You are on page 1of 46

Uneven Democracy?

Subnational Authoritarianism in Democratic Brazil
















1
More than twenty years since Latin America joined the global resurgence of
democracy, scholars have begun to wonder if the deficiencies of these young
democracies the lack of rule of law, rigged elections, human rights violations, to name a
few are bumps on the road of a long transition or permanent features of these regimes.
As scholarship on federalism and decentralization has become more prominent, some
observers have come to focus on subnational governments that appear and behave much
like the erstwhile authoritarian governments that were once replaced at the national level.
As nondemocratic polities that exist within national constitutional systems that define and
defend political and civil rights, subnational authoritarian governments restrict access to
power and exercise authority in ways that challenge these de jure principles. Far from
being a subtype of democracy or authoritarianism, subnational authoritarianism
represents a problem of governance within national democratic regimes. Subnational
authoritarianism challenges the quality of democracy as opposed to its existence.
1

Subnational authoritarianism is both a persistent and a pervasive threat to
democratization in Latin America. These subnational enclaves of limited electoral
democracy have been identified throughout the region as brown areas representing
exceptions to national, constitutional norms (ODonnell 1993; Diamond 1999: 133-4;
Chavez 2003). Speaking about the Mexican case, Wayne Cornelius (1999: 11) predicts
that the subnational political arena will be the principal source of inertia and resistance
to democratization [in that country]. Outside of the region, the uneven development of
democracy has been studied in India, Russia and Central Asia, and Southeast Asia (Beer
and Mitchell 2006; Heller 2000; McMann and Petrov 2000; McMann 2006; Way 2004;
Franco 2001; Sidel 1999).
2
Despite the importance of these authoritarian enclaves, studies of democratization
have neglected the examination of subnational authoritarianism because the literature has
focused almost exclusively on the national level of analysis (Snyder 2001; McMann
2006: 6-7). This national bias, while understandable during a time of national regime
transition and subsequent consolidation, makes less sense after more than two decades of
democratization and after numerous studies have pointed to the problems of democratic
governance embedded in local government. Moreover, patterns of national transition set
the stage for problems of subnational democratic governance. Third wave transitions to
democracy favored the consolidation of provincial authoritarianism by weakening the
centers previous direct control over subnational fiscal and policy authorities. This
allowed authoritarian elites at the local level to strengthen their respective bases of power
by providing them with greater relative autonomy and much-needed resources. In cases
where national politicians cultivated the support of subnational executives and their
political machines prior to founding elections, democratization at the national level was
engineered through a political arbitrage process in which local bosses could secure their
newly (re)gained autonomy and power as the price of their support. After being elected,
national incumbents in these cases found serving the interests of these subnational
bailiwicks a valuable method of maintaining themselves in power. Finding themselves
beholden to the masters of authoritarian enclaves, national leaders playing the democratic
game could gain strategic advantages at the center by advancing the interests of the
authoritarian periphery.
In many studies of subnational regimes, the central analytical dimension shaping
these polities is conflict between incumbents and the opposition. Given that subnational
3
authoritarians prosper in the absence of a strong opposition, the relative cohesion and
organization of opposition is a powerful predictor of electoral turnover (Van de Walle
2006: 78). Inspired by Schattschneider (1960), Gibson (2004; 2005: 107) argues that
subnational authoritarian incumbents are constantly engaged in boundary control, the
politics of closing down their oppositions access to national resources and possible
allies. The Achilles heel of subnational authoritarians, he argues, is that they need access
to national leaders to solidify their own subnational power, but the degree to which they
become embroiled in national politics makes them more vulnerable by opening up
another front at the national level that can weaken and undermine their control over local
bailiwicks. This logic produces a Faustian bargain that determines how effective
subnational incumbents will be in limiting the opposition. The degree of pluralism, then,
in subnational polities is the object of authoritarians strategic concerns, making it an
indicator of the degree of subnational authoritarianism.
Another dimension of pluralism is factionalism. Even in hegemonic party systems
incumbents may find themselves engaging in rearguard actions against rivals within their
own organizations. For example, Gibson (2005: 114-5) describes how Governor Jos
Murat in Oaxaca, Mexico, faced more opposition within the hegemonic party (the PRI)
and by national PRI interests in the 1998-2004 period than he did from politicians of the
opposition PRD. The emergence of opposition and its cohesion can be a function of how
effective party elites are in stemming the flow of rebels to the other side (Van de Walle
2006: 87). Incumbents not only must engage in boundary control against partisan
oppositions but they must also purge the state bureaucracy and their own party offices of
interests linked to rival factions.
4
Given the more recent concern for subnational authoritarianism as a problem of
governance in new democracies the literature on the causes of this phenomenon is
limited. Gibsons work represents an important, initial step in understanding the politics
of boundary control at the subnational level, but it may not be applicable to all cases.
First, the Faustian bargain of which Gibson speaks is less evident in the case of
authoritarian incumbents who have neither the desire nor the resources to venture into
national politics. Second, Gibsons approach works well in the Mexican and Argentine
cases, where the availability of highly disciplined and even hegemonic parties makes the
parameters of the bargaining between national and subnational elites straightforward. The
same cannot be said for Latin American countries with weak party systems such as
Brazil. The absence of hegemonic parties or the overall weakness of the party system
may prevent subnational incumbents from having either a regular voice or leverage in
national decision-making.
Nonetheless, the focus on incumbent-opposition dynamics does offer some
potentially important variables for determining levels of subnational pluralism.
Incumbents have incentives to control access to resources (boundary control), but their
capacity to do so is only as strong as the ability of the opposition to organize and garner
patronage resources to challenge established interests. In their study of electoral
authoritarianism, Hartlyn and McCoy (2006: 47) use the term the paradox of capacity to
describe how oppositions lacking the organizational, technical, and material resources to
provide oversight reinforce the ability of authoritarian incumbents to rule by fraud and
intimidation. By corollary, the greater the array of potential patronage resources
available, the higher the costs to incumbents of exerting controls on access and the lower
5
the costs to the local opposition of garnering these resources to mount their challenge.
The paradox of capacity extends to the bureaucratic and technical resources of the state
apparatus. Incumbents that fail to exert their monopoly of control over subordinates in the
public sector risk losing the coercive, extractive, and distributive services that reinforce
their rule (Way 2006: 167). Patronage leakage, therefore, can empower the opposition
and set up incumbents for an eventual electoral loss.
An alternative to the focus on incumbent-opposition dynamics is the political-
economic and geographic approach. It has long been argued that more diversified
political economies generate larger public sectors and more economic resources
(Schwartzman 1982). Insights from the study of democracy and economic development
suggest that the notable correlations between the maintenance of democracy and patterns
of development may play out even at the subnational level. Geographic factors such as
the size of the region may determine the availability of opposition. Small regions tend to
have more concentrated power structures and homogenous societies. Lacking geographic
diversity, political cleavages are more likely to be less differentiated than in regions with
multiple social and geographic centers of economic and political power (Reis and Castro
1992). Synthesizing these logics into an understanding of economic autonomy, Kelly
McMann (2006) argues that economic differentiation (capitalist development) provides
individuals with a measure of protection from the actions of autocrats by affording them
access to economic lives that are relatively independent of the resources public
authorities command. Where economies are less differentiated, economic autonomy is
low. Perhaps the best illustration of this tendency is the classic image of highly
agricultural, economically marginal regions that are dominated by powerful landed
6
families whose influence pervades government and economy alike (Nunes 1976). The
iconic representative of these polities is the landed colonel, a patron of a political
machine that runs an extended clientelist network.
This study argues that the politics of boundary control are more dependent on
political-economic differences, and particularly capitalist development, across
subnational units and over time. The degree to which subnational authoritarians can
preserve their rule with all of the attendant violations of rights and coercive impediments
to the opposition is only as likely as the political economic context will allow.
Authoritarians in backwater regions have a strategic advantage in maintaining their
positions and consolidating anti-democratic politics for the long haul since their costs of
coercion are relatively low. Of course, the power of these incumbents is not absolute.
These states may be less bureaucratically developed than those in more advanced areas,
thereby allowing for the leakage of patronage resources to the opposition. Yet this
tendency is tempered by the relatively low levels of patronage available in these states. In
more advanced capitalist economies where the degree of economic autonomy is greater,
the opposition is more likely to be developed. Patronage resources are more available in
these cases and the costs to incumbents of maintaining a monopoly over these resources
are high.
Political economic variables may weigh more heavily in democracies where
national parties are weak and provide few continuing or functional links between
subnational incumbents and national elites. In these cases, parties at the provincial level
may exert leverage in national politics but provide fewer of the opportunities or
incentives for authoritarians to pursue national careers that characterize Gibsons
7
treatment of the national-subnational Faustian bargain. Moreover, weak party systems
undermine attempts by incumbents to make their political machines hegemonic. Absent
programmatic and ideological sources of cohesion, undisciplined party organizations are
likely to produce factionalism and thereby encourage the emergence of rivals to
authoritarian leaders. In these cases, political economic variables will provide a more
consistent explanation for the rise and maintenance of subnational authoritarian polities.
The present study predicts subnational authoritarian governments as a function of
relative levels of economic development. Following the aforementioned paradox of
capacity, we expect state governments to be more closed to opposition and more directly
oppressive of civil society in cases where lower levels of economic autonomy have
weakened opposition actors. The political establishment in these states is more likely to
be able to concentrate their control over patronage resources with little threat of leakage
to the opposition.
The next section will discuss empirical measures for subnational authoritarianism
(the dependent variable) and it will be followed by an examination of these measures in
democratic Brazil. The subsequent section will employ a statistical test of independent
variables to assess the explanatory value of political, economic, and geographic factors.
The final section offers some conclusions and suggestions for further research.

Measuring Subnational Authoritarianism
All depictions of subnational authoritarianism revolve around the parochialization
of local politics; the closure of political exchange to a relatively small cadre of elites who
have shared interests in limiting access to public office to all but those loyal to them.
8
Consistent with scholarship on hybrid regimes known as electoral authoritarian or
competitive authoritarian, subnational variants of these polities permit limited forms of
pluralism in political society without wholesale restrictions on the franchise. Political
elites can compete for power, but a small, predictable few capture and hold it repeatedly
(Levitsky and Way 2002). This degree of parochialization allows for alternation in power
and high voter turnout, qualities that keep these polities within the realm of democracies,
but access by the opposition is restricted (Howard and Roessler 2006; Schedler 2006: 10;
Van de Walle 2006). These polities are, in the words of Kelly McMann (2006: 180-1),
strong in participation, weak on contestation.
Besides limited access, the authoritarian exercise of power is another dimension
of these subnational polities.
2
Subnational authoritarian governments violate both the
spirit and the letter of national democratic rules by infringing on the rights of individuals
and groups. This may occur directly through the explicit exercise of restrictions on the
opposition or indirectly by indulging or creating a climate of oppression and intimidation.
The latter is accomplished by breaking down the distance normally reserved in
democratic government between state officials such as police, administrators, and tax
authorities on one side and politicians on the other (Way 2006: 169). This conflation of
political interests and the coercive apparatus allows incumbents to use police forces to
harass and even kill, tax officials to threaten, and other public officials to cajole and
coerce support for state leaders.
As a problem of the quality of democracy, subnational authoritarianism is a
variable of degrees rather than discrete, ordinal types. But unlike national regimes, there
are no data that measure levels of subnational authoritarianism across units within
9
countries or across time. Nothing like Polity IV or Freedom House scores exist for
subnational governments. More important, many of the conceivable indicators for
subnational authoritarianism suffer problems of invalidity or data availability. The
following points attempt to take the most accessible indicators and discuss their relative
utility for empirical study.
Beginning with indicators of subnational pluralism, one accessible and systematic
indicator is the competitiveness of subnational elections for governor. Political theorists
from Madison and Schumpeter (1942) to Przeworski (1991) and Dahl (1971) have
observed that if the outcome of elections is predictable, democratic institutions cannot
keep power in check. Elections in subnational authoritarian regimes are not completely
rigged as the opposition enjoys the possibility, albeit limited, of unseating the incumbent,
but the level of contestation is low. Of course, electoral margins by themselves are
insufficient indicators of subnational authoritarianism because large ones are not
consistent in such polities and slender ones are not necessarily a threat to established
interests. As Gibson (2005: 102) observes, greater competition in southern Mexican
states after 2000 produced smaller margins in gubernatorial contests but established
interests retained their hold on power through the use of electoral fraud. Margins can also
shrink due to one-off factors such as economic crises that turn a large percentage of the
electorate against the incumbent.
Still, some aspects of electoral competition do offer some partial clues about the
degree of pluralism in a subnational polity. While margins by themselves might be
insufficient, the averages and the standard deviations of margins over long periods of
time provide a better sense of tendencies toward parochialization. Gubernatorial and state
10
legislative elections with large mean margins suggest the existence of political
establishments that are infrequently challenged. The dispersion of these margins indicates
how consistently dominant first-round winners are. The number of gubernatorial
elections that go to a second round also indicates how common decisive electoral
victories are over time. Finally, the effective number of parties (ENP) in gubernatorial
and state legislative elections as well as the percentage of seats in the legislature held by
the incumbents party are all indicators of pluralism. The latter is especially important
because oppositions in hybrid or competitive authoritarian regimes tend to build their
capacity for competing against incumbents by capturing and organizing within the
legislature (Levitsky and Way 2002: 56). Data for these indicators are usually available
and if pooled and analyzed over several elections, are likely to present some tell-tale
patterns of relative competitiveness.
The degree of turnover is another basic indicator of pluralism. The rate of
turnover predicts the risk of incumbent defeat. In polities governed as electoral
authoritarianisms, such as the archetypal cases in Southeast Asia, elections are organized
to provide feedback, not changes in government (Case 2006: 96). Turnover may be a
good indicator of factionalism as well if incumbents lose but their parties retain the seat.
Nonetheless, turnover may be a less valid indicator in countries in which institutional
impediments to reelection such as term limits produce a false impression of pluralism. In
these cases, individuals holding the governorship may change regularly, but the same
political families control the official apparatus. This underscores the need for multiple
indicators of pluralism including ENP, margins, and the legislative dominance of
incumbents parties.
11
The use of extra-judicial or illegal techniques are often cited as a defining
attribute of the politics of subnational authoritarians. The aforementioned example of
southern Mexico shows that electoral fraud can keep incumbents and/or their hegemonic
parties in place. Authoritarian techniques by definition extend to the use of physical
coercion, including assassination and other politically-motivated killing. A focus on these
factors, though, favors qualitative comparisons such as Gibsons since data on
subnational electoral fraud are most often case-specific or just anecdotal. Moreover,
electoral manipulation is unobservable ex ante and only detectable in the cases that it is
revealed through media exposs or official investigation ex post (Schedler 2006). These
measures may not even reflect the main reasons for low levels of pluralism. Not all cases
of weak opposition and hegemonic incumbents are the results of manipulative,
undemocratic action by established elites (Hartlyn and McCoy 2006: 46).
Clientelism as a mode of political organization appears in all cases of subnational
authoritarianism, suggesting that it is a defining attribute, though its utility as an indicator
is questionable. In his assessment of the Mexican states, Jonathan Fox (1994: 153) refers
to states that are authoritarian clientelist for the way that clients are persistently
subordinated politically at the subnational level. The rule of political families in the
Argentine provinces of San Luis, La Rioja, Catamarca, and Salta are also prominent
examples (Gibson 2005: 122). Yet like electoral authoritarian regimes at the national
level, the subnational variants do not follow entirely sultanistic tendencies. That is,
personalism and the dispensing of patronage are tools to maintain power, but they are
hardly the lifeblood of politics. Incumbents must still keep power through multiparty
elections and that means that they have to mobilize their voters and demobilize those of
12
the opposition. Moreover, in national polities where the use of clientelism is systemic, it
is difficult to discriminate among subnational governments that employ it more or better
than others.
Following the precept that authoritarian governments must at least act like
authoritarian governments, indicators of arbitrary use of public authority to oppress
members of civil society through human rights violations are prima facie measures of the
degree of authoritarianism. Once again, the availability of systemic data at the
subnational level is a problem, but it is not one that is insurmountable considering the
number of official and non-governmental entities presently collecting information on
rates of official violence, including police homicides. One may object that such acts may
not represent the politics of incumbents but are the actions of others. Still, the consistency
of these problems over time reflects a more fundamental tendency on the part of political
incumbents to countenance or condone these acts as a normal part of the way the
subnational polity functions.
Given the exploratory nature of current studies of subnational authoritarianism,
this paper argues in favor of an eclectic set of indicators that incorporates an array of
measures of pluralism and government behavior. These are presented in the next section
with a focus on the Brazilian experience.

Uneven Pluralism and Democratic Governance in the Brazilian States
A comparison of state governments in Brazil provides a good test of the wider
incidence of subnational authoritarianism in Latin America. First, the choice of Brazil
diversifies the extant scholarships focus on Mexico and Argentina within Latin America
13
(e.g., Fox 1994; Chavez 2003; Snyder 2001; Cornelius et al. 1999; Gibson 2005). This
case selection has favored explanations focusing on changes in hegemonic parties such as
the PRI or highly disciplined parties such as the Partido Justicialista (Peronists). By
contrast, Brazilian parties with very few exceptions are highly undisciplined (Mainwaring
1999). Even the relatively disciplined Partido dos Trabalhadores (the Workers Party) is
hardly hegemonic in its ideological proclivities or its leadership group (Lacerda 2002).
Second, there is great variance in the relations between incumbent elites and oppositions
among the Brazilian states. To be sure, subnational authoritarian governments in Brazil
are what Lucan Way (2006: 170) would call weak authoritarian states since they
confront large and autonomous business classes and civil societies. Though, as I argue
here, the capacity of the opposition differs across subnational units, making some of the
Brazilian states pure electoral democracies, others hybrid regimes, and perhaps one or
two de facto electoral authoritarian cases. Finally, there is much variance in political
economic levels of development across the states (Schwartzman 1982). This allows for
testing political factors and political economic independent variables across time and
space.
Using electoral and human rights data for several of the core indicators discussed
above, I found systematic distinctions in the level of pluralism and democratic
governance among the twenty-seven Brazilian states from 1982 to 2002. The first
subsection below focuses on indicators of pluralism for all 27 states while the second
examines extrajudicial killings as a proxy for governance in 19 Brazilian states.

Uneven Pluralism
14
Using data compiled by Nicolau (1998, 2005) for gubernatorial and state
legislative elections from 1982 to 2002, I calculated average scores for first-round and
second-round margins, their standard deviations, the effective number of parties (ENP)
based on votes in gubernatorial contests and seats in the state legislature,
3
the incidence
of run-offs and re-election of incumbents, and the average percentage of state legislative
seats held by the parties in the winning and rival coalitions (mean legislative
concentration). The distribution of values across the 27 Brazilian states as reported in
Table 1 evinces a three-fold categorization of subnational polities that we label here
conservative, moderate, and broadened competition. As we argue below, these categories
are not internally homogenous. The attempt here is to specify some of the deviant
patterns within categories as well as draw distinctions among the three groups of states
we believe reflect certain similar attributes.
[Table 1 about here]

Conservative competition Eleven states (Mato Grosso, Bahia, Maranho, Cear,
Alagoas, Tocantins, Sergipe, Pernambuco Gois, Paraba, and Amazonas) rank at the top
in order of mean first-round margins. This group contains the only four states to never
have a second-round contest for governor (Mato Grosso, Alagoas, Pernambuco, and
Amazonas) and, with the exception of Paraba, they all re-elected governors at least once
(five did twice, and one, Amazonas, three times).
4
The limited competitiveness of these
states elections is also reflected in average ENP scores for governors races. These states
have the lowest average ENP scores with none exceeding 2.4. (Only three of the other
states in the other categories have 2.4 or lower ENP scores). The dominant parties tend to
15
be the catch-all, right-wing PFL (Party of the Liberal Front/Partido da Frente Liberal)
and PDS (Social Democratic Party/Partido Democrtico Social), and center-right PMDB
(Party of the Brazilian Democratic Movement/Partido do Movimento Democrtico
Brasileiro), commensurate with the conservative profile of these cases. The center-left
social democratic PSDB (Party of Brazilian Social Democracy/Partido da Social
Democracia Brasileira), which is dominant in only the case of Cear, elected a winning
governor only five times (including one re-election) in the other states. The progressive
PT captured no governorships in this cohort and it found itself in only four partisan
coalitions of successful candidates. In virtually all cases, conservative candidates and
right-wing parties retained control of the governorships of these states.
The conservative competition cohort may be subdivided in terms of the dispersion
of first-round margins. The largest standard deviations of first-round margins occur in
this group, with Maranho, Paraba, Sergipe, Alagoas, and Mato Grosso representing the
highest dispersion rates and Tocantins, Gois, Cear, Amazonas, Pernambuco, and Bahia
representing the lowest. This difference reflects distinct conditions facing conservative
establishments in these states in particular elections, although the overall nature of these
regimes is relatively low competition among conservative interests. The variation of
margins reflects first-round victories that are small when notables of the conservative
establishment of these states battle between each other for the governorship and large
when conservatives are united either against a leftist or populist threat.
The legislative profile of these states elections follows the dynamics of the
governors races. Consistent with the relatively larger first-round margins and low ENP
scores of this group, the parties of the winning and rival coalitions represent on average
16
85 percent of the party-seats in the state legislature. This is an indicator of the importance
of gubernatorial contests in these states for maintaining political machines. Even minor
parties have strong incentives to join one of the two rival coalitions to claim rights to the
patronage of the executive office after the election. Not surprisingly, average legislative
margins for the parties of the coalition supporting the winning gubernatorial candidate are
the largest of all of the states. Only in Alagoas does the rival coalition win a larger
percentage of seats, but that occurred during the second half of the 1990s when the
conservative establishment was challenged by the rise of a left-of-center governor,
Ronaldo Lessa. The same dynamic appeared in Mato Grosso where the legislative margin
was uncharacteristically in the single digits. Excluding these two states in the top cohort
the mean legislative electoral margin is 26 percent in favor of the coalition of parties
electing the governor.

Moderate competition The eight states that compose the middling cohort (Esprito
Santo, Acre, Rio de Janeiro, Paran, Amap, Mato Grosso do Sul, Par, and Rio Grande
do Norte) are more developed cases of tendencies present in conservative competitive
states such as Alagoas and Mato Grosso, namely, the erosion of the hegemony of right-
wing parties and personalities and the emergence of more consistent and viable left-wing
or populist challengers. In these states, conservative notables in the PMDB and PFL were
unable to maintain anything akin to the position their co-partisans enjoyed in Bahia or
Maranho. All of the states in this cohort had run-offs and fewer former governors return
to the governors mansion. The number of parties that compete was also greater, as
demonstrated by the higher ENP scores for both gubernatorial and legislative races.
5

17
With the exception of Rio Grande do Norte and Mato Grosso do Sul, which maintained
the lowest joint ENP numbers of the cohort, the state legislatures in this group were not
under the complete control of parties that ran candidates in the gubernatorial races. The
more fragmented polity in these cases undermined the dominance of any one political
machine of conservative notables. Consequently, the center-left PSDB and the leftist PT
made more inroads in gubernatorial and legislative races in these states. The more
consistent challenge of the PT was particularly notable as the party won governorships in
Esprito Santo, Acre, and Mato Grosso do Sul and it joined winning coalitions in Rio de
Janeiro and Amap.
The other salient quality of this cohort is that gubernatorial elections are more
consistently competitive than those in the conservative cases. A wider array of rivalries
develops in these cases that is not present in the conservative cohort. The three states with
ENP scores above the average for the cohort (Rio de Janeiro, Amap, Par) demonstrate a
full range of rivalries: conservative vs. populist, populist vs.leftist, center-left vs. center-
right, etc. The case of Rio de Janeiro is particularly notable as populist governors such as
Leonel Brizola and Anthony Garotinho (PDT) battled conservatives (Wellington Moreira
Franco, PMDB-PFL), leftists (Jorge Bittar, PT), and center-left moderates (Marcello
Alencar, PSDB), often getting the upper hand with larger margins, but sometimes losing
in close contests. Rivalries in Amap and Par forced run-offs four times in these poor
northern states. But these were not just cases of strong PT-led challenges to conservative
parties. In some cases, PSB-PT incumbents defended against conservative and populist
candidates and vice versa (Amap) or conservative candidates challenged center-left
PSDB incumbents (Par).
18
The other states offer numerous examples of the same tendencies, with the
common denominator being more diverse rivalries than those evident in the conservative
competition cohort. Only Rio Grande do Norte seems to reflect tendencies of
conservative competition. In this case, traditional elites in the divided conservative
establishment battled one another and never needed an alliance to fend off a populist or
leftist challenge. Given its mean legislative concentration score and its ENP, Rio Grande
do Norte could easily be placed in the conservative competition cohort if it were not for
the highly divided nature of its right-of-center polity. Its mean first-round margin is the
lowest of the moderate competition cohort.
Several indicators illustrate that legislative elections are more competitive in the
middle cohort than in the conservative group. The average legislative margins of these
states are well below the scores of the conservative cohort. ENP scores are significantly
higher than those of the conservative cohort (t=-2.486, p=.024) and mean legislative
concentration is significantly lower than in the least competitive states (t=3.131, p=.006).
That is not to say that these states are devoid of political machines, but only that several
such groups battle for the political control of the governors mansion and the state
legislature in each election.

Broadened competition Eight cases compose the cohort of the most competitive states
(Santa Catarina, Rondnia, Distrito Federal, Minas Gerais, So Paulo, Roraima, Rio
Grande do Sul, and Piau). Although the ENP scores, mean legislative margins,
legislative ENPs, and mean legislative concentration numbers are not significantly
different from the moderate competition cohort, these states show several qualitative as
19
well as quantitative attributes of higher competition. First-round margins are significantly
different from those of the moderate cohort (t=-5.466, p<.001) as are second-round
margins (t=-2.301, p<.05).
6
Second, the most competitive cohort has more run-offs and
fewer re-elections than the moderate competitive states.
7
Third, as in the cases of
moderate competition, a variety of political agents win office in these cases. PSDB and
PT have more governorships and participate in more winning coalitions than in the other
two cohorts. But more notable is the fact that a larger variety of independent candidates,
populists, and right-wing elites compete for power. This gives competition in the bottom
cohort a breadth as well as an intensity that separates these states from the other two
groups.
Some of the states in the broadened competition cohort are surprising. Piau mixes
a very high legislative concentration score (90 percent) with the lowest first round margin
numbers of all of the Brazilian states. A closer look reveals a long history of deep intra-
conservative competition, independent candidacies, and the success of the PT in recent
years without the reactive conservative alliances seen in many of the least competitive
states. These factors are replicated in other northern and northeastern states with low
first-round margins (Rio Grande do Norte, Roraima, and Rondnia), although in these
cases it is parties such as the PPB, PSDB, PDT, PSL, and not the PT that expand the
range of non-conservative forces.
The other states in the most competitive cohort are not surprising: So Paulo, Rio
Grande do Sul, Santa Catarina, Minas Gerais, and the Distrito Federal. These are the most
socio-economically developed states and therefore places where the widest range of
20
interests are represented. If it were not for their relatively higher first-round margins, Rio
de Janeiro and Paran might also be in this cohort.

Uneven Democratic Governance
Finding comprehensive human rights data for the Brazilian states is complicated
by the lack of a central statistical authority with complete time-series or cross-subnational
data. Acquiring and analyzing data on perhaps the most obvious indicator of official
miscarriage of authority the incidence of extrajudicial killings by police was
especially a fools errand until recently. Using raw data collected by the National Human
Rights Movement (Movimento Nacional de Direitos Humanos, MNDH) from newspaper
reports of homicides, Ronald Ahnen (2007) has recently analyzed data for 19 states and
for a broader time frame (1994-2004) than has heretofore been available. Using this data
on police homicides as a proxy, the current study measures respect for human and
individual rights by public officials. The validity of this indicator is based on the idea that
police abuses of power are both an indicator and a variable: they are indicative of a
generalized ambiguity of law that continually protects the wealthy and powerful, but
they are also instrumentalities of inequality, tools of oppression that are applied
predictably and selectively to the poor and the landless (Pereira 2000).
[Table 2 about here]
Using general homicide rates based on mortality data as reported in the Ministry
of Healths database (DATASUS) and MNDHs raw police homicide data, Table 2
orders twenty-six Brazilian states from highest police homicide rates to the lowest.
8
As
with the raw homicide data from the Ministry of Health, it is likely that the MNDH data
21
are biased due to over- and under-reporting in newspapers. Ahnen (2007) calculates a
correction based on differences between official statistics and the MNDH data. The tests
below include model specifications using the estimated figures employing Ahnens
correction.

The Study
Which variables explain the distribution of pluralism and democratic governance
across the Brazilian states and over time? To answer this question, I designed statistical
models for time-series cross-sectional (TSCS) tests of several core indicators of
subnational pluralism and police homicides as the main proxy for subnational democratic
governance. The pluralism models focus on ENP of gubernatorial elections as the
dependent variable. The comparison of the three state cohorts suggests that as the
effective number of parties increase, margins should decline. This expectation is
consistent with that in the extant literature on ENP and partisan competition (e.g., Hecock
2006) and is affirmed in preliminary tests using the Brazilian data that found a strong
inverse correlation between ENP in gubernatorial races and first-round margins. The
second set of models use police homicides as the dependent variable. This variable is
weighted by the population of each state. Below we discuss each independent variable in
terms of how it should predict changes in pluralism and governance across the
subnational units and over time.

Political Factors
22
According to the aforementioned paradox of capacity we expect factors specific
to the organizational capacity of the opposition to determine the pluralism of the
subnational polity. We measure this variable straightforwardly as the share of seats in the
state legislature of the coalition of parties supporting the runner-up. The models also
include the seat share of the party of the runner-up. Using ENP of legislative seat shares
as a control, we expect that opposition parties with strong positions in the legislature will
increase the competitiveness of gubernatorial races, particularly in the first round where
there are few institutional disincentives to run candidates.
Conceivably, less competitive states are dominated by conservative parties. After
all, these parties tend to be in power in the northern and northeastern states that dominate
the conservative competition cohort. Conversely, the greater role of leftist parties such as
the PT, especially in the more industrialized states of the south and southeast, may
explain patterns of broadened competition. The political models take this into account by
controlling for the presence of PFL and PDS, the most conservative parties, and the PT,
the largest leftist party, in winning coalitions. ENP in gubernatorial contests across units
and time should decline if the conservative parties tend to support the winner and they
should increase where the PT supports the winner. One model also includes the PMDB
and PSDB as catch-all center-right and center-left options, respectively, to test the
consistency of findings for the more conservative and leftist parties. If party identity
matters, then these parties should have no or little discernable effect on ENP.
Other political factors may intuitively affect pluralism. The incidence of runoffs is
one. There were runoffs in over forty percent of all state elections after 1990, when a
second round was first required if a candidate did not receive a majority in the first round.
23
If a runoff occurred in the previous election, political agents may expect another close
call and organize their co-partisans to stave off a repeat of the past. This may mean
establishment parties offer greater concessions of patronage to allies to get their support
in the first-round to secure an early win. If runoffs occur with great frequency, this
calculus will become less useful. The past can also play a role in the case of candidates
that are seeking re-election as incumbents or as ex-governors giving it another go. Sitting
and ex-governors are apt to wield patronage more effectively than their challengers to
secure first-round victories. Therefore, we might expect states with a penchant for run-
offs to be more competitive overall and states with a tendency to see incumbents or ex-
governors be re-elected to be less competitive.
In the police homicide tests, these political factors may all work in the same
intuitive manner they do in predicting degrees of pluralism. The organizational capacity
of the opposition, the presence of leftist parties in winning coalitions, and the high
incidence of run-offs and low rate of re-election should be associated with lower rates of
extrajudicial killings. Conversely, weak oppositions, conservative party dominance in
winning coalitions electing the governor, low run-off rates, and high incidences of re-
election should be linked to more extrajudicial killings.

Socio-Economic Variables
If Table 1 were converted into a map, one might easily conclude that the
conservative competitive states tend to be agricultural, low growth, and generally the low
population states in the northern and northeastern regions. The history of states such as
Maranho, Alagoas, and Bahia evoke images of extractive, heavily agricultural
24
economies with landowner-dominated polities. These are the attributes most associated
historically with colonelismo and patron-client politics in Brazil (Nunes 1976). As
suggestive as this erstwhile impression is of the states of the north and northeast, it is
apparent that not all the states in these regions fall within the conservative competitive
cohort. This begs the question whether the socio-economic characteristics of the Brazilian
states help to explain the overall pattern of their political competition.
The first dimension of socio-economic differences is population and percentage of
the GDP represented by each state for each year. We found a strong correlation between
these two variables as one would presume from the fact that the more populous states
compose a larger share of the GDP. Following much scholarship on the political
economic dynamics of subnational polities in Brazil, we should expect the more
industrialized states to have more competitive politics while the more agricultural ones
should be in the conservative cohort (Stepan 2000; Snyder and Samuels 2004). However,
preliminary tests on the percentage of the economy represented by industry or agriculture
failed to yield confirmation of this presumption. This may have to do with the nature of
the data. Agroindustry is sufficiently capital intensive and industrialized to blur the lines
between agricultural and industrial sectors. The service sector, which is not measured
separately, is now more important and more dynamic than either industry or agriculture.
A comprehensive variable such as the relative share of GDP measures can serve as a
proxy for the overall sectoral development of the subnational economy, so we employ it
in the models.
We also calculated a growth rate in the relative share of GDP term and used it in
several tests. The arguments regarding growth in relative shares are more nuanced with
25
two hypotheses seeming plausible. An increase in a states relative position could create
incentives for new capital to back candidates outside the establishment, especially if
incumbent powers do not respond to the interests of new sectors of the economy.
Alternatively, declines in relative shares of GDP could cause capital connected to the
establishment to force a change so as to not see the state fall further behind its rivals. This
would increase competition among traditional elites, some of whom would appeal to the
reactionary movement while others would defend the status quo. Such an outcome is
suggested by patterns of occasional high competition among the states in the conservative
cohort, but it remains to be seen if this tendency explains change in the other cohorts in
which competition among conservatives is less important. If incumbent powers are
responsive, new capital might invest in the political establishment so as to secure their
property rights, thereby linking improvements in relative economic position of the state
to increased competitiveness.
The second dimension of socio-economic differences we tested was growth and
transformation of the state economy. Given Brazils numerous monetary and financial
crises, working with comparable time-series data for subnational growth is treacherous.
Fortunately, the Brazilian Institute of Geography and Statistics (IBGE) calculates a
growth index that allows for standardized, longitudinal and cross-sectional comparisons.
The index, which begins in 1985 (100) and proceeds to the present, is based on the value-
added to each state economy by all economic activities. Since the index shows an
exponential increase for all states from 1985 to 2002, we used the natural log of the index
as our variable.
9

26
As argued above, we expect socio-economic variables to play an important role in
explaining subnational pluralism and democratic governance. Evidence from the three-
cohort comparison suggests that there might be some merit to the role of relative GDP
shares, a position somewhat consistent with historical approaches to Brazilian politics.
We are less certain about the effects of changes in the economy and growth on
democratic governance. Economic modernization could be associated with greater
pluralism and therefore more oversight and limits on arbitrary uses of political authority.
Alternatively, such transformations can generate backlashes by incumbent authorities
against political rivals.

Government Size Variables
Most of the literature on gubernatorial politics emphasizes the central role of
patronage in linking political ambitions and policy-making in Brazil (Samuels 2000,
2003). Government spending is the lifeblood of patronage politics, so we expect
expenditures and the weight of government in the state economy to affect the
competitiveness of gubernatorial and legislative elections. Yet it is plausible to imagine
two alternative hypotheses. First, that as the size of government expands, the availability
of patronage to incumbents makes challenging the political establishment of the state
more difficult. By undercutting the electoral capacity of the opposition, incumbents are
able to secure their own re-election or elect their successors. This is consistent with the
image of conservative machine bossism. Alternatively, the greater availability of
patronage encourages political rivals to the incumbent to organize their own forces to
garner the more considerable rewards of executive office. Expenditures thus reinforce the
27
ambitions of the opposition and raise the costs to incumbents of preventing patronage
leakage. Poorer states are less likely to create these opportunities for the opposition due
to the lack of patronage, but this too may depend on the ability of incumbents to more
effectively control the state apparatus. Despite the possible caveats, we expect
government size to be positively correlated with pluralism and inversely with police
homicides.

Other Control Variables
Fiscal, demographic, and social development variables may all potentially
impinge upon the relative pluralism and democratic governance of subnational polities in
Brazil. Regarding the fiscal, the capacity of governors to secure fiscal resources is a sine
qua non of their ability to dispense patronage. Although all the state governments can
extract tax revenues through their control of the ICMS (value-added tax), real revenues
vary by state due to the different sizes of tax bases. Poorer states rely more heavily on
fiscal transfers that are constitutionally guaranteed. Recent work that applies rentier
theory to fiscally federal cases argues that this dependence on transfers determines the
characteristics of the regime (Gervasoni 2006). Accordingly, one may expect that states
with a greater dependence on fiscal transfers are less dependent on the need to cultivate
popular support since they do not require the taxation of their people. Conservative
oligarchies can count on fiscal transfers to avoid taxing (or representing) much of the
population and diverse capitalist interests. And they can still produce sustained popular
support through the distribution of patronage. Therefore, greater fiscal autonomy through
28
taxation and less of a reliance on fiscal transfers should be associated with more
pluralism and better democratic governance.
The first two pluralism models tested below include regional dummies to account
for geographic factors. The specifications control for five regions: northeast, north,
southeast, south, and center-west. States were designated to regions following the IBGEs
categories. The regression tables do not report the coefficients for these dummies to
facilitate the presentation. Models 3 and 4 in the pluralism tests do not include the
regional dummies. In the governance models, the lag of homicide rates as reported by
DATASUS is used to control for crime as a cause of residual police violence in these
models. Since these rates are known to underestimate the real number of homicides, we
employ Ahnens (2007) correction to these figures.

Results
The pluralism models provide support for the proposition that relative capitalist
development predicts the competitiveness of subnational polities in Brazil across space
and time. Evidence for the importance of political factors in general was inconsistent. As
Table 3 demonstrates, unit shares of GDP were strongly and consistently correlated with
higher ENP scores in the Brazilian states from 1982 to 2002. This is true controlling for
political factors such as the strength of the opposition, ENP of seats in the legislature,
differences in which parties supported winners, the incidence of run-offs and re-election,
government size, and other fiscal and economic factors. Model 1 offers some support for
the argument that the capacity of the opposition matters, though it is the seat share of the
chief rivals party and not the wider coalition of parties that explains ENP of
29
gubernatorial contests. However, once this variable is retested in model 2, it loses
statistical significance and changes sign.
10
The removal of the rival coalition variable
might have explained the loss of significance of this variable, but model 3 demonstrates
that share of seats of the rival coalition without this other factor is insignificant.
[Table 3 about here]
The effect of partisan identity of the winning coalition is uncertain. As predicted,
the conservative parties are associated with lower ENP scores, but so is the leftist PT. As
we would predict the PT to be part of winning coalitions in more competitive systems,
this was an unexpected finding. We tested this factor further in model 2 by adding the
main catch-all party of the center-right, the PMDB, and the center-left PSDB. These
coefficients were also negative and significant (though PSDB is insignificant in model 3).
As a result, there is no clear interpretation of partisan identity and ENP scores across the
Brazilian states. Conservative parties may not be any more associated with parochial
subnational polities than are the other major parties.
Notably government size as a proxy for the availability of patronage does not
have an effect on subnational pluralism. In preliminary tests we examined the hypothesis
that this factor may interact with capitalist development or industrialization, but the
resulting interaction term proved insignificant in all tests. Neither dependence on fiscal
transfers nor the economic growth factors proved significant in models 1 and 2,
suggesting that it is the complexity of the capitalist economy, its structure and not its
performance, that is the key predictor of subnational pluralism. Model 3 shows that fiscal
transfers are positively and significantly correlated with greater pluralism, but this defies
the logic of rentier theory which predicts the inverse relationship. Rentier effects may still
30
play a role in that transfer dependence may limit economic development, thereby
requiring further testing of these variables (Wantchekon and Asadurian 2002). Models 1-
3 explain a reasonable degree of the variance of ENP scores.
We retested the core specification in a fourth model that regressed an index of the
dependent variable. Using the first factor of a principal components analysis (PCA) of
ENP variables, absolute value of first-round margins, the reelection variable, and
legislative concentration (the percentage of seats held by the two largest parties), we
constructed an index term as an alternative to the ENP of gubernatorial elections to see if
this would change the results of the core model. Since the index term measures
subnational autocracy, the signs are reversed for the purposes of interpretation. The
results generally verify those found in the first three models with the exception that share
of seats of rival coalitions is shown to deepen authoritarian tendencies, a counter-intuitive
outcome that will require further testing of the PCA model. GDP share remains
significant and its sign runs in the expected direction. The logged growth term becomes
significant in this model, suggesting that not only more developed but better performing
economies have more plural political societies.
Regarding democratic governance, Table 4 presents the models predicting police
homicides per hundred thousand inhabitants in each unit and over time. Once again,
relative shares of GDP performed as expected with the more developed and complex
economies representing lower rates of extrajudicial killings. Unlike the first three
pluralism models, the growth term in relative shares was consistently significant and
positive in all tests. This finding suggests that economies undergoing more rapid changes
in development patterns generate political tensions that result in violations of human
31
rights. This is consistent with erstwhile arguments such as Huntingtons (1968) that rapid
economic transformation undermines traditional political systems in ways that lead to
breakdowns in order rather than transitions to democracy.
[Table 4 about here]
Political factors in the governance models proved inconsistent in explaining
police homicides. Neither of the variables measuring the organization of opposition
achieved statistical significance. The PFL/PDS variable was consistently significant but
with a negative sign, indicating that conservative parties are associated with lower levels
of extrajudicial killings. In Model 2, the PMDB variable also proved negative and
significant (p<.1), demonstrating that incumbents with center-right and right-wing
support are not more likely to countenance violations of human rights. Indeed, they lead
polities with lower police homicide rates. Taken together with the findings in the
pluralism models, there is no consistent support for the notion that partisan identity, and
especially conservative leadership, is a distinctive attribute of subnational
authoritarianism. Of course, the weakness of parties in Brazil throws partisan
explanations into doubt, making this an area for further research into political variables.
The only other political factor that proved significant and consistently so in all
tests was the incidence of run-offs in the past.
11
Both governance models demonstrate
that states with a history of competitive first-round gubernatorial contests produce
tensions that are associated with extrajudicial killings by public servants. Once again, this
variable may work in much the same way the growth in relative shares of GDP works.
That is, societies facing acute change in economic structures and greater competition for
32
state power undermine status quo arrangements, leading to the use of extrajudicial
responses by incumbent interests.

Conclusions
The results of the statistical tests offer a preliminary suggestion that the progress
of democracy at the subnational level is arduous and potentially contradictory to national
de jure principles of democratic governance. The most consistent finding is that capitalist
development and different levels of economic autonomy across units and over time
enhance both pluralism and the exercise of democratic rule. Simpler and lesser developed
economies offer fewer options for citizens to become economically autonomous from the
state and incumbent interests. These conditions lend these polities a conservative
tendency, though this is not necessarily reflected in the continuation of conservative and
right-wing parties. Partisan identities and the organization of political society appear less
important than social and economic structures.
The results of the governance models also suggest that changing subnational
authoritarian polities is treacherous. Socio-economic transformation of subnational
polities in Brazil is associated with the tendency to pursue anti-democratic solutions to
the inevitable social tensions wrought by modernization. Moreover, there is no
ameliorating role played by the public sector. The size of the public sector neither
provides enough patronage to reduce the threats felt by incumbents nor does it offer any
of the social protections that lessen tensions in rapidly changing societies in more
advanced capitalist states. These observations underscore the more pessimistic view that
competitive authoritarianism represents a problem of the quality of democratic
33
governance that will not be easily resolved as part of a larger and sustained process of
democratization. Further democratization in these cases may be possible, but the process
of change may itself deepen authoritarian tendencies in the medium-term.
Structural and not institutional explanations seem to enjoy the strongest support in
the present study, though these findings may provide guidance to future work on
institutional approaches. Much current work on decentralization, and particularly the
fiscal and policy dimensions, tends to view the process as dominated by differences in
national political institutions such as party systems and inter-governmental structures
(e.g., Willis, Garman, Haggard 1999). The analytical approaches of these studies embrace
collective action logics that see subnational interests existing in competition or contingent
cooperation with national actors. This framework territorializes national politics but it
does not take territorial interests seriously. As Tulia Falleti (2005) has argued, the
interests of governors and state legislatures are not reducible to a universal set of
preferences. These actors have preferences that have different scope and dimension based
on the practices, informal understandings, and institutional structures of the subnational
polities in which they are embedded. Beginning with a political economic approach to
unpacking the way that territorial interests are themselves shaped by factors endogenous
to subnational societies and polities will produce more powerful hypotheses about the
role of political institutions. In this regard, understanding the politics of boundary
control and integrating intergovernmental relations; that is, national-subnational
interaction, in the analysis will require an understanding of socio-economic change and
how it affects institution-formation and political practice.

34











Table 1 Indicators of Pluralism in the Twenty-Seven Brazilian States, 1982-2002

Gubernatorial Elections State Legislative Elections
State
1st Rd.
Mean
Margin
St.
Dev.
2nd d.
Mean
Margin St. Dev.
ENP
(votes)
Run-
offs
Re-
elected
1 Rd.
Mean
Margin St. Dev.

ENP
(seats)
Mean
Legislative
Concentration
Conservative
Competition
Mato Grosso 29.4 18.9 - - 2.1 0 1 4.9 26 4.6 86.9
Bahia 28.5 14.7 17.2 - 2.3 1 1 32.2 17 4.7 79.3
Maranho 28 28.8 4.5 4.7 2.1 2 1 37.9 24.6 5.4 83.3
Cear 27.4 10.9 0.1 - 2.2 1 2 28.3 14.8 3.9 83.4
Alagoas 25.4 22.9 - - 2 0 2 5.1 24.2 5.5 83.9
Tocantins 22.3 7.5 11.8 - 2.2 1 1 28.2 26.7 4 92.8
Sergipe 21.3 23.5 7.4 3.5 2.2 3 2 27.1 42.4 5.7 86.8
Pernambuco 19.9 12.3 - - 2.1 0 2 9.2 11.3 5.4 83.6
Gois 19.5 10.6 9.7 4.4 2.4 2 2 15.6 27.7 4.5 81.7
Paraba 18.9 24.4 9.9 6.9 2.2 3 0 16.7 35 3.5 88.0
Amazonas 18.6 11.9 - - 2.2 0 3 36.9 15.9 5.4 85.5
Moderate
Competition
Esprito Santo 18.3 16.4 22.8 14.5 2.5 2 0 11 12.2 7.6 48.6
Acre 18.3 13.5 8.3 1.3 2.6 2 1 11.8 5.5 5.1 74.3
Rio de Janeiro 17.8 14.8 14.1 2.7 3 2 1 14.1 14.3 7.5 56.0
Paran 13.7 18.9 12.7 3.1 2.6 3 2 15.8 23.5 5.4 77.0
Amap 13.1 4.4 15.1 12.7 3.2 4 1 2 24.3 7.7 56.8
Mato G. do Sul 13.1 13.6 15 10.7 2.3 2 2 1.4 24 5.8 82.0
Par 11.7 17.9 9.3 11.1 2.8 4 2 8.2 39.3 5.5 79.7
Rio G. do Norte 9.2 5.1 13.1 12.6 2.4 2 2 5.6 23.2 3.7 90.3
Broadened
Competition
Santa Catarina 9 19 1.1 0.7 2.6 2 1 9.6 21.6 4.1 75.4
Rondnia 8.4 11.7 15.6 8.4 3.9 4 0 4.1 34 6.9 49.1
Distrito Federal 7.8 18.4 4.1 3.4 2.8 3 2 11.2 25.6 8.5 63.2
Minas Gerais 7.2 17 11.7 8 2.8 3 0 6.5 25.1 6.8 62.6
So Paulo 7.2 16.6 11 5.6 3.4 4 1 11.8 12.6 6.2 56.6
Roraima 7.1 5.8 9.6 5.1 2.7 4 1 12 19.4 7.7 69.0
Rio G. do Sul 7.1 8.6 8.5 9.4 2.9 4 0 1.5 28 5.4 62.7
Piau 3.7 9 7.7 5.1 2.3 3 0 12.3 30.6 3.5 90.0
35


Table 2 Police Homicide Data
States
Average
Yearly
Homicides Years
All Police
Homicides per
100,000
People
Esprito Santo 1092.89 94-02 36.31
Roraima 79.44 94-02 25.96
Acre 116.75 94-01 21.90
Mato Grosso 422.67 98-00 17.46
Alagoas 480.11 94-02 17.23
Amap 69.50 98-01 15.65
Pernambuco 1211.73 92-02 15.58
Sergipe 245.40 97-01 14.17
Mato G. do S. 284.75 99-02 14.07
Amazonas 374.60 94-98 13.92
Paraba 451.89 92-00 13.29
Rio de Janeiro 1873.57 95-00 13.29
D. Federal 247.00 94-00 12.56
Bahia 1372.20 96-00 10.67
Tocantins 118.75 95-02 10.65
Gois 476.00 94-05 9.86
Cear 669.29 92-98 9.25
So Paulo 3177.14 95-01 8.81
Par 464.00 94-00 7.76
Rio G. do N. 191.90 92-01 7.08
Piau 187.50 95-00 6.70
Rio G. do S. 568.00 94-01 5.68
Santa Catarina 285.00 00-01 5.47
Paran 279.50 99-02 2.98
Minas Gerais 476.67 99-01 2.74
Maranho 86.00 98-99 1.56
Note: Police homicide data based on raw MNDH scoring.




36



Note: Figures are regression coefficients with unbalanced data. Numbers in parentheses are panel-corrected
standard errors. Coefficients in bold are statistically significant: * p<.1; ** p<.05; *** p<.001.
Table 3 Estimates of Effective Number of Parties (ENP) in Gubernatorial
Elections, 1982-2002




Variable Model 1 Model 2 Model 3 Model 4
Share of seats of rival party .010* -.001 -.004
(.005) (.004) (.014)
Share of seats of rival coalition -.007 -.004 .018*
(.005) (.005) (.009)
ENP of seats in legislature .085** .060** .050**
(.036) (.026) (.021)
PFL/PDS -.524*** -.471*** -.486*** .920***
(.134) (.089) (.106) (.143)
PT -.390** -.654*** -.612*** .247
(.163) (.15) (.135) (.414)
PMDB -.258** -.258** .553***
(.085) (.097) (.086)
PSDB -.157* -.119 .621**
(.087) (.105) (.302)
Run-offs
(t-1)
.003
(.147)
Re-election -.397*** -.391*** -.381**
(.106) (.099) (.111)
GDP share 4.570*** 4.147*** 2.951*** -7.902***
(.898) (1.028) (.770) (.953)
Growth in GDP share -.169 -.210
(.227) (.540)
lnGrowth .026 -1.402*
(.132) (.783)
Government size .103 .490 -2.333
(.585) (.423) (1.442)
Fiscal transfers -.267 .410** .933
(.445) (.172) (.980)
Constant 2.260*** 2.747*** 2.756*** 6.073
(.638) (.251) (.302) (3.862)

N 80 108 106 106
R
2
.43 .43 .35 .29
Wald X
2
71.75 24989.92 1651.37 46492.79
Prob > X
2
0.000 0.000 0.000 0.000
37


Table 4 Estimates of Police Homicides per Hundred Thousand
Population, 19 Brazilian States, 1994-2002

Variable Model 1 Model 2
Share of seats of rival party -0.102
(0.153)
Share of seats of rival coalition -0.133 -0.156
(0.110) (0.111)
PFL/PDS -3.666** -4.300**
(1.342) (1.955)
PT -2.863 -2.831
(6.293) (4.455)
PMDB -3.435 -2.965*
(2.173) (1.701)
PSDB -2.307 -1.888
(3.726) (3.347)
Run-offs
(t-1)
7.172** 6.856***
(3.261) (1.921)
Re-election 0.251
(3.746)
GDP share -145.05** -143.95**
(72.267) (59.886)
Growth in GDP share 5.374** 4.794**
(2.157) (1.928)
lnGrowth -5.449 -3.070
(14.248) (10.409)
Government size -24.023 -13.846
(70.499) (21.951)
Fiscal transfers 6.439
(26.521)
Corrected homicide rate
(t-1)
0.003* 0.003*
(0.002) (0.001)
Constant 49.511 37.421
(71.848) (51.431)

N 40 41
R
2
.30 .28
Wald X
2
5.21 6.67
Prob > X
2

0.157

0.083



Note: Figures are regression coefficients with unbalanced data. Numbers in parentheses are panel-corrected
standard errors. Coefficients in bold are statistically significant: * p<.1; ** p<.05; *** p<.001.


38
Works Cited

Ahnen, Ronald. 2007. The Politics of Police Violence in Democratic Brazil. Latin
American Politics and Society 49:1 (Spring): 141-64.
Beer, Caroline and Neil J. Mitchell. 2006. Comparing Nations and States: Human Rights
and Democracy in India. Comparative Political Studies 39:8 (October): 996-
1018.
Case, William. 2006. Manipulative Skills: How Do Rulers Control the Electoral
Arena? In Electoral Authoritarianism: The Dynamics of Unfree Competition, ed.
Andreas Schedler. Boulder, CO: Lynne Rienner.
Chavez, Rebecca Bill. 2003. The Construction of the Rule of Law in Argentina: A Tale
of Two Provinces. Comparative Politics 35:4: 417-37.
Cornelius, Wayne. 1999. Subnational Politics and Democratization: Tensions between
Center and Periphery in the Mexican Political System. In Subnational Politics
and Democratization in Mexico, eds. Wayne A. Cornelius, Todd A. Eisenstadt,
and Jane Hindley. La Jolla, CA: Center of US-Mexican Studies, University of
California, San Diego.
Cornelius, Wayne A., Todd A. Eisenstadt, and Jane Hindley, eds. 1999. Subnational
Politics and Democratization in Mexico. La Jolla, CA: Center of US-Mexican
Studies, University of California, San Diego.
Dahl, Robert A. 1971. Polyarchy: Participation and Opposition. New Haven: Yale
University Press.
Diamond, Larry. 1999. Developing Democracy: Toward Consolidation. Baltimore: Johns
Hopkins University Press.
39
Falleti, Tulia. 2005. A Sequential Theory of Decentralization: Latin American Cases in
Comparative Perspective. American Political Science Review 99 (August): 327-
46.
Fox, Jonathan. 1994. The Difficult Transition from Clientelism to Citizenship: Lessons
from Mexico. World Politics 46:2 (January): 151-84.
Franco, Jennifer. 2001. Elections and Democratization in the Philippines. New York:
Routledge.
Gervasoni, Carlos. 2006. A Rentier Theory of Subnational Authoritarian Enclaves: The
Politically Regressive Effects of Progressive Federal Revenue Redistribution.
Paper presented at the meeting of the American Political Science Association,
Philadelphia PA, August 31.
Gibson, Edward. 2005. Boundary Control: Subnational Authoritarianism in Democratic
Countries. World Politics 58 (October): 101-32.
______. 2004. Subnational Authoritarianism: Territorial Strategies of Political Control
in Democratic Regimes. Paper presented at the 2004 meeting of the American
Political Science Association September 2-5.
Hartlyn, Jonathan and Jennifer McCoy. 2006. Observer Paradoxes: How to Assess
Electoral Manipulation. In Electoral Authoritarianism: The Dynamics of Unfree
Competition, ed. Andreas Schedler. Boulder, CO: Lynne Rienner.
Hecock, R. Douglas. 2006. Electoral Competition, Globalization, and Subnational
Education Spending in Mexico, 1999-2004. American Journal of Political
Science 50:4 (October): 950-61.
40
Heller, Patrick. 2000. Degrees of Democracy: Some Comparative Lessons from India.
World Politics 52:4 (July): 486-501.
Howard, Marc Morj and Philip G. Roessler. 2006. Liberalizing Electoral Outcomes in
Competitive Authoritarian Regimes. American Journal of Political Science 50:2
(April): 365-81.
Huntington, Samuel P. 1968. Political Order in Changing Societies. New Haven: Yale
University Press.
Laakso, Marku and Rein Taagepera. 1979. Effective Number of Parties: A Measure
With Application to West Europe. Comparative Political Studies 12: 3-27.
Lacerda, Alan Daniel Freire de. 2002. O PT e a Unidade Partidria como Problema.
Dados 45:1: 39-76.
Levitsky, Steven and Lucan A. Way. 2002. The Rise of Competitive Authoritarianism.
Journal of Democracy 13:2 (April): 51-65.
Mainwaring, Scott P. 1999. Rethinking Party Systems in the Third Wave of
Democratization: The Case of Brazil. Stanford: Stanford University Press.
McMann, Kelly M. 2006. Economic Autonomy and Democracy: Hybrid Regimes in
Russia and Kyrgyzstan. New York: Cambridge University Press.
McMann, Kelly M. and Nikolai V. Petrov. 2000. A Survey of Democracy in Russias
Regions. Post-Soviet Geography and Economics 41:3: 155-82.
Nicolau, Jairo Marconi, ed. 1998. Dados Eleitorais do Brasil (1982-1996). Rio de
Janeiro: IUPERJ/UCAM.
41
______. 2005. Dados Eleitorais do Brasil (1982-2004). Electronic version:
http://jaironicolau.iuperj.br/Site/dados%20eleitorais%20do%20Brasil%201982-
2004.html
Nunes Leal, Victor. 1976. Coronelismo, Enxada e Voto: o Municpio e o Regime
Representativo no Brasil. So Paulo: Editora Alfa-Omega.
ODonnell, Guillermo. 1993. On the State, Democratization and Some Conceptual
Problems: A Latin American View with Glances at Some Postcommunist
Countries. World Development 21: 1355-69.
Pereira, Anthony W. 2000. An Ugly Democracy? State Violence and the Rule of Law in
Postauthoritarian Brazil. In Democratic Brazil: Actors, Institutions and
Processes. Peter R. Kingstone and Timothy J. Power, eds. Pittsburgh: University
of Pittsburgh Press.
Przeworski, Adam. 1991. Democracy and the Market: Political and Economic Reform in
Eastern Europe and Latin America. New York: Cambridge University Press.
Reis, Fbio Wanderley and Mnica Mata Machado de Castro. 1992. Regies, Classe e
Ideologia no Processo Eleitoral Brasileiro. Lua Nova 26: 81-131.
Samuels, David J. 2000. The Gubernatorial Coattails Effect: Federalism and
Congressional Elections in Brazil. The Journal of Politics 62:1 (February): 240-
53.
______. 2003. Ambition, Federalism, and Legislative Politics in Brazil. New York:
Cambridge University Press.
Schattschneider, E.E. 1960. Semisovereign People: A Realists View of Democracy in
America. New York: Holt, Rinehart and Winston.
42
Schedler, Andreas. 2006. The Logic of Electoral Authoritarianism. In Electoral
Authoritarianism: The Dynamics of Unfree Competition, ed. Andreas Schedler.
Boulder, CO: Lynne Rienner.
Schmupeter, Joseph. 1942. Capitalism, Socialism, and Democracy. New York: Harper
and Row.
Schwartzman, Simon. 1982. Bases do Autoritarismo Brasileiro. Rio de Janeiro: Campus.
Sidel, John. 1999. Capital, Coercion, and Crime: Bossism in the Philippines. Stanford:
Stanford University Press.
Snyder, Richard. 2001. Politics After Neoliberalism: Reregulation in Mexico. New York:
Cambridge University Press.
Snyder, Richard and David J. Samuels. 2004. Legislative Malapportionment in Latin
America: Historical and Comparative Perspectives. In Federalism and
Democracy in Latin America, ed. Edward Gibson. Baltimore: Johns Hopkins
University Press.
Stepan, Alfred. 2000. Brazils Decentralized Federalism: Bringing Government Closer
to the Citizens? Daedelus 129 (Spring): 145-69.
Van de Walle, Nicolas. 2006. Tipping Games: When Do Opposition Parties Coalesce.
In Electoral Authoritarianism: The Dynamics of Unfree Competition, ed. Andreas
Schedler. Boulder, CO: Lynne Rienner.
Wantchekon, Leonard and Tamar Asadurian. 2002. Transfer Dependence and Regional
Disparities: The Case of Nigeria. Working Paper No. 152 Center for Research on
Economic Development and Policy Reform, Stanford University.
43
Way, Lucan A. 2006. Authoritarian Failure: How Does State Weakness Strengthen
Electoral Competition? In Electoral Authoritarianism: The Dynamics of Unfree
Competition, ed. Andreas Schedler. Boulder, CO: Lynne Rienner.
______. 2004. The Sources and Dynamics of Competitive Authoritarianism in Ukraine.
Journal of Communist Studies and Transition Politics 20:1 (March): 143-61.
Willis, Eliza, Christopher Garman, and Stephan Haggard. 1999. The Politics of
Decentralization in Latin America. Latin American Research Review 34:1: 7-56.

44
Notes


1
Conceivably, the proliferation of subnational authoritarian regimes could produce the collapse of national
democracy, though the author is unaware of any cases of democratic breakdown that fits this
characterization.
2
This distinction differentiates subnational authoritarianism as a problem of the quality of democracy from
electoral authoritarianism as a regime type. Observers of the latter phenomenon restrict their definition
exclusively to indicators of access rather than the nondemocratic exercise of power. See Schedler 2006 (5-
6).
3
We calculated the effective number of parties using the conventional Laakso and Taagepera (1979)
equation. This formula is represented as N = 1/ x
2
i
where x
i
is the percentage of seats held by the i-th
party. Following this formula, if the distribution of seats favors two parties in similar proportions, with a
third minor party taking the remainder of seats, N will be some number between 2.0 and 3.0.
4
Although re-election for governor were not legalized until 1998, several governors in some cases returned
for nonconsecutive second or more terms during the pre-98 period.
5
An independent-sample t-test confirmed that the difference in means was significant at the .001 level (t=-
4.983).
6
Neither first-round or second-round margin standard deviations are significantly different between the two
cohorts, suggesting that margin degrees are distinct but not their dispersion. Both cohorts are consistently
competitive.
7
Tests of differences in means found statistical significance at the .1 level.
8
Rondnia is excluded for insufficient police homicide data.
9
We calculated the rate of change in the value-added index for each state and ran the growth rate as a
control. However, the transformed index proved insignificant in every test. We report the coefficient for the
natural log of the index and not the growth rate in our models.
10
The number of observations increased from 80 to 108 between the models because all of the pre-1990
cases were excluded due to the run-off variable in the first model. Run-offs became legal only after the
1990 contests.
45

11
We also tested the ENP variables in multiple specifications but found these indicators of pluralism to be
insignificant in all tests.

You might also like