given party forms
a coalition with PP (Progressive Party) and zero otherwise. The parties of prim ary
interest for this analysis are the Wor kers’ Party (PT) and the Brazilian Social
Democratic Party (PSDB) as those parties have polarized the national elections si nce
1994.
However, in the subsequent analysis several other parties will be considered
according to their size and distribution on the political spectrum. Indeed,
competitive elections are an important component of the theory developed in this
dissertation, which suggests that as parties exhaust their ideological appeal, they
will access other dimensions of political linkages. In this sense, one of the main
objectives of this study is to assess the validity of the measurements of the two
dimensions of political competition – ideology and clientelism -- to unveil how these
dimensions interact to produce the electoral outcome desired by the political
parties.
The econometric model of the coalition hypothesis is shown below:
Party_Coalition = β0 + β1 Ideology + β2Clientelism + β3Party_Size_Seat +
β4Party_Size_Vote + β5District_Magnitude +
β6Presidential_Coattail + β7(Log)Population + β8GDP_per_Capita
+ β9Urbanization + β10Verticalization + β11Region + e
Table 5.1 shows the estimation of eight logistic regression models explaining
coalition formations for selected parties. Besides the main parties of interest, the PT
(big, programmatic, and center-left) and the PSDB (big, programmatic, and
centerright), I also include other parties that are representative of both size and
position on the clientelistic-ideological spectrum, such as the PMDB (big,
programmatic, and center-right), the DEM (big, programmatic, and right), PSB
(median, programmatic, and center-left), the PTB (median, programmatic, and
center-right), the PMN (small, programmatic, and left), and the PRP (small,
clientelistic, and right) (see Figure 4.7 in Chapter 4 to visualize party placement).
Overall, the explanatory variable ideology, which measures the dispersion of
coalitions in both directions, is remarkably consistent. It show a robust and
statistically significant (p<.001) relation for all but one party (PMN). The overall
negative correlation with the dependent variable indicates that as the absolute
difference of the parties’ ideology increases (a measure of dispersion), the likelihood
of forming a coalition with a given party decreases. In other words, it indicates that
coalitions tend to be formed with parties that are ideologically proximate. This
finding represents strong evidence that ideology indeed matters in coalition
formation.
I also test the effect of party ideology over time to account for party shifting
along the political spectrum as suggested in the literature (e.g. Hunter 2007, 2010,
Power and Zucco 2009, Samuels 2004). To test this effect, I run restricted models for
the elections of 2002, 2006, and 2010 (Table 5.8, Appendix C). In general, it shows
that the effect of ideology decreases with the exception of the PSDB, for which this
effect increases, and the PRP that remains stable. The effect of ideology in three
other parties (PSB, PMN and PTB) is not statistically significant. I repeated the
process with restricted models for 2006 and 2010 (Table 5.9, Appendix C) and the
overall effect of ideology decreases even further in comparison with the previous
models (Table 5.8, Appendix C), with the exception of the PSB, for which the
coefficient regains statistical significance with a higher value that the fully specified
model. The effect of ideology on PT, the PMDB and the PTB loses statistical
significance and the PMN increases while also gains statistical significance. The
103
effect of ideology on the PSDB decreases in comparison to the first restricted model
but it is higher than the fully specified model (Table 5.1).
Table 5.1:
Party’s Electoral Coalition Formation (1998-2010)
DV: Party’s Electoral Coalition Formation
PT PSDB
PMDB
DEM
PSB
PTB
PMN
PRP
b/se b/se
b/se
b/se
b/se
b/se
b/se
b/se
ideology
-0.224*** -0.211*** -0.171*** -0.271*** -0.144*** -0.096*** 0.013 -0.173***
(0.03) (0.05)
(0.05) (0.04) (0.03) (0.03) (0.02) (0.03)
clientelism
-1.026 1.206*
1.101 0.990 0.181 1.089* -0.398 1.523**
(0.59) (0.58)
(0.60) (0.74) (0.56) (0.47) (0.55) (0.53)
size_vote_100
0.002 0.003
-0.003 -0.002 0.001 -0.002 0.001 -0.007***
(0.00) (0.00)
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
size_seat
-0.012*** 0.006**
0.006** 0.005* -0.003 0.009*** 0.007** 0.002
(0.00) (0.00)
(0.00) (0.00) (0.00) (0.00)
(0.00) (0.00)
magnitude
-0.029 -0.047*
0.041 0.030 -0.023 0.025
-0.003 0.102***
(0.02) (0.02)
(0.02) (0.02) (0.02) (0.02)
(0.02) (0.02)
population (log)
-0.147 -0.155
-0.207 -0.280 0.015 -0.035
-0.403** 0.058
(0.14) (0.12)
(0.14) (0.16) (0.13) (0.13)
(0.15) (0.14)
gdp_per_capita
0.033 -0.060
0.242*** -0.106 0.185** -0.017
0.181** -0.026
(0.06) (0.06)
(0.06) (0.08) (0.06) (0.06)
(0.06) (0.07)
urbanization
-0.078** -0.002
-0.020 -0.042 -0.070** 0.033
-0.045* -0.008
(0.03) (0.02)
(0.02) (0.03) (0.02) (0.02)
(0.02) (0.02)
verticalization
-0.094 0.465*
0.461* 0.023 0.850*** -0.012
0.314 -0.289
(0.19) (0.18)
(0.19) (0.20) (0.19) (0.20)
(0.21) (0.25)
north
-0.199 0.031
-1.254*** -0.389 -0.504* 0.194
0.058 1.089***
(0.18) (0.19)
(0.22) (0.27) (0.20) (0.18)
(0.19) (0.22)
southeast
-0.353 0.065
-1.841*** 0.151 -0.864* -0.656
-0.838* 0.399
(0.35) (0.35)
(0.39) (0.47) (0.36) (0.38)
(0.36) (0.66)
south
-0.894* 0.126
-2.052*** 0.209 -1.692*** -0.691
-1.792*** 0.610
(0.37) (0.37)
(0.40) (0.50) (0.39) (0.38)
(0.41) (0.45)
centerwest
-0.235 0.136
-1.023*** 0.411 -0.208 0.144
-0.378 0.925***
(0.25) (0.24)
(0.26) (0.31) (0.26) (0.23)
(0.26) (0.26)
constant
2.242 1.186
0.933 3.419 -1.586 -1.163
3.761 -2.681
(2.00) (1.79)
(2.01) (2.31) (1.96) (1.85)
(2.12) (1.93)
N
2361
2361
2365
1535
2244
2293
2133
1633
BIC
2192.0
2136.5
1969.8
1333.9
1961.5
2111.0
2038.1
1625.3
* p<0.05, ** p<0.01, *** p<0.001
104
The effect of clientelism on coalitions does not change substantially when the
restricted models are considered (Table 5.3). It is worth mentioning that the PT i s
only statistically significant for the 2002 period and subsequent elections. A further
restriction of the model for the 2006 and subsequent elections show that the PT
loses statistical significance just as in the fully specified model. The same pattern
also applies for PMDB. The effect of clientelism on the PSDB coalitions, on the other
hand, is statistically significant for the fully specified model but it loses statistical
significance in the two remaining restricted models. The same applies for the PTB.
Conversely, the DEM, the PSB, and the PMN are not statistically significant in the
remaining three models. Finally, the PRP remains statistically sig nificant throughout
the models.
When both dimensions of political linkage are analyzed for the whole period
considered in this study, it appears that the PSDB is slightly more consistent on its
coalition formation on the ideological dimension than is the PT. At the same time, the
PSDB appears to extend further on the clientelistic spectrum when forming
coalitions, as compared to the PT. In fact, as Table 5.2 indicates, for the 1998-2010
period, the standard deviation of the ideology of the parties under the PT’s coalition
is bigger than that of the PSD B, suggesting that the PSDB’s electoral coalitions were
less dispersed than that of the PT. The dispersion of the PT’s coalitions decreased for
the 2002-2010 period and decreased even further for the 2006-2010 period but is
still more dispersed than is the PSDB ideological partners. The PSDB’s ideological
coalition dispersion also went down when the period was restricted to 2002-2010,
went slightly up for the 2006-2010 period though it remained lower than the
105
19982010 period. This finding suggests that the PT’s coalition strategy has been
more aggressive in maintaining ideological coherence when compared to the PSDB’s
strategy. It may be the case that the PT still remains the preferred partner of the
center-left, even as the party has shifted over time to the center in order to enlist
new allies and become a more credible political party after Lula’s presidential
electoral defeat in 1998 (Samuels 2004, Hunter 2007, 2010).
Table 5.2: Standard Deviation for Party Ideology and Clientelism
Ideology
Clientelism
PT
PSDB
PT
PSDB
1998-2010
(PSDB in
Executive)
1.919837
1.094385
.0830145
.091273
2002-2010
(PT in
Executive)
1.841787
1.016936
.0843211
.0892931
2006-2010
(PT in
Executive)
1.656561
1.036928
.0898822
.0914972
Conversely, looking into the clientelistic-programmatic dimension of party
competition, the logic changes on the clientelistic dimension. In this case, the PSDB
appears to reach further for partners than the PT is willing to do. Although the PT’s
clientelism standard deviation increases, as the period is restricted to capture
changes over time, it remains smaller than the PSDB. In sum, considering the
clientelistic-programmatic and left-right spectra the PT’s coalition partners appear
to be less consistent ideologically but more consistent programmatically consistent,
while the PSDB coalition partners are the opposite: more consistent ideologically but
less consistent programmatically.
106
In order to compare the relative importance of ideology and clientelism for
the Workers’ Party (PT) and the Brazilian Social Democracy Party (PSDB) over time,
Table 5.3 restricts the models for three electoral periods. The analysis shows that
over time the importance of ideology for the PT decreases to the point of even
becoming not statistically significant when the model is restricted to the elections of
2006 and 2010. In other words, the Workers’ Party coalitions become more
heterogeneous over time. As for the PSDB, the effect of ideology increases in 2002;
that is, the coalitions become more homogenous, and the effect of ideology
decreases in 2006 but it is still more homogenous than in 1998.
One explanation for this overall pattern may be that parties are not only
broadening their coalitions to capture as many votes as possible, but also that
political parties are themselves moving within the political spectrum. As Power and
Zucco (2012) demonstrate, both the PT and the PSDB show a clear shift toward the
center and to the right of the political spectrum, when compared to their original
position prior to governing. For the PSDB, their shift to the right happened in 1997
and the PT’s shift happened in 2005. T he effects of these shifts are different for each
party however, as it appears that the PSDB’s new coalitions become more
ideologically similar, while the PT’s coalitions appear to have become more
haphazard over time.
Table 5.3: PT and PSDB Electoral Coalition Formation (1998
and 2006-2010)
DV: Party’s Electoral Coalition Formation
-2010, 2002-2010,
PT PSDB PT__02 PSDB_02
PT_06 PSDB_06
b/se b/se b/se b/se
b/se
b/se
107
ideology -0.224*** -0.211*** -0.166*** -0.276***
-0.066
-0.246***
(0.03) (0.05) (0.03) (0.05)
(0.04)
(0.06)
clientelism -1.026 1.206* -1.339* 1.109
-1.219
1.024
(0.59) (0.58) (0.64) (0.67)
(0.72)
(0.75)
size_vote 0.002 0.003 0.003 0.008**
-0.002
0.009**
(0.00) (0.00) (0.00) (0.00)
(0.00)
(0.00)
size_seat -0.012*** 0.006** -0.011*** 0.007**
-0.005
0.004
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
magnitude -0.029
-0.047*
-0.042
-0.130***
-0.000
-0.166***
(0.02)
(0.02)
(0.03)
(0.04)
(0.04)
(0.05)
population(log) -0.147
-0.155
-0.205
-0.234
0.098
-0.313
(0.14)
(0.12)
(0.18)
(0.17)
(0.24)
(0.22)
gdp-per_capita 0.033
-0.060
-0.015
-0.193**
-0.125
-0.416***
(0.06)
(0.06)
(0.06)
(0.07)
(0.08)
(0.10)
urbanization -0.078** -0.002
-0.058*
0.007
-0.063
0.028
(0.03)
(0.02)
(0.03)
(0.02)
(0.04)
(0.03)
verticalization -0.094
0.465*
(0.19)
(0.18)
north -0.199
0.031
-0.208
0.286
-0.136
0.664**
(0.18)
(0.19)
(0.19)
(0.21)
(0.22)
(0.25)
southeast -0.353
0.065
-0.448
0.852*
0.292
2.127***
(0.35)
(0.35)
(0.39)
(0.41)
(0.55)
(0.58)
south -0.894*
0.126
-0.822*
0.882*
-0.246
2.516***
(0.37)
(0.37)
(0.41)
(0.42)
(0.59)
(0.62)
centerwest -0.235
0.136
-0.131
0.678*
0.154
1.477***
(0.25)
(0.24)
(0.27)
(0.30)
(0.37)
(0.40)
o.verticalization 0.000
0.000
0.000
0.000
(.)
(.)
(.)
(.)
constant
2.242
1.186
3.105
3.053
-1.030
5.310
(2.00)
(1.79)
(2.64)
(2.54)
(3.40)
(3.32)
N
2361
2361
1925
1925
1388
1388
BIC
2192.0
2136.5
1884.7
1705.6
1446.4
1345.3
* p<0.05, ** p<0.01, *** p<0.001
In this sense, a close analysis of the data seems to suggest that a re-shifting of
the political parties along the political spectrum in general (Power and Zucco 2009)
and of the PT in particular (Hunter 2007, 2010, Samules 2004) weakened the
importance of ideology in coalition formation. This argument is endorsed by the
House Speaker, who is a strong proponent of political reform, especially the
elimination of the proportional representation system. In a recent debate with two
political scientists he asserted that “[n]obody sits down to form coalitions looking
for people who think like you do. You sit down to form a coalition thinking about
how many votes the party has, what are the viable candidates in the other party who
108
will take votes away from you, what do I have to do to preserve my seat” (Cunha,
May 21, 2015).
The effect of the size of the party on coalitions, which assesses the hypothesis
that smaller parties coalesce with bigger parties to beat the electoral quotient,
seems to matter but only for the measure of party size accordi ng to the number of
seats in the House of Representatives (as opposed to the size measured by the total
votes received in the election). For the PT the relationship is negative, indicating that
this party tends to form coalitions with smaller parties, while the PSDB tends to
form coalitions with bigger parties. The effect of district magnitude on coalition
formation varies across parties and, although negative for both the PT and the PSD B,
it is only statistically significant for the latter.
The proxies to test for the effects of modernization (population, GDP per
capita, and urbanization) are all negative for the PT and the PSDB. This indicates that
these parties tend to form fewer coalitions in areas that are more developed.
The important caveat is that urbanization is only statistically significant for the PT.
This makes sense given the Worker’s Party’s longstanding history of working with
unions that are located in the most industrial parts of the country.
The effect of verticalization, or the legal imposition for parties to replicate in
the states the same electoral coalitions for president, on coalition formation varied
among the parties. This resolution predicted that parties could only reproduce
presidential coalitions for the other elections. The expectation was that parties
running presidential candidates would have more restricted electoral coalitions. Yet,
for the PT its effect is negative and not statistically significant. On the other hand, for
109
the PSDB the effect is positive and statistically significant, indicating that the
verticalization clause instituted for the 2002 election explains the PSDB coalition
formation.
With respect to region, the regional effects are mostly not significant though
the direction of the effect varies. Regional effects are negative for the PT and positive
for the PSDB. The South is only statistically significant for the PT. Using the
Northeast region as a baseline indicates that the PSDB is more likely to form
coalitions in all other regions, while the PT is less likely to form coalitions in other
regions as compared to the Northeast.
Conclusion
The question we seek to answer in this chapter is whether electoral
competition can be explained as a function of ideological preferences and
clientelistic strategies. Toward this aim, I integrated two levels of political
competition identified as major determinants of political linkage: ideology and
clientelism. After having described and tested the validity of the measures
associated with these two dimensions in Chapters 2 and 3, I have integrated these
two dimensions to test the combined utility theory, which suggests that competitive
polarized elections in developing democracies constrain parties to seek crucial votes
on a secondary clientelistic dimension when they have exhausted the main
ideological dimension of political linkage. This theory posits that programmati c
parties, stronger in public electoral markets, when competing in national polarized
elections will form electoral coalitions with clientelistic parties to preserve their
110
own party brand while garnering votes in private electoral markets, where
clientelistic parties tend to have more reach.
This study finds evidence that ideology linkages predominate when it comes
to determining electoral coalitions in the electorally competitive Brazilian case.
However, it also suggests that ideology has become a less important determinant of
electoral coalitions since the 2002 presidential election, which marked a definite
shift of the Workers’ Party from the left to the center-left of the political spectrum
and as a consequence repositioned the Brazilian Social Democrat Party from the
center-left to the center-right of the political spectrum. The analysis shows that as
ideology dispersion of the parties in a coalition increases, the probability of for ming
a coalition with the given parties decreases. This empirical finding corroborates the
reports from political elites during the qualitative field interviews conducted in the
states of Bahia and Rio Grande do Norte and in the Federal District during the
summer of 2013. The informants appear to share the common view that since the
2002 presidential election, ideology became less important in defining electoral
coalitions.
Regarding the effect of clientelism, on the other hand, the secondary
dimension does not yield conclusive evidence. The importance of this dimension is
that the direction and statistical significance changes depending on the party
considered, suggesting that the parties mostly rely on a wide spectrum of parties
when forming coalitions in the clientelistic dimension. Focusing specifically on the
two nationally polarizing parties, it appears that the PSDB counts on a wider range
of clientelistic parties in its coalitions than the PT does. Although the PT coefficient
111
is consistently negative it is not statistically significant for the whole period
considered. It becomes statistically significant following the 2002 presidential
election but it loses statistical significance for the period since the 2006 election. The
PSDB, on the other hand, is consistently positive for all the periods considered
but only statistically significant when the full model is specified (Table 4.3).
The dispersion of the ideology and clientelism coalition formation indicated
by the standard deviation for both parties (PT and PSDB) also shows interesting
results. On the ideology dimension, both parties seem to narrow the ideological
proximity of the parties in their respective electoral coalitions. However, this
coalitions is wider for the PT than its is for the PSDB, which indicates that the PT has
been more aggressive in its strategy to form coalitions by attracting more parties
from the extremes of the political spectrum when compared to the PSDB. On the
clientelistic dimension, however, the analysis inverts somewhat. In fact, the PT
appears to increase the spectrum of the parties comprising its coalition over time,
but it remains slightly lower then that of the PSDB, which remains more or less
constant over time. In sum, the PT has a wider ideological configuration of parties in
its electoral coalition, while the PSDB has a wider clientelistic configuration of
parties comprising its electoral coalition.
The analysis presented in this chapter is a stylized model that only considers
the decision of one party to coalesce with another party. In practice, parties may
need to weigh the benefits of entering coalitions with multiple partisans. Indeed,
smaller parties may decide not to join coalitions with bigger parties. This is the case
for parties that consistently garner one or two seats each election. As we will discuss
112
in the next chapter, smaller parties systematically avoid coalition with bigger parties
because their party quotient, or the seat that is reserved for each party within the
coalition, is low thus making it difficult for them to compete with other parties in the
same coalition. Rather, they prefer to join forces with other smaller parties.
113
CHAPTER 6 Politics in Practice: How Ideology and Clientelism
Works in Brazil
In order to shed light on the methodological findings from Chapters 3, 4, and
5, this chapter draws on field research and case studies in Brazil to capture the
dynamics of electoral politics in a modernizing setting. Field research in two
northeastern states, a region known for its historic reliance on clientelistic politics
but that have nevertheless seen the rise of programmatic partisan politics, allows us
to examine when, why, and how partisan coalitions form. Rather than infer partisan
preferences based on patterns of electoral dispersion, for example, we can ask
partisan operatives to explain why they are willing to align with other parties.
The research for this chapter draws on case studies of Bahia and Rio Grande
do Norte. These states share cultural similarities, demographic characteristics, and
have historically been dominated by oligarchies. Both have also undergone shifts in
the long-lasting pattern of political competition with the ascension of opposing
leftist parties to power more recently. The incursions of the left in Rio Grande do
Norte started earlier in 2002 and lasted two consecutive terms when the right
regained power in the 2010 election. In Bahia, on the other hand, the left only gained
power in 2006 and was re-elected in 2010. While the left has won a third mandate
in 2014, the right has gained the capital, Salvador. The influential capital city elected
Anto nio Carlos Magalha es Neto (DEM), the grandson of a powerful political boss in
the state of Bahia, as mayor in 2012 making him a very strong
contender for the governorship in future elections.
114
These two states are representative of a wave that brought to power more
programmatic, mostly leftist, parties in a region that has historically been
characterized by the dominance of oligarchic groups linked to clientelistic parties,
usually on the right of the political spectrum (Montero 2010). Virtually all states in
this region had more pragmatic parties replacing clientelistic ones since 1988. For
instance, the PDT won the 2006 election beating the Sarney machine for the first
time since 1960 in Maranha o; the PT won in Piauí in 2002 and Sergipe in 2006; the
PSB won in Alagoas in 1998 and in Pernambuco in 2006, both replacing the PMDB,
and also defeating the PSDB in Ceara this same year; finally, the PSDB won
in Paraí ba in 2006 replacing the PMDB.
Why did the right manage to return to power in Rio Grande do Norte while in
Bahia they seem to be slow to recover their strength? Is this a function of their
ability to form a regionally autonomous electoral coalition? Or, is it likely that
national polarization influences state-level electoral coalitions as the quantitative
models in Chapter 5 suggest? To answer these questions, I conducted elite
interviews with federal congressmen, state representatives, mayors, rank-and-file
party members, journalists, and one vice-governor to investigate the mechanisms
under which parties’ competition strategies are devised in competitive multiparty
polarized elections. I then investigate political linkages on two dimensions --
ideology and clientelism -- and examine how these two seemingly opposite
dimensions are integrated through coalitions with the purpose to win elections.
Before delving into the findings, I first provide some historical background
information on the case studies.
115
Rio Grande do Norte
The state of Rio Grande do Norte is situated in the Northeast. It has an
estimated population of about 3.3 million distributed along 167 municipalities. The
Gini index is .49 and the Municipal Human Development Index is .684 (IBGE). The
recent political history of Rio Grande do Norte has been marked by polarization
between two families: Alves and Maia. The legacy of the Alves’ family starts with
Aluí sio Alves who began his political career in the Rural Democratic Union (UDN), a
rightist party that opposed the Brazilian Labor Party (PTB) of president Getu lio
Vargas and subsequently Juscelino Kubitsheck. However, he consolidated his
leadership after being elected governor in 1960 through the Social Democratic Party
(PSD), after he had served four consecutive mandates as congressman.
Subsequently, the vice-governor, Walfredo Gurgel, was elected governor in the 1965
electoral dispute with Alves’ support.
In 1966, Alves joined the National Renovating Alliance (ARENA), where he led
the minority group. Even though this party supported the military regime, his
political rights were revoked in 1969 as a consequence of internal party disputes. He
later joined the opposition party Brazilian Democratic Movement (MDB), which
eventually became the Party of the Brazilian Democratic Movement (PMDB) during
the political transition that called for direct elections at the subnational level in
1982. During this year he suffered an electoral defeat against the Maia family, which
had already come to power by direct appointment during the military regime.
The Maia family had won three consecutive governorships. In the first two,
Tarcí sio Maia (ARENA) and Lavoisier Maia (ARENA) were respectively appointed for
116
the state government in 1975 and 1979 during the military regime. The third, Jose
Agripino Maia (ARENA), was elected by popular vote in 1982. After initially served
the government of Alves and migrated to the Maia’s political machine, Geraldo Melo
(PMDB) returned again to the political group led by the Alves family and was elected
governor in 1986. After that, power alternated between these two families: Agripino
Maia (PFL) in 1990 and Garibaldi Alves Filho (PMDB) in 1994 and 1998, when the
re-election rule took effect. The back-and-forth between these families in power was
interrupted by the election of Wilma de Faria (PSB), who was elected in 2002 and re-
elected in 2006. However, with the support of the Maia family, Rosalba Ciarlini
Rosado (DEM) was elected governor, thus granting the
return of the old oligarchic forces to power (Trindade, 2010).
Although the election of Faria as governor in 2002 clearly represents an
anomaly vis-a - vis the dominance of these two political families, she was not a
political outsider. In fact, she became a public figure in 1979 as the wife of the
Governor Lavoisier Maia, who assigned her the role of leading community work
sponsored by the state government, a traditional role reserved for First Ladies in
Brazil. Later Faria served the government of her cousin by marriage, Agripino Maia.
Eventually she was elected House Representative in 1986 through the PDS, a rightist
party that evolved from ARENA. Surprisingly, during this period her voting record
was aligned with social causes defended by workers’ movements, which eventually
led her to join a left-of-the-center party, PDT (Oliveira 2005). Afterward, she was
elected mayor of Natal in 1988, and after switching to yet another left-of-the-center
party (PSB), she was elected again for the same position in 1996 and re-elected in
2000. Finally, Faria was elected governor in the second round in what appeared to
117
be a changing point in the oligarchic state politics.
Several factors should be taken into consideration when accounting for
gubernatorial elections: the composition of the electoral coalition, the presidential
coattail, position of the other political forces, and the candidates’ electoral record. In
this sense, the reduced size of the coalition seems to play a smaller role in Faria’s
election. In fact, if the coalition size were a crucial factor, the main beneficiary would
have been the other contenders – Fernando Freire (PPB) and Fernando Bezerra
(PTB) – who had stronger electoral coalitions. Instead, it appears that Faria’s
popularity as a trice-elected mayor of the capital and the largest city in the state, laid
the groundwork for her electoral success. As a strong third candidate running
against two polarized political forces, she gained enough votes to make it to the
second round. In the runoff election she benefitted from the direct support of two
important actors. First, she garnered the support of President Lula da Silva (PT),
who the PSB already had as a partner in the national election due to the absence of a
local candidate from the PT. Second, she secured the support of Agripino Maia (PFL),
who backed the candidacy of Bezerra (PTB) in the first round (Spinelli 2006).
The re-election of Faria in 2006 happened under different political
circumstances. Now an incumbent, Faria was able to form a broad coalition of
parties on the right, center, and left of the political spectrum, against the two families
that coalesced to launch the candidacy of the previous two-time governor, Garibaldi
Alves Filho (PMDB). Alves Filho also received the support of Rosalba
Ciarlini Rosado (PFL), who belonged to the third most powerful clan in the state
(Spinelli, 2006). Although Rosado had won a Senate seat, this coalition was not able
118
to elect the governor. Once again this election was competitive and decided only in
the second round. Faria (PSB) framed her campaign as a woman running against
traditional political forces and, most importantly, stressed her association with
President Lula da Silva. Alves Filho (PMDB), on the other hand, stood as a moderate
opposition candidate with vast administrative experience and added that the
unusual coalition of the two formerly-opposing forces represented a new political
era in the state (Spinelli 2006). It did not work this time around but in the 2010
election the story was different.
The election of 2010 presented a slightly different dynamic. Faria could not
run again for governor as she had already served two consecutive terms. Thus
constrained, she stepped down to run for Senator and her vice-governor, Ibere de
Souza (PSB), took office and became the running candidate. Although supported by a
small coalition mostly composed by leftist parties, different from the previous
elections, the left was divided between Souza and Carlos Eduardo Alves (PDT),
another member of the Alves family who defected from the PMDB and joined the
PDT because there was no space in the former party to accommodated the electoral
ambitions of a growing family (A. Alves, personal communication, July 29, 2013).
Rosalba Ciarlini Rosado (DEM) on the other side formed a broader coalition
that included parties from the center and right. Additionally, she received the
informal support of part of the PMDB (Spinelli 2010), which decided not to formally
join the coalition due to internal divergence (Felí cio 2010). Moreover, Rosado
attracted the discontent of otherwise opportunistic political leaders who once
belonged to the Faria government who had suffered from accusations of corruption
119
that involved the governor’s son and brother (Spinelli 2010). Considering this
scenario, Rosado was elected in straightforward fashion in the first round, thus
granting the return of the old political forces to the helm of the state’s politics.
Bahia
The state of Bahia is situated in the Northeast and has an estimated
population of about 15 million distributed among 417 municipalities. The Gini index
is .49, and the Municipal Human Development Index is .66 (IBGE). The recent
political history of the state has been characterized by the dominance of the Anto nio
Carlos Magalha es (ACM) oligarchy. ACM rose to prominence during the military
regime when he was indirectly elected as governor, with the support of the regime,
in 1971. He stepped down after his term ended in 1975. As there was no re-election
rule at the time, ACM returned to power under the same circumstances in 1979, and
then supported the victorious 1982 candidacy of Joa o Durval Carneiro (PDS) in the
first direct election for governor after the military coup in 1964. Opportunistically
allied to the forces that promoted the transition from the military regime to
democracy, he was appointed to the Ministry of Communications of Jose Sarney
(PMDB) in 1985 (Terra, Senado Federal). His only sign of weakness was that his
candidate for Bahia Governor, Josaphat Marinho (PFL), suffered an overwhelmi ng
defeat in the 1986 election.
In 1990, ACM returned to elected office by winning the governorship of Bahia.
Subsequently, he successfully ran for the Senate and also helped to elect his choice
candidates: Paulo Souto (PFL) in 1994, Ce sar Borges (PFL) in 1998, and
120
Paulo Souto (PFL) again in 2002. In a remarkable turn of events, s till as a senator in
2006, ACM saw his sitting candidate lose the gubernatorial election against Jacque
Wagner (PT). Wagner, who was a former Labor Minister under Lula da Silva’s
presidency, represented a momentous turn to the left i n state politics. Many saw this
election as the defining moment of the defeat of oligarchic forces led by ACM, who
died in 2007. In fact, many members of his group, Ce sar Borges for instance, joined
another party in support of the Lula da S ilva and Rousseff government. However, his
grandson, Anto nio Carlos Ma galha es Neto (ACM Neto), has inherited his political
fortunes and offers a realistic alternative pathway back to power for the political right
in Bahia.
An analysis of the election of Wagner (PT) in 2006 and the concomitant
retraction of the dominance of one of the strongest oligarchies in Brazil should also
take into consideration both local and national factors: the popularity of Wagner,
realignment of the political forces, presidential coattail effects, and the public’s
exhaustion with the old political oligarchy. In reference to the popularity of Wagner,
he had been elected to Congress three times in 1990, 1998, and 2002. After
unsuccessfully running for governor in 2002, he was appointed as Minister of Labor
and Ministry of Institutional Relations, which are responsible for the relations
between the Executive and Legislative; both positions gave him substantial national
visibility. During the gubernatorial election of 2006 he had not only increased his
visibility vis-a -vis 2002, but also was seen as a special interlocutor of Lula da Silva, a
popular president who had prioritized the election of Wagner, thus giving Wagner
the advantage of presidential coattails.
121
Paulo Souto (PFL), who was serving a second non-consecutive gubernatorial
term, counted on the structure of ACM’s machine, dubbed “Carlismo.” In spite of his
incumbent position, he could no longer count on the federal political machine under
ACM as Lula da Silva’s presidency stripped him off of the seemingly monopolistic
control of federal appointments in the state as well as central positions within the
federal government (Dantas 2007). The relative weakness of ACM as a consequence
of his lack of direct ties with the federal government (Montero, 2010) allowed
Wagner the opportunity to aggregate important opposition forces under the same
coalition, which listed parties mostly from the left, some parties from the center, and
parties from the right of the political spectrum. Especially important for this
coalition was the support of the PMDB, a party with great reach in the state and
whose leader Geddel Vieira Lima was also a Minister in the Lula da Silva’s
administration. Other factors also contributed to consolidate Wagner’s election in
the first round: one refers to the neutrality of the PSDB which, in spite of being a
preferred PFL ally in national politics, was a historical foe of ACM’s machine. The
other factor is related to the informal support of the mayor of Salvador, Joa o
Henrique Carneiro (PDT) to Wagner (Dantas 2007). Paulo Souto (PFL), on the other
hand, formed a homogenous rightist coalition that, like in the previous election, was
relatively small but that, differently from the last previous election, did not have the
opposition (PT, PMDB, PSDB, and PSB) fragmented into different candidates.
The election of 2010 took place under a slightly different scenario, as is
usually the case. Although the outgoing president Lula da Silva was very popular and
his support was again crucial, the most important re-alignme nt happened in the
122
configuration of the electoral coalitions. While the forces that opposed the ACM
machine in the previous elections were mostly united, this time around they were
fragmented and three important candidates presented a real chance of winning
based solely on their previous electoral strength. Paulo Souto (DEM) tried to regain
strength after both being defeated in the previous election and the passing of ACM in
2007. He counted on the unprecedented support from the PSDB, which happened by
imposition of the PSDB’s central committee given its presidential ambitions. Geddel
Vieira Lima (PMDB) withdrew his support for Wagner and ran for governor on an
extensive coalition formed mostly by small rightist parties.
However, this new combination of factors did not upset the re-election plans of
Wagner who kept the PSB in his coalition and then officially added the PDT. U nder
this political configuration, Wagner achieved outright victory in the first round.
I turn now to political linkage variables that focus on the parties’ strategic
behavior to connect to voters (ideology and clientelism) and strategic organization
to connect to other parties (electoral coalition formation). Interviews with political
elites reveal that both strategies are at work.
Ideology
The conventional view contends that parties in Brazil lack institutional
strength and are undisciplined and non-programmatic, which contributes to
undermining the ideological linkage and, consequently, advancing non-policy based
strategies, such as clientelism and personalism (Ames 2001, Mainwaring 1999,
Roberts 2002, 2010). Other scholars contend that leftist votes are ideologically
driven in Brazil (Holzhacker and Balbachevsky 2007).
123
Recent electoral success of more programmatic parties in Brazil has reignited
the debate regarding the consistency of the party system with a burgeoning
literature challenging conventional wisdom by empirically demonstrating the
strengthening of these party systems. In tune with this view, Chapter 3
demonstrates, both at the aggregate and individual level, tha t th e Brazilian party
system indeed shows some degree of party differentiation. Remarkably however, the
overwhelming majority of party operatives interviewed for this project reported
that there is lack of ideological linkage between party and voters.
Among the interview respondents on the political right, there seems to be a
consensus about this lack of ideological linkage. During field research to Brazil in
2009 for instance, a deputy from the Democratas (DEM) stated in an interview
that:
Today the voter very often votes for the candidate, he does not vote for the
party, very often he does not know what the party’s platform is, what the
programmatic line is. Perhaps he remembers some of the party’s point
and he identifies himself with this. But he does not vote for the party, he
votes for the candidate. I believe that when the deputy Felipe Maia goes
on the street to ask for votes, the voter is not voting for the Democratas,
he is voting for Felipe Maia (Felipe Maia, 2009).
When asked about the political linkage between parties and votes in a more
recent interview, another politician wit hin the same party quickly replied that there
was no such linkage and that the voter choses the candidate, not the party. He later
complained that it had become more difficult for him to get re-elected and credited
such challenges to the lack of militancy within his party. He also added that the great
124
advantage of the leftist parties was the ability to count on an active militancy (P. Azi,
personal communication, July 23, 2013).
The lack of party affinity can even be reproduced at a higher political level as
far as the relationship between mayors and deputies is concerned. In Brazil, national
parties’ strategies are generally conceived during the mayoral election, which takes
place two years before the general elections. Although there is no indication that the
mayoral elections influence presidential elections (Nicolau 2008) or that this
influence can even be negative (Zucco 2008), the state and house representatives
link their electoral prospects to their ability to recruit mayors as cabos eleitorais, or
supporters (J. Dias, personal communication, July 31, 2013). As the PSDB deputy J.
Ju nior put it, the election of congressmen was directly linked to the elections of
mayors (personal communication, August 21, 2013). He explained that elected
mayors endorsed the candidacy of congressmen [two years later] and their support
was definitive for their election, much more so than for the president. Another
deputy from the PMDB adds that the voters would follow their mayors when as ked
to vote for a Congressman but they would freely choose their president and
governor independent of the mayor’s request. He added later that nowadays it
would be impossible to win an election without the endorsement of the mayors (L.
Vieira, personal communication, August 27, 2013).
However, while J. Ju nior linked this support to mayors within the same party
(personal communication, August 21, 2013), two mayors from different states
posited that they were free to choose whom they would support. When asked about
how the process of choosing his Legislative Deputy was established, a PDT mayor
125
said that it was not ideological because the number of parties makes it difficult to
differentiate among the various parties. He held that the deputies had specific
interests in certain areas and they approached the mayors but it was a person-
toperson contact as opposed to party affinity (A. Rocha, personal communication,
July 24, 2013).
Another PMDB mayor indirectly corroborated this view. When asked if party
leaders dictated what candidates they should support, he answered that mayors
were free to choose whichever candidates for deputy they wanted to endorse. When
further asked about the name of the candidates he endorsed, he replied that for the
State Assembly he endorsed Gesane Marinho. Then, when finally asked about which
party she belonged to, he answered that he did not know and further suggested that
her party was PMN but pondered that what really mattered was that she was from
his region. Finally, he disclosed that for the House of Representatives he endorsed
Henrique Alves because this was a matter of party affiliation [they belong to the same
party, PMDB] (F. Sousa, personal communication, July 31, 2013).
On the left, conversely, there seems to be a general affirmation of the
important role ideology plays in politics. This discourse is articulated mostly around
the party’s program and policies, as opposed to a direct political linkage with the
voter. In fact, the explanation revolves around the realignment of one of the main
parties on the left, the Workers’ Party. Since 2003, the PT has reformulated its
program and broadened its electoral coalitions to include parties more at the center
and center-right of the political spectrum (Hunter 2007, 2010; Samuels, 2004).
126
Some deputies stated that the Workers’ Party became more like the PMDB or
“peemedebizou-se” (A. Alves, personal communication, July 29, 2013) as a
consequence of this transition.
Politicians at the center-ri ght aligned with the Brazilian Democratic
Movement Party (PMDB) affirmed the general tendency of strategic political
alignments with the Workers’ Party. The PMDB has participated in all but one
governmental coalition since redemocratization (Freitas 2012). For this reason the
PMDB has gained a reputation for being a quid-pro-quo “fisiológico” party. As one
party insider described it, the PMDB abandoned the prospect of running candidates
for national elections to focus on more a quid-pro-quo kind of politics that comes
with governing; for instance, the exchange of legislative amendments for
congressional funding (L. Lima, personal communication, August 27, 2013).
Similarly, another state representative from PMDB held that his party was very much
criticized for being a fisiológico party and reckoned that to some extent it was indeed
a fact (Morais, personal communication, August 5, 2013). Finally, a former party
member justified his existence by stating that he was elected through PMDB for
seven consecutive mandates but that his relationship with his party had worn out.
The party had become very fisiológico and, as a consequence, he could no longer
agree with the direction the party was taking (J. Dias, personal communication, July
31, 2013).
Workers’ Party members, however, defend the party’s program and policies
and separate them from the party’s strategy to win elections. For instance, J. Neto
suggested that the broad governmental coalition caused confusion since the other
127
parties [in the coalition] did not have a national project [for the country]. Asked if
this “confusion” prevented the PT from implementing its program, he answered
negatively, adding that the reason was because the PT had a line and the other
parties knew this and were expected to follow it. He then enumerated the specific
policies as foreign policy, social policies, and transparency and democratization of
the Brazilian society (J. Neto, personal communication, July 25, 2013). This opinion
was shared by another PT member who emphasized the party’s national project and
argued that other parties in the coalition would eventually come to support this
project (Luiza Maia, personal communication, July 25, 2013).
A rank-and-file party member recognized the existence of a party program
but stressed the difficulty of advancing such a project in view of the heterogeneity in
the coalition. In his view, the party’s coalition occupied the center-left and tried to
promote income redistribution as the main axis of the PT’s goals. However, his party
could not move forward with other progressive issues such as the reduction of the
working hours (M. Braga, personal communication, August 8, 2013). Another
congressman from the same party had an identical perspective about the constraints
for the party. As a consequence of the heterogeneous governmental coalition, they
were unable to approve some amendments such as fiscal reform and the reduction
of working hours to 44 hours per week (G. Simo es, personal communication, August
27, 2013).
Finally, yet another PT party member offered his diagnosis for the issue of
ideology, holding that the lack of ideological constraints was a huge problem in
Brazil because there are many parties without a coherent line of thought. Even
128
within the PT, which was a more ideological party, t here were representatives that
occupied a wide spectrum of positions that ranged from the left to the right. The
interviewee suggested that other parties such as PMDB did not have any ideological
profile whatsoever. He contended that besides the PT, perhaps the only parties that
had some ideological coherence were the PSol and the PC do B. (C. Vacarezza,
personal communication, August 13, 2013).
Clientelism
The understanding that clientelism besets Brazilian politics, in general,
(Avelino Filho 1994) and rightist parties, in particular (Montero 2010), is challenged
by scholars who claim that clientelism has declined due to systemic (Borges 2011,
Hagopian et. al. 2009, Hunter and Power 2007) or institutional (Borges 2011)
reasons. Sugiyama and Hunter (2013), for instance, sustains that Brazil has created
massive poverty alleviation programs whose implementation spurs modernizati on
while defying established practices of the traditional political system, such as
clientelism and patronage. It has also achieved economic stability, which has paved
the way for the expansion of the middle class and, consequently, for the demands on
the political system. Although still a very unequal society and with unequal regional
development, even by developing countries’ standards, recent political outcomes
suggest substantial changes in patterns of political competition, especially in the less
developed areas of the country. I assessed these claim qualitatively during my field
interviews. Clientelism is still a phenomenon, but the nuanced perception that it has
been losing strength in Brazilian politics appears to vary based on one’s ideological
129
orientation and party membership. Party operatives on the left tend to have a more
positive view on progress in decreasing clientelism, whereas views of those at the
center and right are less so.
The majority of those interviewed stress ed that most requests received from
their constituents are related to personal benefit. A PDT mayor, for instance,
described his interaction with voters as a constant pressure for favors. He contended
that the overwhelming majority of the voters in his municipality were interested in
personal benefits with only a minority interested in his program (A. Rocha, personal
communication, July 24, 2013). A PDT state deputy added that only voters who had
strategic importance received special attention, since they knew all voters wanted
was individual benefits (A. Alves, personal communication, July 29,
2013).
This idea was endorsed by another representative from the PR who
emphatically responded that the most frequent request from his constituents was
for employment. When further asked how he dealt with such requests, he replied
that it was very difficult to place people in jobs given his limited resources. He
conceded that he tried to clarify this to the voter but suggested that powerful voters,
given their status as local leaders, required special attention. Although he claim ed to
treat all voters equally, he acknowledged that each municipality had its own weight
(G. Soares, personal communication, July 30, 2013). Thus, even though he was
cautious about admitting the straightforward use of clientelistic strategies -
patronage is illegal after all - he suggested that depending on the weight of his
supporter the request for employment would be considered.
130
Later in the interview I asked if voters offered to sell their votes and his
answer was a resounding yes. He recognized that those who had money won
elections (G. Soares, personal communication, July 30, 2013). Another representative
from the PSD was asked about what was necessary for him to receive the support of
local elites and he answered that realistically there was a system of mutual support,
but this also included friendship and favors that he provided along his political
career that people remembered on election day (J. Dias, personal
communication, July 31, 2013).
Institutional mechanisms supporting clientelism were also identified in
interviews. The use of patronage, another mechanism through which clientelism
operates, is institutionally visible in such case as the Rio Grande do Norte’s State
Assembly. A. Alves points out that 99.9% of the public servants in the state
legislature’s Lower House have their jobs as a result of political influence (personal
communication, July 29, 2013). He added that only in 2013 would the State
Assembly hold its first meritocratic public entrance exam to fill staff positions. At
this point, his chief of staff interrupted the interview to add that the legislation has
required such entrance exams since the 1988 Constitution but that this law is only
now taking effect in this state (Cunha, personal communication, July 29,2013).
During my visits to the State Assembly, one particular office was always busy
with waiting lines that exceeded 20 people on any given day. I learned that this
representative’s legal staff offered constituents legal consultations free of charge.
Later, in an interview with an assistant in this office, I asked what the main demands
of the constituents were. He explained that since the representative ran his
131
campaign calling himself “the people’s lawyer,” most of the requests were for
assistance associated with retirement benefits and the resolution of legal claims.
When prompted about whether there were employment requests he emphatically
replied yes. He added that he received many of those requests but that it was easy to
say no because he had a good argument when he said that the deputy was the
people’s lawyer and he was willing to advocate for the voters’ legal issues but that
employment is not a legal issue (T. Moura, personal communication, July 29, 2013).
Another institutional source for clientelistic linkages was described by a high
ranking politician who stated his support for single member plurality district
(SMDP) as opposed to the actual proportional representation system (PR).
According to him, the SMDP system represented a vote with local identity but that
the politicians were concerned that their lack of identity with specific regions would
prevent them from amassing votes by acquiring the endorsement of local leaders . He
added that such mechanisms were vitiated since they were based on private
interests and on the buying of such endorsements (R. Faria, personal
communication, August 1, 2013).
Importantly, representatives perceive a notable change on the part of the
electorate, which they attribute to modernization. A mayor recognized that the voter
was changing and becoming more demanding. He assented that from then on,
candidates would no longer win an election by just doing construction works. He
argued that with the growth of the socio-economic class “C,” one’s politicians were
obliged to create new jobs because everything else had become just a part of their
obligations. He finished by saying that the tendency of the human being was to
132
always want more (F. Sousa, personal communication, July 31, 2013). When
prompted to specify the time when he noticed such changes, he answered that he
had noticed this difference in his second term in office but that there was still those
who only seek personal benefit, concludi ng by saying he could not lie about it (F.
Sousa, personal communication, July 31, 2013).
Similarly, the vice-governor of Rio Grande do Norte agreed that vote buying
indeed exists but added that collective issues [as opposed to personal interest] had
been advancing and concluded that in each election the voter was increasingly free
and demanding (R. Faria, personal communication, August 1, 2013).
The difference between left and right does not come from their
acknowledgement of clientelistic practices, but how their mandate is conditioned by
the nature of their main voters. Poli ticians on the right cautiously admit that the
control of the state apparatus (A. Alves, personal communication, July 29, 2013), the
buying of mayors’ endorsements (F. Souza, personal communication, July 31, 2013),
and straight vote buying (G. Soares, personal communication, July 30, 2013) are
generally decisive clientelistic resources in elections. Conversely, politicians on the
left generally invoke their ties with social movements (J. Guimara es, personal
communication, August 21, 2013; J. Neto, personal communi cation, July 25, 2013) as
an ideological dimension linking them to the voter.
Coalitions
The study of electoral coalitions has stressed the role of formal institutions
conditioning the policy-seeking behavior of political parties. In this sense, a party’s
ideology is a crucial element in explaining and predicting electoral coalition
133
formation. The office-seeking behavior of the political parties, however, requires
consideration of informal institutions, of which clientelistic linkages are essential.
The introduction of clientelism as a second dimension operating in the political
system helps to explain oversized ideologically heterogeneous electoral coalitions
and, consequently, party development and electoral outcomes in multiparty
presidential regimes in developing democracies. This logic is articulated by the
combined utility theory, which proposes that polarized competitive elections in
modernizing national electoral markets force programmatic parties to coalesce with
clientelistic parties to gain access to regional private electoral markets.
Socioeconomic changes that bring about modernization create a public electoral
market, which demands programmatic-oriented representation determining
patterns of political competition on the ideological dimension. However, uneven
socioeconomic changes perpetuate private electoral markets, which demands
patronageoriented representations determining patterns of political competition on
a clientelistic dimension.
This chapter considers the office-seeking approach as the foremost important
element guiding electoral formation given the electoral polarization of
programmatic parties. The policy-seeking goal of parties remains an important
strategy between elections. The focus on office-seeking strategies relaxes the
assumption that parties place a premium on ideology as the main determinant
driving coalitions. That is, policy-seeking parties will place a higher value on
ideologically contiguous coalitions. However, office-seeking parties may relax the
134
requirement of ideologically consistent coalitions, provided they can enter those
coalitions without being severely punished by an ideologically demanding
electorate.
Political competition in multiparty democratic developing countries occurs in
a very heterogeneous cross-national electoral market. Thus, the national success of a
party depends on its ability to prospect votes in public goods and in private goods
markets. However, a programmatic party cannot overtly pursue clientelistic politics
because it would risk losing public market voters, hence defeating its purpose of
building a party brand as the most effici ent means of electoral linkage in the public
goods market. By the same token, an expanding programmatic party
cannot alienate clientelistic votes, especially if it has exhausted its universe of
expansion in the public goods markets and needs to compete with other
programmatic parties.
In expanding competitive the public goods marketplace as a consequence of
structural changes in multiparty systems, the success of programmatic parties
depends on their ability to extract additional votes from private goods markets at
the subnational level. Theory put forward in this dissertation is that pragmatic
parties can do this without compromising the party brand by coalescing with
political machines already established in those private goods markets. Does this
happen in practice? What do interview subjects from the northeast of Brazil reveal?
Elections for all positions, except for mayors and aldermen, happen
concomitantly in Brazil. Generally, the strategies driving electoral coalition
formation, with the ultimate goal to win elections, are mainly driven by two
135
motivations: access to free broadcasting time for the majoritarian election and
overcoming the electoral coefficient in proportional elections (J. Negra o, personal
communication, July 23, 2013). The free broadcast time is the main media resource
to reach the voters. Since this time is allocated to each party independent of the
existence of a candidate, the consequence is that the party owns its time and, in this
case, if said party enters an electoral coalition the party’s time will be added to the
coalition. In this sense, the size of the party will determine the bargaining
power when entering the coalition.
For proportional elections, the free broadcasting time is less important
because the higher number of candidates reduces each individual’s appearance to
meager seconds. Hence, the electoral coalitions for proportional elections are d riven
by the necessity to beat the electoral coefficient. In this sense, small and even
medium size parties that do not have enough votes to reach the electoral coefficient
from electoral coalitions, for the vote-counting effect, are considered as one single
party.
Parties generally try to form competitive tickets that include a presidential, a
gubernatorial, a senatorial, and House candidates, but the coordination of these
coalitions varies according to each election, the particularity of each state, and the
introduction of new legislation.66 In general, the states have autonomy to form their
coalitions; however, states’ idiosyncrasies may require a top-down approach by the
party’s central committee (R. Faria, personal communication, August 1, 2014). As
one rank-and-file member explained, the coordination of the electoral coalition is
desirable programmatically, but it does not always happen because of coronelism and
136
the interests of local political families (M. Braga, personal communication, August 8,
2013).
In the 2010 election, for instance, the P T’s central committee intervened in
the state of Maranha o, forcing the local party to support the candidacy of Roseana
Sarney (PFL), whose father was Jose Sarney (PMDB). Securing an alliance with the
Sarney family was crucial for the PT as the elder was the Senate chair, an important
ally of the PT in the federal government, since he could ensure PMDB loyalty in
Congress to support the PT’s legislative agenda. Other times, local interests
supersede national alliances as in the case of Rio Grande do Norte in 2010. During
this election the PMDB preferred not to form a coalition with the PT and instead
freed its members to endorse whatever party it wanted (H. Morais, personal
communication, August 5, 2013). In fact, many parties that do have national
ambitions, i.e. strong presidential candidates, may focus on strategies that give them
the best chance of electing the highest number of congressmen. L. Lima (PMDB)
explains this approach saying that since the PMDB did not have a national project [to
elect the president], it prioritized states’ elections because the strength of the PMDB
came from their senators and congressmen (personal communication, August 27,
2013).
Parties may establish general guidance on preferred parties with which to
form electoral coalitions but very few parties follow it (D. Almeida, personal
communication, August 8, 2013). The PC do B’s guidance, for instance, was not to
coalesce with the PSDB and the DEM, but specific cases allowing such coalitions
could still be considered (A. Bueno, pers onal communication, August 8, 2013). The
137
leftist party, the PSol, was identified by other partisans as the only party that
systematically restricted their coalitions based on ideological principles (C.
Vacarezza, personal communication, August 13, 2013). This information was
confirmed by a rank-and-file member of the PSol who stated that his party
prioritized ideological coherence and that the general orientation of the party was to
only form coalitions with other left-learning parties, such as the PSTU and the PCB
(J. Guimara es, personal communication, August 21, 2013).
The national polarization between the PT and the PSDB loosely defines the
pattern of coalition in the states. C. Vacarezza (PT), for instance, claimed that, as far
as coalition formation applied, the replication of coalitions i n the states was perhaps
the only coherent behavior that stemmed from the national polarization between
the PSDB and the PT. In this case, he added, the PSDB represented a more neoliberal
ideological position (personal communication, August 13, 2013). Similarly, A.
Imbassahy (PSDB) acknowledged that the PT and the PSDB were programmatic
parties that rival each other at the national level and, for that reason, both parties
reciprocally rejected coalitions with one another. He added that to a certain extent,
the PSDB tends to follow the national line on forming subnational coalitions
(personal communication, August 13, 2013).
Although the main factor driving coalitions in proportional elections is the
electoral coefficient hurdle that parties need to overcome, individual candidates in
medium and small parties may devise strategies to maximize their own chance of
being elected once that coefficient is achieved. Big parties achieve their size because
138
they have vote-champion candidates who bring sufficient votes not only to elect
themselves, but also garner the votes to boo st other party members through the rule
of proportionality. Leaders of smaller parties thus may choose to form coalitions
with other smaller parties because they fear that by coalescing with bigger parties
their votes would not be enough to beat the other candidates within the same
coalition (J. Negra o, personal communication, July 23, 2013).
A state deputy in Rio Grande do Norte confirmed the importance of party
size, saying that small parties do not like to for m electoral coalitions with bigger
parties as the former tend to not elect anybody due to the proportional
representation (G. Soares, p ersonal communication, July 30, 2013). The calculus on
the potential to be elected is based on previous electoral performance and thus
candidates prefer to form coalitions with other parties that will deliver enough votes
to reach the electoral quotient without upsetting the chances of the strongest
candidate within the coalition. The candidates with less or no potential to be elected
are called “esteiras” (conveyor belts), as their role is restricted to contribute votes to
the pile. When asked if the esteiras know that they are only instrumental candidates
to the coalition, one informant answered that the majority of candidates did not
know they were esteiras due to the lack of understanding about the electoral game.
He further explained that, as newcomers in politics, the candidates are convinced of
their victory even as it is all but certain they will not get elected (T. Moura, personal
communication, July 29, 2013).
Finally, another important element driving coalitions is the phenomenon of
capilaridade, defined as the ability of a party to prospect for votes in remote areas,
139
or, in other words, the degree of reach that a party has in remote areas. J. Neto
(personal communication, July 25, 2013) referred to it as the strength of some
parties in specific regions, since some parties do not have sufficient density at the
state level but in do in specific areas. M. Braga (personal communication, August 8,
2013) stated that most parties only had regional expression, while only three parties
were nationally expressive (PT, PSDB, and PMDB).
With national polarized competitive elections among programmatic parties,
the differential vote necessary to win elections has to be garnered in a
nonideological dimension. Since more clientelistic parties tend to have greater reach
in more rural areas, they become desirable partners in coalitions for it. A
congressman referred to the necessity to form broad coalitions, saying that it was
natural for the leftist parties to have access to an electorate that was more
ideological but acknowledged that they did not have access to all constituents, such
as religious and regional groups. In this sense, he added that it was necessary to
form coalitions so as to increase the potential to gain this electorate (D. Almeida,
personal communication, August 8, 2013).
To justify the importance of his party as a strategic national partner of the PT,
H. Morais explained that the PMDB was the biggest Brazilian party and the one that
had the biggest reach throughout the municipalities (personal communication,
August 5, 2013). He added that it would be a big mistake for its presidential
ambitions if the PT failed to bring the PMDB into its coalition. In fact, some
representatives pointed out the phenomenon of capilaridade driving electoral
coalitions being specially important in the majoritarian elections (G. Simo es,
140
personal communication, August 27, 2013). L. Lima goes further to explain that it
was more important for presidential elections (personal communication, August 27,
2013). For governor, he added, there was a problem related to the widespread
nature of parabolic antennas, which allowed the voter to get TV signals straight from
Sa o Paulo, thus missing the locally-produced content. According to him, it was
important to build a local structure that would bring the message directly to the
voters. It was important to be present to deliver the message, he concluded.
Conclusion
Brazil’s political structures have shown some signs of change with the
election of programmatic parties for the presidency and for the governorship,
particularly in the Northeast after 2002. This chapter focused on two states as case
studies, highlighting these changes and taking into consideration the operation of
ideology and clientelism and the interaction of these two dimensions in defining
electoral coalition formation.
In the state of Rio Grande do Norte, we saw that the rise of PSB to power
against two oligarchic families leading the PMDB and PFL/DEM parties. The
polarization between two historical political forces opened up space for a third
candidate who had shown increasing electoral strength as a popular mayor of the
most important city. The presidential coattail effect then appeared to have defined
the election of Wilma de Faria (PSB) for her first mandate. The re-election, though,
happened by way of the re-alignment of both the left under Faria and the right ,
including the once opposing oligarchies, under Alves Filho. However, this alliance
was not enough to overcome the effects of the presidential coattails with strong
141
support from Lula da Silva for Faria.
In the state of Bahia, several factors explain the rise of the left under the
Workers’ Party. First, you see the oligarchy’s exhaustion and the defection of ACM
after 16 uninterrupted years in power. The PT’s candidate, Wagner, also benefited
from the presidential coattail effect as he did from the re-alignment of the
opposition.
Looking specifically at the variables ideology and clientelism, which together
account for electoral coalition formation, the qualitative interviews show that, i n
general, ideology has lost importance as a factor driving electoral coalitions. On e of
the main factors indicated by party officials during interviews was that the
pragmatic move of the PT towards the center of the political spectrum had a notable
effect. This strategic move broadened the scope of the electoral coalitions made by
the PT thus making ideological criteria less important than the electoral calculus in
terms of winning office. A state representative from the PT in Bahia claimed that as
her party started to become stronger it realized that it would not be able to win
majoritarian elections alone. She added, however, that the PT was initially more
prone to form electoral coalitions with parties with which they had a history of
alliances (L. Maia, personal communication, July 25, 2013).
Indeed, this strategy was confirmed by a member of an historically allied
party, who stated that it has always been discussed within the PC do B that in order
to win elections it was necessary to broaden electoral coalitions (D. Almeida,
personal communication, July 25, 2013). However, the party personnel also
suggested that, although less ideologically consistent and given the autonomy of the
142
states, there was a general guideline for electoral formations that loosely followed
the national polarization of the PT and the PSDB, with each party orienting their
counterparts in the states to prioritize electoral coalition that excluded the main
opponent in national elections.
As a consequence of the strategic move of the parties to relax the ideological
dimension as the main determinant of electoral coalition formation, the electoral
calculus prevailed and parties sought electoral coalitions that would deliver the
highest electoral prospects. In this sense, clientelism, a second dimension of
electoral linkage, became important because strong programmatic parties tried to
garner additional votes by forming electoral coalitions with parties that had access
to more remote areas of the country. This effect was described as capilaridade in
reference to the degree of reach of certain parties in more rural areas of the country.
CHAPTER 7 Conclusion
This dissertation has argued that political competition in multiparty
developing democracies takes place in two dimensions, those of ideology and
clientelism, and that parties resort to electoral coalitions to extract votes in both
domains in order to create political linkages with voters. The mechanisms under
which political competition takes place still generate heated debates about the
evolution of the party system in relatively new democracies. The political party
system and the extent of its institutionalization are at the center of this contention.
Some scholars suggest that nascent democracies lack ideological depth and
thus rely on clientelistic strategies to link voters to the electoral process (e.g.
Kitschelt and Wilkinson 2007, Nyblade and Reed 2008, Piattoni 2001). In contrast,
143
others point out that empirical evidence from contested elections suggests that
parties have advanced as a consequence of systemic changes (e.g. Borges 2011,
Hagopian et. al. 2009, Hunter and Power 2007). In this debate, ideology and
clientelism tend to be confounded in one singular political dimension. Particularly in
the Brazilian case, there has been a propensity to assert that political parties on the
right are clientelistic and that parties on the left are programmatic. This study
breaks from this tendency in important ways by suggesting that ideology and
clientelism can operate distinctively. The proposed combined utility theory
developed in this dissertation explains how both ideological and
clientelistic/programmatic strategies interact during the political process.
The general combined utility theory indicates that competitive polarized
elections in developing democracies constrain parties to seek crucial votes on a
secondary clientelistic dimension when they have exhausted its potential to attract
voters based on ideological appeal. It also posits that programmatic parties, which
are prevalent in public electoral markets, compete in national polarized elections by
forming electoral coalitions with clientelistic parties to preserve their own party
brand while garnering votes in private electoral markets, where clientelistic parties
better perform. The confirmation of this theory is particularly relevant as most
elections around the world take place in modernizing settings where parties
compete in environments that combine both public and private markets.
To test the combined utility theory, this dissertation uses Brazil as a crucial
test case. The choice of this country is justified by the uneven socio-economic
development across the country alongside the existence of both programmatic and
144
clientelistic parties competing in subnational elections as well as its having had
polarized national elections for six consecutive elections. Two parties have
dominated national contests for the presidency with the center-rightist PSDB and
the center-leftist PT capturing the greatest vote share. The center-rightist PSDB
controlled the execu tive from 1995-2002 and the Workers’ Party has won the last
three electoral cycles (2003-present).
This polarization at the national level has helped to shape politics at the
subnational level, whereby shifts in electoral competition have undermined the
oligarchic forces across the country. While the oligarchy has always made inroads
and alliances with the federal government, their stronghold has always been in local
politics, particularly in the less developed Northeast region. The recent decline of
regional political bosses and the growth of programmatic parties in the Northeast
has been particularly notable.
The theoretically-derived hypotheses to assess ideology, clientelism, and their
interaction through coalitions are then described further in this dissertation. In
order to compete in public markets, political parties need to be sufficiently
differentiated to respond the ideologically-oriented demands of constituents.
Indeed, the findings presented in Chapter 3 indicate that the Brazilian political party
system is ideologically differentiated. Importantly, this study confirms that the
Brazilian party system is differentiated both at the aggregate electorate level as well
as the individual voter level, as indicated respectively by different statistical
modeling. At the aggregate level, the unidimensional deterministic model of spatial
electoral competition for gubernatorial runoff elections confirms this claim.
145
This method allows for probabilistic inference on the runoff election outcome
by assuming ideological transference of votes between the first round and the runoff
election. That is to say, the model assumes that voters retain their ideological
(left/center/right) preference between the electoral rounds. The model results
suggest tha,t in runoffs, the electorate choses the most proximate party to the one
she selected in the first round of the election. At the individual level, this study
further confirms the ideological differentiation of the Brazilian party system
by using a national survey dataset.
The analysis on ideological partisan linkages makes an important
contribution by adding nuanced analysis about how these partisan linkages take
place at the state level, where personalism is more likely to prevail. In this sense,
subnational elections provide a high-bar for testing whether parties have developed
meaningful ideological linkages with voters. The evidence strongly supports the
ideological differentiation of the political parties in Brazil. Most remarkably, even
competition between parties within is the same ideological spectrum shows partisan
differences across Brazil where subnational political competition is often deemed
prone to patron-client relationships.
Uneven socio-economic development – where pockets of deep poverty and
unemployment have created few opportunities for upward mobility – can
nevertheless counter the electorate’s ability to vote freely. Clientelism and
patronage, which are longstanding traditional practice in Brazil, suggest the
existence of a private electoral marketplace. To test the hypothesis that a private
electoral marketplace fuels clientelistic politics independent of ideology, this
146
dissertation developed a two-pronged strategy.
First, Chapter 4 develops an entirely new measure of clientelism. This
measure is based on the assumption that programmatic parties tend to have votes
clustered in adjacent municipalities while clientelistic parties tend to have more
dispersed votes (Ames 2001). On the one hand, clustering refers to representatives’
strategy to garner votes in contiguous municipalities. This phenomenon maximizes
limited campaign resources, as it does not necessarily require the mediation of local
bosses to access the electorate. Consequently, clustering makes the candidate more
accountable to the electorate. Dispersion, on the other hand, is associated with the
phenomenon of dominance, which is related to the ability of representatives to
amass a high percentage of vote shares in municipalities that contribute to the
majority of his individual vote. Dominance is more common in rural areas where the
mediation of the local boss is crucial for the deliverance of such a high vote share. As
a consequence, dominance renders the candidate less accountable to the electorate.
To capture these phenomena, I draw on electoral and geospatial data to
measure clustering and dominance among parties. I initially used the Herfindahl
Index, a measure of concentration that is typically used to assess companies’
monopolist behavior, and developed the Cluster Index, which improves on the
previous measure to account for the clustering of votes according geographically
contiguous municipalities. Both variables are the result of millions of computational
iterations that run for months at the time. Although the distribution of the parties on
a two dimensional political competition using the measures discussed suggests that
in Brazil rightist parties tend to be more clientelistic and leftist parties tend to be
147
more programmatic, the model lacked statistical evidence to confirm this tendency.
One important finding relates to the effect of the interaction term of party
competition in private and public markets on the outcomes of clientelistic party
behavior. In larger private markets (i.e. larger private sector as a share of the local
economy), the residual pool of clientelistic votes is smaller, thus requiring more
public resources for the dispensation of patronage. In smaller private markets (e.g.
smaller private sectors as a share of the local economy), the supply of clientelistic
voters is higher, thus reducing the amount of public resources necessary for
patronage. This finding is also supported by qualitative interviews. The main
conclusion is that more developed municipalities tend to have a smaller pool of
clientelistic votes which, in turn, requires more public resources to feed the
clientelistic machine. Less developed municipalities can make do with fewer public
resources as they do not have to compete with other economic sectors to leverage
votes.
To test the overall utility theory, this dissertation draws on the unique
measure of clientelism (Cluster Index) to analyze whether electoral coalition
formation is a function of parties’ ideological preferences and clientelistic strategies
to win elections. Thus, Chapter 5 represents the second prong of the strategy to test
for the role of clientelism and ideology in electoral politics. Specifically, this chapter
tests the hypothesis that electoral coalitions will form with a heterogeneous
composition of programmatic and clientelistic parties. The model results indicate
that electoral coalitions on the ideological dimension are more prevalent. However,
ideology has become less important since the 2002 election. This empirical finding
148
corresponds with reports from political elites found during field research. The
common view among them was that ideolog y has lost importance in defining
electoral coalitions since 2002 when Lula won the presidential election.
The models on the role of clientelism in electoral coalition formation,
however, are not as conclusive. The analysis suggests that polarizing parties on the
left resort to less clientelistic partners than do parties on the right. Focusing
specifically on the two nationally polarizing parties, it appears that the PSDB counts
on a wider range of clientelistic parties in its coalition than the PT does. The
dispersion of the ideology and clientelism variables on coalition formation, indicated
by the standard deviation for both parties (PT and PSDB), also shows interesting
results. On the ideology dimension, both parties seem to narrow the ideological
proximity of the parties in their respective electoral coalitions. However, this
dispersion is wider for the PT than it is for the PSDB, which indicates that the PT has
been more aggressive in its strategy to form coalitions by attracting more parties
from the extremes of the political spectrum when compared to the PSDB. On the
clientelistic dimension, the result is the opposite. In fact, the PT appears to increase
the spectrum of the parties comprising its coalition over time, but it remains slightly
lower than that of the PSDB, which remains relatively constant over time. In sum, the
PT has a wider ideological configuration of parties in its electoral coalition while the
PSDB has a wider clientelistic configuration of parties comprising its electoral
coalition.
Brazil’s political structures have shown some signs of change with the
election of programmatic parties for the presidency and for the states’ governorship,
149
particularly in the Northeast after 2002. In order to flesh out the quantitative
findings, qualitative interviews were conducted in the states of Rio Grande do Norte
and Bahia from July to August 2013. Rio Grande do Norte is a notable case because it
is where the PSB first defeated the Alves and Maia oligarchies, which respectively led
the PMDB and the DEM parties. In the state of Bahia, the PT defeated 16
uninterrupted years of rule by the PFL/DEMs, headed by the ACM oligarchy.
Looking specifically at the variables ideology and clientelism, which together
accounted for electoral coalition formation, the qualitative interviews reveal that
ideology has lost importance as a factor driving electoral coalitions. Party personnel
indicated that the diminishing importance of ideolog y was due to the pragmatic
move of the PT towards the center of the political spectrum. This strategic move
broadened the scope of electoral coalitions made by the PT, thus making ideological
criteria less important than the electoral calculus to win office. However, party
personnel also suggested that their central leadership recommended that the local
chapters form electoral coalitions that followed the national polarization of the PT
and the PSDB. That is, the PT shoul d avoid forming electoral coalitions with the
PSDB, and vice-versa.
Other findings from Chapter 4 suggest that modernization operates to
attenuate the effects of clientelism on political competition. In the Cluster models,
the size of population, which is a demographic variable operating in the same log ic
of modernization theory, was also consistently confirmed. Nevertheless, both GDP
per capita and urbanization were not in the predicted direction.
150
In the following sections I will address the limitations and significance of this
dissertation and conclude with suggestions about future research.
Limitations
Informal institutions are complex structures to understand because they lack
the evident set of rules that frame and inform their mechanisms. As an informal
institution, clientelism has proved an elusive candidate for quantitative studies.
While our models suggest reliable measurements, more research is necessary to
improve the validity of the parties’ clientelistic behavior. Specifically, studies would
benefit from a country specialist to assess party placement at the state level on a
second dimension of political competition in order to advance the understanding of
political linkages beyond ideology. Ideolog y and clientelism are difficult variables to
disentangle, as there seems to be a general assumption that parties on the left tend
to be more programmatic while parties on the right tend to be more clientelistic.
This understanding makes it difficult to isolate each variable when conducting field
interviews with political elites and party operatives. Scholars should be aware of this
conflation and devise strategies to make this difference clear and therefore be better
able to assess the nuances that drive each dimension of these political linkages
independently.
Another caveat relates to the applicability of specific models, since they are
contingent upon the nature of the electoral system and adaptations should be
considered. For instance, the aggregate-level analysis of ideological competition
would be restricted to systems that practice runoff elections, in which case, it could
also be applied to subnational runoff elections. This level of analysis would be
151
especially useful as ideological politics are generally perceived to be lacking at the
local level where mayoral elections have greater potential to mobilize personal
votes. This model can also be applied to presidential elections with runoffs.
While this dissertation demonstrates that coalition formation is subject to the
operation of both ideology and clientelistic politics both quantitatively and
qualitatively, it appears that the statistical evidence concerning clientelistic linkages
is not conclusive. Several factors may have contributed to this methodological
shortcoming. One could be the dynamic nature of the political parties, which shift
position on the ideological political spectrum over time as suggested by Power and
Zucco (2009). Such shifts may cause some noise in the assessment of coalition
formation at the ideological level and may be particularly problematic for the
clientelistic level due to the elusive nature of this concept. Another possibility for
this shortcoming in the assessment of coalition formation could also be related to
the flexibility of the parties in adapting to local politics. In this sense, all else being
constant, it may be the case that the same party may behave more on a clientelistic
fashion in a private electoral marketplace and more programmatically in a public
electoral marketplace. Although the data allows for such nuance, more time would
be necessary to develop this idea.
Finally, the field research posited its own challenges. First, the sensitivity of
the subject makes it hard for some elected officials to openly and frankly address the
issue of clientelism -- especially when related to vote buying, which is an electoral
crime. For this reason, officials are more likely to talk about the “demand”
for clientelism from voters rather than t he “supply” of vote-buying exchanges.
152
Another limitation is related to the number of states and representatives
interviewed. The number of states visited was contingent upon funding. In this
sense, it would have been ideal to travel to more states to increase the sample of
both developed a nd less-developed areas. I compensated this limitation by spending
most of my time in the Federal District where representatives of all states congregate
from Tuesday to Thursday. Political elites do not grant interviews to scholars easily, a
factor that was complicated by the timing of the field research. Initially scheduled for
June of 2013, my field trip had to be postponed to avoid an unprecedented and
massive protest that took place in Brazil during that month. As a consequence, I
arrived in Brazil during the congressional recess, which reduced the number of
available representatives. To compensate, I interviewed other political informants
such as journalists, party personnel, and congressional staff.
Significance
This dissertation has focused on electoral competition in a multiparty
developing democracy by integrating the ideological and clientelistic dimensions of
political linkages. The findings represent an important contribution to theory
building on partisan politics (Amorim Neto and Santos 2001), ideology
differentiation (Luna and Zechmeister 2005, Power and Zucco 2009, Rosas 2010),
clientelism (Kitschelt, 2000, Remmer 2010, Roberts 2002, 2010), electoral coalitions
(Amorim Neto 2002, Golder 2005, Machado 2009, Mignozzetti, et . al. 2011, Nicolau
1996), and modernization (Borges 2011, Hagopian et. al. 2009, Hunter and Power
2007, Sugiyama and Hunter 2013).
153
This dissertation contributes to the literature on party development in
emerging democracies (e.g. Borges 2011, Hagopian et al. 2009, Hunter and Power
2007, Lyne 2005). It draws on spatial models of voting behavior in subnational
elections to advance an innovative endogenous statistical model to forecast runoff
elections for governors in Brazil. Additionally, it applies a new methodological
approach based on alternative-specific conditional logistic regression models. In this
perspective, the ideological models offer compelling evidence that ideology is the
main level of electoral competition, which helps to settle the debate on the
ideological differentiation of Brazilian political parties. Moreover, it also marginally
confirms strategic voting theories (Cox 1997).
This dissertation also offers a major contribution to the understanding of
clientelistic politics. It does so by building a completely new measurement of
clientelism based on the geospatial characteristics of the distribution of votes among
parties and how these parties devise strategies to garner these votes. It also
incorporates interesting insights from field research interviews that corroborate the
statistical models. A notable contribution relates to evidence suggesting that
developed areas (public markets) may require even more resources to meet the
demands of clientelistic-based voters.
Finally, the novel integration of both ideology and clientelism to explain
partisan politics through electoral coalitions also offer an insightful approach to
understand party development in modernizing environments. Since ideology
appears to be the main dimension of political competition, parties may resort to
other strategies, such as electoral coalitions, to gain access to electoral markets that
154
ideology alone otherwise can not. More studies, however, are necessary to refine the
combined utility theory developed in this dissertation to explain the two dimensions
of political competition (ideology/clientelism).
Future Research
The models developed herein to understand patterns of political competition
could be extended beyond Brazil to assess party competition in other multiparty
developing democracies. In this sense, some improvements are suggested. For
instance, future research could refine the deterministic model of spatial distribution
of votes by testing the probabilistic alternatives in runoff systems. While the
deterministic model applied in this study is simpler and more satisfying for runoffs
with two candidates, a probabilistic approach based on the spatial distribution of
votes may improve the predictability of the outcomes of runoff elections.
The individual-level analysis using conditional logistic regression has broader
application potential, as it is not constrained by the nature of the electoral rules but
only by the availability of national surveys that include self-ideological placement
and party choice plus external measures of party ideological placements. That is to
say, researchers will find many opportunities to apply these analytic and
methodological techniques for analyses of elections around the world.
The clientelistic models also seem to have a broader application potential as
they are determined by the availability of electoral data that could link
sociodemographic characteristics to votes. The validity of the measurements, in this
case, would depend on the availability o f data and the degree of socio-demographic
homogeneity/heterogeneity of the electoral districts in which this data could be
155
aggregated. For instance, if the state is relatively homogeneous
sociodemographically and the data available is aggregated by state, then it should
suffice. However, in relatively heterogeneous socio-demographic electoral districts,
the level of aggregation should reflect this difference at the lowest level possible,
either
by municipality or electoral zones.
Electoral coalitions also inform the theory developed herein but should not
limit its application. In fact, formal coalitions are easier to follow as they are
recorded by the electoral commissions; but even when the system lacks a clear
definition of electoral coalition, informal alliances, the previous governing coalition,
or the ex-post-facto governing coalition could offer alternatives to formal electoral
coalitions. The biggest challenge with this would be to track informal alliances, since
its mapping depe nds on the availability of written sources such as newspapers or
party documents.
Looking further, researchers should also account for the dynamics of
multiparty political party systems. In the case of Brazil, partisan politics have been
polarized between two programmatic parties (the PT and the PSDB), but the
electoral viability of thi rd parties may displace any or both parties in national
elections. In fact, all but the presidential election and re-election of Fernando
Henrique Cardoso have been decided in runoffs since re-democratization in 1989,
which suggests the viability of third parties. In the 2014 elections, for instance,
Marina Silva (PSB) jumped to second place in the polls, ahead of the Ae cio Neves
(PSDB) and behind Dilma Rousseff (PT). Eventually the second round was disputed
156
again between the PT and the PSDB, but the PT has lost critical political capital
following Rousseff’s inauguration into her second term amid an economic recession
and corruption scandal involving members of her party. Regardless of the
configuration of the main national electoral contenders, modernization suggests that
ideology plays an increasingly import ant and definitive role in partisan politics in
Brazil, with apparently ebbing clientelism. More study is necessary to determine
how parties access the dimension of clientelistic linkages in modernizing polities.
