ResearchPDF Available

The relevance of socioeconomic factors over votes dispersion

Research

The relevance of socioeconomic factors over votes dispersion

Abstract and Figures

The purpose is to identify the relevance of socioeconomic factors over the dispersion of votes for Brazilian Congress.
Content may be subject to copyright.
1
Are Brazilian Electoral Districts Actually Built? The Relevance of Socio-economic Conditions
on the Dispersion of Votes for Brazilian Congress
Glauco Peres da Silva
1
Graziele Silotto
2
Abstract
How are the votes for the Legislative spread out across Brazil’s territory? Ames (1995a, 1995b,
2001) failed to answer this question when providing the concept of informal districts. Since then, it is
assumed that politicians are free to exercise constituency service or to provide pork to any part of
their district. Since the Brazilian electoral system is based on an open list proportional representation
system in districts of large magnitude, scholars state that politicians have strong incentives to create a
personal links with the electorate. Moreover, it is implicit that the only individual constraints faced
are the other existent links, created by other politicians, with the electorate. However, recent data has
shown that informal districts are related to socio-economic conditions and that all politicians are
subjected to them. Adequate measurements for the dispersion of votes, such as the Horizontal
Cluster, provide evidence that a hierarchy of cities based on commuting, transportation flow, press
coverage, flow of commerce, and other socio-economic activities, highly constrain political activity
in Brazilian electoral districts. The hierarchy of municipalities provides a layer in which political
activity takes place, and which political scientists must take into consideration within their analyses.
Key-words: Personal Vote, Concentration of votes, hierarchy of municipalities, congressmen.
1
This work was funded by the Center for Metropolitan Studies (CEM/CEBRAP, USP), process nº 2013/07616-7 and the
Brazilian Center for Analysis and Planning, process2011/08536-1, of the São Paulo Research Foundation (FAPESP).
The opinions, hypotheses, and conclusions or recommendations expressed do not necessarily reflect the views of
FAPESP.
2
The authors would like to acknowledge and thank the database organization done by Bruno Hirsch and Bianca Flório
and thank all the participants in the electoral studies group organized at the NECI – FFLCH/USP.
2
I. Introduction
The specialized literature dealing with the concentration of votes for politicians in Brazil’s
legislative elections presumes, at least implicitly, the existence of zones where politicians control the
local political disputes. This somewhat commonplace interpretation gains further resonance with
works grounded on the idea of ‘electoral connection’ as a summarizer for the Brazilian political
system, which attributes the country’s institutional combination as the causes for such a phenomenon
(e.g., Mainwaring, 1991; Armijo et al., 2006). Since then, it is understood that politicians are the
ones to seek the formation of these “redutos”
3
and, from this constitution, organize the electoral
competition within the country.
However, the use of new indicators for measuring the concentration of votes for candidates
attempting a federal congressional representative seat a prime example of such phenomenon
reveals that there is a pattern in the dispersion of these votes towards different candidates in different
election years. Individuals with political careers within the same cities display the same kind of
voting dispersion across the district. This unexpected behavior by this traditional literature suggests
that there is something within the dynamics established between the politician and the voter that is
common to them, which results in the observed voting dispersion. The question arises: what does
explain the patterns of dispersion of votes in Legislative elections?
This observation not only introduces a new way to comprehend the pattern of electoral
competition in Brazil, but it also brings new elements to evaluate the effectiveness of incentives
coming from the electoral rule. This rule would be responsible for creating a favorable environment
for the personal vote, and consequently, the formation of electoral constituencies. Thus, not only is it
observed that these incentives are not enough to create a common pattern throughout the country, but
it would also bring to the discussion hitherto neglected elements: the introduction of socio-economic
relationships within political interactions. The analysis intended here presents a dimension that has
been undervalued by a tradition grounded on institutionalism when dealing with electoral behavior in
Brazil. By considering formal institutions as the focus for understanding the workings of the political
system, a particular positioning of the so-called rational choice institutionalism, such studies have
given a secondary role to components that would be associated with informal institutions, such as the
socioeconomic environment in which formal institutions operate. The adoption of important
elements of the Political Geography, notably Agnew (1995), allows a deeper analysis of vote
concentration. Thus, the contribution of this research is quite significant insofar that it recovers
3
Ames (1995a) defines it as “electoral fortress, where a deputy dominates a group of contiguous municipalities” (p. 410)
3
undervalued aspects in face of the theoretical framework on which the analytical literature regularly
rests. The relationship between politician and voter would be theoretically expanded beyond
clientelism.
From this initial effort we intend to investigate whether the voting patterns of congressional
representatives from a same area are similar, stemming from the hypothesis that the network of
relationships established among Brazilian cities, as observed through the hierarchy of cities (IBGE,
2008)
4
established by the commercial trades between its residents, the media outlets read, the daily
work pendulum flow, etc., form an environment in which the stream of information flows among
people in such a way as to influence the obtainment of votes by a particular candidate. It evaluates if
by knowing the city in which the politician establishes their career it becomes possible to determine
in which cities their votes will be concentrated, from a network of relationships across the cities of
São Paulo. The descriptive findings in this study suggest a strong relationship between these
dimensions. Besides this introduction, the paper is organized into five other sections. The next
section presents the research literature and the problem to be investigated and then presents the
methodological procedures, the descriptive results, and discusses an evaluation model of the regions
regarding vote concentration. A last section with final considerations closes the paper.
II. Presentation
The literature shows that the traditional interpretation of the Brazilian political system consists
of incentives for political behavior based on electoral rules and, to this end, goes back to an
arrangement established by the 1988 Federal Constitution
5
. By directly referring the electoral rules, it
states that its permissiveness, whose origins comes from the open-list proportional representation,
would lead to high degree of intra-party competition, thus creating incentives for the individualist
behavior of politicians and, consequently, weakening parties within the country (Carey e Shugart,
1995; Samuels, 2002). Its consequences could be seen in parliamentarians whose behavior lacks
cohesion and discipline, guided by personal voting (Carey and Shugart, 1995) and, more importantly
for the present work, geared towards localist strategies over national discussions, thus reducing
parties to “patronage machines" (Samuels, 2002).
4
The IBGE, or the Brazilian Institute of Geography and Statistics, is a public institution, created in 1934, in order to
collect and provide information data about Brazil through researches and analysis.
5
Such an arrangement between a strong federalism, with electoral rules favoring individual behavior from politicians,
combined with a centralized Executive Power, would generate a democracy doomed to failure. See, for example, Ames
(2003) Lamounier (1992) and Mainwaring (1991).
4
This diagnosis derives from another dimension in the relationship between the elected and the
voter: the electoral connection. By articulating the electoral and legislative arenas, politicians would
make use of budgetary resources to allocate funds to their spatially defined electoral bases or
constituencies, so that these benefits may be converted into future votes. These constituencies would
be located in clearly defined areas within a territory dominated by politicians, in which they could
avoid the competition, much like electoral strongholds where they would surely be able to win voters
by offering them pork-barrel policies or even doing constituency service. Thus, politicians whom
would solely be interested in maintaining their offices (or would be, according to Mayhew (1974)
office seekers) would be able to reach their objectives by offering clientelist policies to the voter,
who would repay this behavior by offering votes.
According to this argument, the action destined to each politician’s specific constituency would
explain the dispersion of votes for legislative offices in Brazil. Each candidate’s votes would be
geographically located where their constituencies reside, therefore places where the candidate would
allocate attention and efforts in exchange of votes. The sending of pork and the intermediate work
alongside the electorate would explain both the actions of parliamentarians in Congress as well as
their relationship with voters, notably clientelistic. In the interest of this research, this explanatory
chain attributes to the individual willingness of politicians the regionalization observed in their votes.
That is to say, regional vote dispersion patterns over a given territory would be created from
cultivating ties between politicians and voters.
Ames (1995a, 1995b, 2003) is one of the leading authors to consider the prevalence of electoral
connection and clientelistic personal vote as being typical to the Brazilian political system.
According to the author, although candidates can seek out for votes in all over the state in an
election, many of them prefer to focus their campaigns in geographically defined regions for several
reasons, including the high costs of travelling. It is important to say that Brazil adopts a proportional
rule with open list for legislative elections. Each state is a electoral district. The magnitude of each
district ranges from 8 (less populated) to 70, the case of São Paulo, the most populated state. These
components, all together, explain the high cost of reaching the electorate and then, also explain the
concentration of political efforts in the region within the district. These regions would form the so-
called informal districts. The establishment of these sites would be the realization of electoral
incentives, the visible face of the argument that “thus, Brazilian politics favors the provision of local,
geographically separable benefits” (Ames, 2001:79). The regionalization of votes would initially
depend on electoral incentives and decisions from politicians to cultivate these relationships.
However, recent researches have questioned the several results derived from election rule
incentives, such as the capacity for clientelist practices via pork in the electoral outcome for the
5
national legislative (Pereira and Renno, 2007); the standardization of the congressional
representative’s individual political actions geared towards clientelism (Carvalho, 2003; Nicholas,
2006); and even difficulties from politicians themselves in finding their own electoral constituency
(Santos, 2003). In a very recent paper, Lago and Rotta (2014) also found no effects of the
implementation of budget amendments over reelection rates of congressional representatives elected
by states in south of the country, reinforcing Pereira and Rennó (2007) findings. The observation of
incentives coming from the electoral system has not thus been able to account for the election results
for the national legislative.
Coming from a different approach, this study intends to follow the idea that the vote, as Agnew
(1987) states, is the result of a geographically conditioned process in a manifold dimension.
Elections for the national legislative, in an open list proportional representation system of high
magnitude, are subjected to dynamics that are both national and local. A particular candidate is not
exempt from the influence of the party’s political positioning, for example if its stance is pro-
government or oppositional, but also subjected to the sub-district context of the dispute over votes, in
addition to the ties established with their constituency, not necessarily clientelistic. From this
perspective, analyses regarding votes for a congressional representative office in Brazil need to
combine both dimensions.
In addition to merely individual categories, whether from voters, whether from politicians, the
political geography approach seeks to combine interaction processes that occurs in a more structural,
ontological explanation sphere with others that take place at a local level. To this effect, it primarily
uses the concept of context, where “(t)he context refers to the geographical scope of specific
influences, the limits thus set to practical reason, and the distinctive meanings or discourse associated
by the human agent with living with these influences” (Agnew, 1996: 130). This conception contains
multiple interacting factors for justifying electoral results, where incentives from distributional
policies make for a specific dimension. Agnew (1996) poses this point as follows:
“(C)ontext refers to the hierarchical (and non-hierarchical) ‘funnelingof stimuli across geographical
scales or levels to produce effects on politics and political behavior. These effects can be thought of as
coming together in places where micro (localized) and macro (wide-ranging) processes of social
structuration are jointly mediated. As a result, politics can be mapped not simply as the geographical
outcome of non-spatial processes of political choice, but as a spatialized process of political influence
and choice.” (p.132)
6
Although quite broad, this perspective is interesting while dealing with problems faced when
analyzing the dynamics within Brazilian politics. The approach stemming from a political geography
allows for another perspective insofar that it considers that:
“the hierarchical-geographical context or place channels the flow of interests, influence and identity out
of which political activities emanate. This approach assumes, therefore, that political behavior is inevitably
structured by a changing configuration of social-geographical influences as global-local connections shift over
time.” (p.132).
According to this same perspective, Johnston (2001) presents an important argument for the
discussion presented in this paper. Within the more specific debate on the definition of electoral
geography, the author argues that:
“(i)ts main concerns are with: (…) the processes by which territorially defined electoral constituencies
are defined within different electoral systems; the geographies of representation which result from the
pattern of votes across constituencies; and the geography of political power—of ‘who gets what, where’
because of politically- and electorally-influenced decisions on the location of public goods and
services.” (p.4374).
This observation allows to disassociate events understood as being bound together in the
Brazilian case. The process of forming constituencies is understood as the same process in the voting
distribution pattern, and also identical to the location of public goods and services. Insofar as
clientelist practices strengthens a link between politician and voter, which is rewarded with votes,
these three dimensions would stem from the same process, more notably the incentives stemming
from the electoral rule. The theoretical separation at different levels allows the debate to overcome
difficulties from the empirical findings. In this sense, Agnew (1996) shows the importance of context
for voting decisions summarized in six arguments. One of these arguments deserves special attention.
The author says: “(t)he microgeography of everyday settings home, work, leisure, etc. – can
stimulate local distinctiveness that is reflected in voting patterns” (p.4376)
Johnston thus complements this idea:
“Most voting decisions result from interactions, many indirect, between political parties and electors,
which include local interactions that influence election outcomes. (…) Many candidates performed best
in their home areas, creating a ‘friends and neighbors effect.’ (…): voting decisions are influenced by
spatially biased information flows (…) which generate not only support for local candidates but also
7
more general ‘neighborhood effects’ whereby an area’s majority opinion is accentuated if those initially
favoring minority views are won over through a process of ‘conversion by conversation.’” (p.4375).
Even political party campaigns are geographically conditioned, obtaining different results
according to the context in which they occur. Both rely on the biased information flow. The electoral
impact of these elements occurs due to the restricted information flow that affects different voters in
different ways; that is, depending on the context. Pattie and Johnston (2000), concerned with the
importance of the everyday dynamics in electoral results, state that:
“Electorally relevant information (political cues) flows through social networks and stimulates
responses in the form of partisan decisions: if the information reaching an individual through her/his
conversations predominantly favors one party, then he/she is more likely to vote for that party,
irrespective of prior predispositions, than if the information was less biased in that particular direction”
(p.42).
It is necessary here to underline that the influence of the geographical aspect appears in this
argument line in a rather more sophisticated manner than within the context of parochialism.
In studies on elections in Brazil, the most straightforward reference to this geographical
dimension in politics was given by Sonia Terron, for whom even when dealing with presidential
elections it is necessary to notice that “political, economic, and social factors interact in different
geographical scales and can determine significant differences in voting behavior" (2009:12). In this
sense, there is a connection between the voters’ political preferences and social context in such a way
that it becomes necessary to provide further attention to interactions between citizens who live in the
same areas, the impact of national and local economy in the government’s popularity, and
geographically-oriented election campaigns.
Studies dealing with such voting socioeconomic factors can be traced back to Key (1949). In a
study on voters in Southern USA, the author showed that voters decision were based on what he
named friends-and-neighbors effect”, which suggests that voting choices are prone to the
candidate’s place of origin or place of residence. Therefore, voters would prefer local candidates,
which is justified by calculating the promotion of local interests by the electorate, even if this same
electorate is not organized towards the defense of a common interest (Key, 1949:37). The
electorate’s decision over a certain candidate would be based on valuing their region within the
national parliament and not necessarily grounded on a clientelistic relationship. At any rate, the
electorate cannot be considered to be independent from the environment they live and it would be
8
responsible, via localism, for providing the electoral nucleus to concede votes to a local candidate or
to a candidate from contiguous cities (p.132).
According to Terron (2009:30), candidates would depend on this friends-and-neighbors effect
to secure their electoral base. As for the voters, their voting decision would be influenced by the
context closest to their daily lives, which is independent from specific candidates or political parties,
but would be involved in the exchange of information with residents from neighboring towns. This
exchange process is influenced by access to radio and local television networks covering the regions
in which they reside, sales of printed newspapers, as well as the use they make of transportation
available between cities (in view of the proximity and the ease of access), which in turn eases access
to commerce in nearby towns and daily trips to work, school, etc., since “the biased information
flows are likely to be spatially concentrated” (Pattie and Johnston, 2000:42).
Thus, the incorporation of geographical space and the social and economic relationships
underlying studies on voting behavior and the concentration of votes may reveal processes that
would otherwise remain invisible and contribute to better estimate the effects of factors that may also
determine voting choices. The idea is not to indicate a new explanatory variety for voting choices.
The voter may still be seen as a rational voter who opts for the programmatic position closest to their
ideal point, in a Downsian sense. The most relevant factor here is in how the information regarding
politicians generally reaches voters before they make their decision. It is argued that the voter is
highly influenced by their daily lives, bounded by their daily circulation space, available sources of
information, the cities visited, etc., at the time when deciding on a vote. The information flow is not
homogeneous within the territory and tends, therefore, to favor different candidates in different
regions. In short, the information is biased.
Electoral geography may therefore contribute to studies on electoral competition, as well as to
comprehend the relationship established between the Legislative power and the electorate, and how
they behave, considering the spatial dimension in order to show, according to Terron (2012):
“[the] establishment of electoral territories and connections between political actors and
their electoral bases; for the politician it may provide relevant information for campaign
strategies; and for the citizen, in may become a surveillance tool for representatives of
'their territory' "(p.17).
This theoretical construction is in accordance with common empirical observations on the
dynamics of national politics, notably during election periods. The candidates and news press items
on different candidatures for the legislative reference the districts’ sub-regions. In the case of São
Paulo, for example, there is a large number of news stories in which we find a relationship between
9
the candidate for the National congress and a specific region of the territory
6
. That is a political
argument, and it is employed to convince voters that a particular region must elect their
representative. Thus, it becomes important to understand the underlying socioeconomic dynamics to
the formation of areas where votes are concentrated. It is within this context that the study intends to
investigate these aspects, now in a more descriptive manner by evaluating whether the voting
patterns for members of Congress in São Paulo are suited to the IBGE city hierarchy network. This
network is a proxy of the socio-economic interactions that take place along the territory.
III. Methodological Discussion
The argument on which this work draws from stems from the principle that such current
literature dealing with the regionalization of votes does not illuminate important factors when
considering the importance of the territory. Basically, one should consider that the population is not
evenly dispersed throughout the territory. This observation, discussed in Avelino et al. (2011), for the
creation of indicators on the spatial concentration of votes creates an environment to evaluate how
politicians strategically deal with this aspect. If the voting of a particular candidate is concentrated in
a region where the total population is low, the candidate is not expected to be elected. This would not
be an appropriate strategy for any politician. Such a scenario implies that in each politician’s region
of concentration there exists at least large enough cities to ensure electoral success. On the other
hand, there are not enough cities in the state of São Paulo, for the case in point, for politicians to be
able to avoid competition amongst themselves. However, it is much more interesting to presume that
the socioeconomic dynamics underlying population displacements condition politicians to deal with
another dimension of political interference, beyond their personal dispositions and interactions with
other politicians. A campaign in a certain city
7
which is also a regional center could reverberate in
several other cities due to the socioeconomic dynamics involving municipalities.
In this case, the problematic is reinforced with the help of graphics. The following maps
illustrate the main characteristics of the problem with which this research aims to work. The maps
6
Some of the following links provide a sample of the politicians concern with the state’s regional aspect in the dispute
for a congressional representative seat in São Paulo for the 2014 election:
http://www.jj.com.br/noticias-6715-bigardi-comenta-a-importancia-eleger-representantes-da-regiao-;
http://portalprudentino.com.br/noticia/noticias.php?id=38514&titulo=bragato-e-ed-sao-os-unicos-deputados-eleitos-de-
pp; http://www.atribuna.com.br/elei%C3%A7%C3%B5es-2014/baixada-santista-elege-seis-deputados-para-s%C3%A3o-
paulo-e-bras%C3%ADlia-1.407846; http://www.revide.com.br/gerais/ribeirao-elege-tres-estaduais-e-dois-federais/;
among others.
7
City and municipality are used interchangeably in this paper regarding the size of the population.
10
refer to different candidates in different elections: orange for the 2002 election, green for the 2006
election, and yellow for the 2010 election.
Figure 1 – Maps of votes dispersion measured by Location Quotient for selected candidates in São
Paulo between 2002 and 2010.
The first two maps refer to candidates who have been mayors in cities near Santos, in the
south coast of São Paulo. The two following maps present candidates with votes concentrated in the
region of Presidente Prudente, western border of the state. The last ones refer to candidates with an
area of influence in Vale do Paraiba, east São Paulo border. What is interesting to note is that the
maps, which present data on the Location Quotient
8
of these candidates in each of the cities in the
8
See Silva and Davidian (2013) for a technical presentation in adapting this typical Regional Economy indicator in
evaluating the spatial distribution of votes. See also Benavid-Val (1991) for more detailed discussion about the index.
11
respective elections, show very similar concentration areas
9
. This fact would be a coincidence if one
considers the traditional literature. Moreover, a fact that draws attention is that while candidates have
trajectories dependent on a certain city regional center, such as Santos, São José dos Campos or
Presidente Prudente, their areas of electoral concentration expand throughout the territory in the
same geographical direction.
If we take the case of Santos under a more detailed analysis it would be possible to presume
that if three candidates had their political origin in this city, each could follow a different path: one
could expand their area of influence to the south of the state, another could incorporate cities
between Santos and São Paulo (MRSP metropolitan region of São Paulo), and the third could
expand their votes to the north coast, for example. However, as reflected by the maps above, this is
not observed. They all show the same expansion pattern towards south and along the coastline, which
is also true for other candidates in the same region. This is also true when observing the maps and
considering politicians with a trajectory in other cities in the São Paulo coast such as São Vicente or
Praia Grande. They all show similar maps. This finding recurs throughout the state for different
candidates in different elections. Politicians from certain regions have nearly identical dispersion
maps.
The construction of the maps refers to the municipalities hierarchy established by the IBGE
whose aim is to present the relationships network between Brazilian municipalities based on
commercial relations, population flow, etc. In the words of the publishers:
"In the 2007 update, the subject of this publication, we sought to define the hierarchy of urban centers
and delimit the regions of influence associated with them, from aspects related to federal and corporate
management and the provision of equipment and services, in order to identify the territorial points from
which decisions are issued and the command exercised in a network of municipalities. To this end, we
used specific research data and, secondarily, data from other surveys also carried out by the IBGE, as
well as records from public agencies and private companies "(IBGE, 2008).
Note that the main concern was to establish a connection between cities and thus outline the
network of relationships between them. As an example, by exploring the state of São Paulo, two
formative dimensions are presented for this network for the year 2007: the flow of transportation and
newspapers. With them it is possible to display maps with networks formed in each dimension and
relate them to voting concentration maps. These representations are shown in Figure 2.
9
Circa 30% of elected members presented clearly defined voting dispersion maps as those presented in the analyzed
period. All maps can be provided upon request to the authors.
12
Figure 2 – Comparison between the transportation flows (upper left), newspapers coverage (upper right) and the maps of dispersion of votes.
Transportation flows
Newspapers coverage
13
In each one of these maps one may see, whether by the transportation or newspaper coverage, a
certain spatial pattern that regionalizes the state of Sao Paulo in fifteen different areas that are
associated with this network. Even more striking, the voting concentration maps used as example suit
some of these areas, such as Baixada Santista and Vale do Paraíba, as highlighted in the maps in
Figure 2. In each one of them there is a pole-city, Santos and São José dos Campos, around which
the transportation and communication network is established. According to this construction, the
areas under the influence of each of these pole-cities would have a particular dynamic and it would
explain the observation of voting concentration areas as shown on the maps. In this context, the areas
for obtaining votes by a given candidate would be subject to the establishment of such networks. It is
worth noting that the formation of this network of relationships between cities is a national issue, and
it is not limited to the political divisions of the states, since they are of a different nature. In this
study we shall only consider the networks internal to the electoral districts, since what is being
analyzed is the voting concentration in the elections for Federal Congress representative as from
1994.
The IBGE sorts the cities in 5 subdivided levels. These ordinances are established from their
self-sufficiency in the provision of goods, services, and public facilities for its inhabitants. Level 1
cities are divided into 1A, 1B, and 1C. Level 2 cities are divided into levels 2A, 2B, and 2C; level 3
cities are subdivided into 3A and 3B, just as the level 4 cities. This leveling is done until they reach
level 5, which has no subdivisions. The municipalities are thus allocated in eleven categories, where
the lower the numerical value assigned, the greater its importance to cities that are hierarchically
below the municipality.
In other words, the city influence network hierarchizes municipalities by pondering the
importance of larger municipalities for smaller ones, whose inhabitants depend on the structure of the
first, whether for access to higher education services, trade, or access to health and culture. Thus, São
Paulo, the better structured city and with the most services in the country, is the only one at level 1A,
the highest possible level. Rio de Janeiro and Brasilia are just below, at level 1B. Smaller
municipalities with little structure are in the lower levels of the hierarchy, as is the case of Adolfo in
the state of São Paulo, with 3631 inhabitants, at municipality level 5.
The operationalization of the areas that divide the state of São Paulo was done in the following
manner: the city of São Paulo is the city that influences all others. In order to identify the sub-areas,
the criterion established was to accompany each city along the hierarchy until the last level before
São Paulo, i.e., traversing a path from the lowest levels to the highest level prior to 1A. The last
municipality in the hierarchy before São Paulo was considered to be the pole-municipality. If there
were no cities among the lower levels and Sao Paulo, then only in such cases would these cities have
14
São Paulo as reference. Therefore, the pole-cities, with their respective hierarchical levels, are:
Araraquara (2C), Bauru (2C), Botucatu (3A), Campinas (2A), Jundiaí (4A), Marilia (2C), Ourinhos
(3A), Presidente Prudente (2C), Ribeirão Preto (2B), the Metropolitan Region of São Paulo (1A),
Santos (2C), São Carlos (3A), São José do Rio Preto (2B), São José dos Campos (2C), and Sorocaba
(2C)
10
.
So, the methodological evaluation from here on lies with the construction of indicators to allow
for a spatial assessment of each candidate’s voting results and the comparison of this data with
information obtained from the Regic-IBGE (2007), in order to determine the correspondence
between them. There has been progress, even if preliminary, within the discussions of vote
concentration strategy with the proposals of new concentration indexes and spatial statistical
calculations to make use of electoral geography. We may highlight in this case the work of Avelino
et al. (2011), in which the authors present a measurement based on Regional Economics, the G
index, in order to assess the overall concentration of votes by a candidate for Federal Congress in
Brazil. Another work, by Silva and Davidian (2013), complements the previous one by discussing
indexes that allow the observation of regional concentrations across the district. One of such indexes
is the LQ index, or location quotient, originally proposed by Benavid-Val (1991: 75)
11
. Its use allows
us to understand the relative importance of each city in the candidate's voting. When the LQ equals 1
it means that the candidate received the expected amount of votes in that municipality if the
distribution of votes was spatially homogeneous, given the total number of votes received by the
candidate and the total amount of votes in the city. The other index is the HC, or Horizontal Cluster,
adapted from Fingleton et al. (2005). This indicator, derived from LQ, shows the difference between
the amount of votes actually received by a certain candidate in any given municipality and the
number of votes necessary for the LQ to be equal to 1
12
. These indicators allow for the creation of
maps, such as the ones in Figure 1, highlighting specific concentration areas for each candidate. Due
to it being expressed in votes, in this research, we shall prioritize the use of the HC for the
descriptive data, and use the G index for control. The HC index redistributes votes throughout the
cities in such a way as to show the positive or negative balance of votes that each candidate obtained
in relation to what would be obtained if the distribution of votes across the territory were
10
For the correspondence of each municipality into regions, check http://goo.gl/X4qUwz.
11
The LQ is adapted for electoral data by the following formula: 

=

, being

the total votes of party i in city
c,
=

,
=

e =

.
12
HC would be equal to



, that is, when the LQ is equal to 1, we obtain

=
, which allows us to rewrite it
as 

=


1
.
15
homogeneous. Thus, its average value is zero by definition. Its calculation allows the redistribution
of votes among cities, considering the existence of surplus or lack of votes in relation to a
homogeneous distribution. In this case, the homogeneous distribution occurs by comparison with the
size of the electorate in each city. Therefore, the HC has no high or low limits, with the positive sign
indicating excess votes and the inferior sign a lack of votes.
These indexes allow us to circumvent difficulties that may conceal the candidate’s real voting
concentration in municipalities. By knowing the areas of concentration it becomes possible to
understand the strategies adopted by the candidates as well as to better understand the choices of
voters in the process of choosing their candidates, relating it to the socioeconomic, political, and
spatial context. Thus, studies on the Brazilian electoral geography have produced an important
research agenda that aims to better understand the arguments that restrict the clientelistic relationship
between politician and voter. Even if they are still incipient, these studies attempt to consider the
importance of spatial aspects. With this information in hand, we proceed to the evaluation of data.
IV. Descriptive Results
The first result of interest is the distribution of candidates and congressional representatives
elected among the different regions of the state. Table 1 below presents this information.
Table 1 – Regional distribution of candidates, congressmen and population
The table shows that the locations contain a similar number of candidates. The average
relationship between population and candidates in each region reaches its maximum value for the
1998 2002 2006 2010
Region Candidates Elected Population Candidates Elected Population Candidates Elected Population Candidates Elected Population
Araraquara 9 1 475.205 6 1 507.930 8 1 537.015 6 1 564.373
Bauru 16 2 905.417 15 1 968.988 19 1 1.049.894 17 0 1.067.978
Botucatu 6 1 411.893 6 1 449.595 5 1 479.668 12 1 501.080
Campinas 63 6 4.220.066 94 7 4.634.534 102 7 5.085.759 106 8 5.356.156
Jundi 7 1 575.724 7 1 667.528 11 0 728.640 8 1 788.202
Marília 12 0 632.573 10 0 675.602 13 3 706.708 15 1 714.680
Ourinhos 2 0 190.474 0 0 202.942 1 0 211.368 3 0 219.277
Presidente Prudente 11 1 776.783 8 1 807.163 9 1 831.240 6 0 851.258
Ribeirão Preto 19 2 1.904.970 19 1 2.050.741 31 3 2.191.432 40 2 2.320.385
Santos 31 2 1.553.840 24 2 1.747.135 40 2 1.878.790 41 3 1.937.702
São Carlos 6 0 300.050 5 1 328.506 3 1 353.312 4 1 371.408
São Jo do Rio Preto 24 4 1.961.453 30 2 2.071.875 29 2 2.190.422 33 4 2.281.178
São Jo dos Campos 35 2 1.853.761 32 2 2.065.084 40 4 2.233.498 46 3 2.351.676
MRSP 311 42 16.496.140 374 42 17.769.576 555 40 19.084.436 613 41 19.533.584
Sorocaba 29 2 1.860.761 22 3 2.085.204 41 3 2.265.387 42 2 2.403.261
Total* 581 66 652 65 907 69 992 68
* The total of Elected column is not equal to 70 because it is not possible to define the especific region of some congressmen
16
region of Ourinhos, with 126,566 inhabitants per candidate in average, while the lower value is
obtained in the Metropolitan Region of São Paulo (MRSP), with 41,702 inhabitants per candidate.
The overall average is 71,601 inhabitants per candidate from all regions, with a standard deviation of
28,173 inhabitants per candidate. The median of this relationship is 66,240 inhabitants per candidate.
This suggests that the distribution of candidates among the different regions is relatively
homogeneous, with the largest number of candidates being associated with the regions with the
largest population, while regions with less population present fewer candidates. It should also be
noted that the larger number of elected candidates is in the region with the highest population. While,
on average, 61.6% of those elected are in the Metropolitan Region of São Paulo, almost 48% of the
population lives in this region. The correlation of this relationship between the regions is 99%,
reduced to 92% if the metropolitan area is disregarded in the comparison. That is, the observation of
these regions seems to attribute an effective regional approach. With this in mind, we shall pass on to
evaluate the results. The first results are in the following table.
Table 2 – Average HC per region for all candidates (1998 – 2010)
The data show, as expected, that the averages of the regions are generally very close to zero.
The only exception is the São Paulo Metropolitan Region, where the p-value for the test against the
null average hypothesis being equal to zero is 9.87%. Still, the general average is equal zero and,
given the wide variability in the observed data, the regions do not show significant differences
between them.
Region Mean Standard Deviation N
Araraquara 0,20 394,27 64.116
Bauru 0,31 312,77 138.918
Botucatu 0,27 172,84 92.612
Campinas 0,43 642,84 274.274
Jundi -0,15 833,22 24.934
Marília 0,26 207,17 138.918
Ourinhos 0,46 197,55 39.182
Presidente Prudente 0,34 204,40 192.348
Ribeirão Preto 0,48 436,61 220.844
Santos 2,29 841,15 85.488
São Carlos 0,66 701,12 24.934
São José do Rio Preto 0,30 224,11 523.614
São José dos Campos 1,21 795,96 153.166
RMSP -8,57 2.379,80 128.232
Sorocaba 0,69 466,81 195.910
Total 0,004 716,54 2.297.490
17
At this moment it is interesting to notice the behavior of the averages between the regions for
groups of congressional representatives who made their political careers in different regions. In this
sense, congressional representatives were separated according to the reference city
13
and the averages
for each region were calculated. There are two scenarios for each region: congressional
representatives whose careers are in the regional pole-city and those whose political career were
established in other cities within the same region. It is expected that congressional representatives
shall not have exactly the same voting pattern among themselves at a municipal level, but high
correlation at a regional level. Moreover, it is expected that the voting for these candidates is also
highly correlated in patterns established in other regions. This means that any given region, even if
neighboring another, which does not have a constant and facilitated flow between them would result
in poor correlation of votes between them. If the proposition hereby defended is correct and the
network between cities in each of the regions is not established, the votes for a candidate from one of
these cities will not flow towards the other. This would hold true for all candidates in each of the
regions. A larger effort would be necessary, which only a few of the candidates would be able to
manage, to overcome this type of barrier and compete for votes throughout the entire territory.
Moreover, it is expected that the regions from which these congressmen originate have
superior HC averages to the other regions, except for cases where congressmen have a widely
dispersed voting. In this case, they must have votes beyond the limits established by the area of
influence to other regions with greater connections with that region. Table 3 below shows the data
for regions in which individuals were elected at least once in their careers.
Table 3 – Average HC for congressmen grouped by original region
* significant at 10%; ** significant at 5%; *** significant at 1% § None congressmen came from this region
Note: Values with no statistical significance were omitted.
13
The criterion used to determine the region of the politician at this time was the verification of the city in which he run
for city council or mayor. In the model tested further on, as it shall be explained, this criterion has been expanded.
Região 1 2 3 4 5 6
§
10
11
12
13
14
15
Araraquara 1 3,800*** 122*** 513***
Bauru 2 365*** 484*** 78***
Botucatu 3 1,586*** 55***
Campinas 4 211*** 694***
Jundi 5 10,344***
Marília 6 25* 909*** 152***
Ourinhos 7 628*** 257*** 264***
Presidente Prudente 8 14*** 24** 1,580***
Ribeirão Preto 9 199*** 1,273***
Santos 10 4,110***
São Carlos 11 9,626***
São José do Rio Preto 12 438***
São José dos Campos 13 2,050***
RMSP 14 897***
Sorocaba 15 310*** 1,253***
18
Table 3 shows that averages are much higher in regions where congressional representatives
built their careers. In Araraquara, the average HC in the region is almost 2,800 votes; in Marilia, it
reaches almost 1,000 votes. A particularly striking fact is that in some other regions the values also
show positive figures for the HC a fact that occurs with all congressional representatives from this
region. This happens in regions neighboring the region where the congress member belongs to, such
as Bauru and Ourinhos. As for Presidente Prudente, further west region of the state, there is a rather
high voting in its own region, and in some other neighboring regions such as Ourinhos and Marilia.
In São Carlos, the average values are the highest among the five areas considered. The same happens
in Sorocaba: values in the region are quite high, but we also find positive values in other regions in
the state. The same holds true for Bauru, Santos, and Campinas and this pattern can be seen in any
one of the fifteen regions.
Table 4 – Average HC for candidates of Bauru region per city of origin
The observations of the results for each individual region provide another important piece of
information. Since the voting pattern for candidates is influenced by a component that is common to
all of them, the political career of two individuals built in different cities, but within the same region,
should show similar regional voting patterns. To put it another way: if we take into account two
random politicians that built their careers in different cities, but within a same region, we would
expect the voting distribution to be similar across these regions. The votes for each individual should
be located in the region of influence of the same pole-city, without overflowing these limits. For this
reason and with this in mind, we now assess the results shown in the above table for some of the
Region Bauru Jaú
Araraquara -4,81 6,75
Bauru 250,33*** 361,83***
Botucatu -2,15 0,51
Campinas 5,99 -25,52
Jundi -34,87 -42,71
Marília -1,76 -4,94
Ourinhos -1,05 -2,69
Presidente Prudente -3,58 -2,66
Ribeio Preto -10,08 -10,30
Santos -21,52 -30,75
São Carlos -13,72 -11,39
São Jo do Rio Preto -4,94** -5,91
São Jo dos Campos -17,47 -20,92
MRSP -172,37*** -212,06
Sorocaba -11,66 -13,72
Total 0,00 0,00
19
analyzed regions. Table 4 shows the data for the Bauru region. There are two cities in the region in
which congressmen built their political career: Bauru and Jaú.
The first highlight is that values are systematically higher in the Bauru region, being the only
region in which the averages are statistically different from zero. Furthermore, the magnitude of
these averages is very high (815 and 1058, respectively). Furthermore, what is even more striking is
that the correlation between the regional averages HC is 0.9993. If we observe the values by
municipality, the correlation becomes negative, very close to zero (-0.04). This finding suggests that
although there are differences in the values observed per city, the pattern between regions is basically
the same.
The very same result can be seen for Santos, as shown in table 5 below. There are two
congressional representatives for Santos, one in São Vicente and another in Praia Grande. Regional
correlations are also extremely high even if we consider the RMSP. The lowest of these regional
correlations is 0.996, whereas if we consider the values at the municipal level, the maximum
observed correlation is 0.45.
Table 5 – Average HC for candidates of Santos region per city of origin
The last analyzed region is Campinas. The pattern repeats itself, with one significant
exception. The congressional representative elected for Limeira presents different voting from the
others. This is justified by the high voting dispersion by this politician. The calculated G index for
Region Santos São Vicente Praia Grande
Araraquara -18,75 -24,70 -19,64
Bauru -16,33** -21,77* -16,77
Botucatu -11,34** -14,24 -11,85
Campinas -39,64*** -53,04*** -36,40
Jundi -62,27 -88,31 -59,88
Marília -10,99** -13,97 -11,43
Ourinhos -11,25 -15,28 -12,27
Presidente Prudente -9,56** -12,62* -9,53
Ribeio Preto -17,72** -28,40* -21,77
Santos 908,04*** 1232,82*** 726,05***
São Carlos -33,22 -42,45 -31,45
São Jo do Rio Preto -9,17*** -12,40** -9,62
São Jo dos Campos -24,62 -33,75 2,60
MRSP -307,84*** -412,19*** -210,92
Sorocaba -20,79** -30,06* -24,07
Total 0,00 0,00 0,00
20
the micro-regional level is 0.047, whereas for the Mogi Mirim congressional representative it
presents a much more concentrated value of 0.634. This high dispersion explains the different pattern
and lower correlations. These, in fact, would be even smaller (circa 0.4) if we exclude the MRSP.
This is the only case in which this exclusion alters the correlations at the regional level not only for
the region of Campinas, but for all the previous regional analysis.
Table 6 – Average HC for candidates of Campinas region per city of origin
In this sense, one can note that the assessment of the general voting dispersion is a key factor
when understanding the phenomenon in question. It becomes a crucial control, among others, such as
the region’s population size, which allows us to reveal the pattern in a candidates’ voting dispersion.
For this reason, we stem from a model that allows the introduction of such controls.
V. Regression
The proposed model considers the dependent variable as the difference between the HC
standardized averages for candidate i in region r in the year a, and the HC overall average obtained
for all candidates that year in that region. This will be named variable Z, which is calculated as
follows:

=






(1)
Region Campinas Mogi Mirim Piracicaba Limeira
Araraquara -9,52 -10,81 -22,28 -13,91
Bauru -7,18 -9,41 -1,05 -6,61
Botucatu -4,85 -5,85 -1,09 -5,00
Campinas 166,10*** 151,66*** 263,37*** 167,78***
Jundi 3,76 -33,25 -31,07 -27,61
Marília -5,34* -6,23 -7,18 -4,10
Ourinhos -7,28 -4,16 6,64 -8,09
Presidente Prudente -4,91* -5,57 -1,93 -5,04
Ribeio Preto -7,35 -12,78 -22,80* -7,79
Santos -35,37** -25,14 -57,98 -36,20
São Carlos -10,31 -17,13 -37,17 -21,40
São José do Rio Preto -3,93** -5,35 -3,08 -7,16*
São José dos Campos -11,30 -17,35 -38,38 -13,85
MRSP -238,99*** -178,78 -355,40*** -217,47**
Sorocaba -11,60** -11,91 -23,92* -10,64
Total 0,00 0,00 0,00 0,00
21
wherein Z is the rate observed for individual i in region r in the year a;
is the average HC for the
same individual i in the same region r; S is the observed standard deviation and n is the number of
observations. The objective of this procedure is to correct the overall variance for each group by
standardizing the data between different regions, which allows a comparison between different
regions and different years, also facilitating the interpretation of the estimated parameters. The graph
below shows the estimates for the probability density functions of variable Z for the 1998 and 2010
elections. The average for all years are around -2.6 and -2.7 and standard deviations are also circa
2.1. The four curves present the same general behavior pattern for changes in Z. Only in slightly
smaller values than -5 and about -2.5 do we find a misalignment of these functions. But the pattern
between them is quite similar.
Graph 1 – Probability Density Function Estimation for Z variable per year
Elaborated by authors
As for the elected, the pattern remains the same, but softer and somewhat more symmetrical.
There is a peak just above -5 and another one around zero. The average for this group is similar to
the full set of candidates, around -2.35, but the standard deviation is higher, around 2.3. The average
value of the Z variable estimated in both groups suggests difficulties from candidates to obtain voting
results higher than the average obtained by all candidates in each region, including those who were
0 .1 .2 .3
Densidade
-10 -5 0 5
Z
1994
1998
2002
2006
2010
kernel = epanechnikov, bandwidth = 0.5063
Eleitos
0 .1 .2 .3
Densidade
-10 -5 0 5
Z
1998
2002
2006
2010
kernel = epanechnikov, bandwidth = 0.2915
Todos os candidatos
Density
Density
Elected
All candidates
22
elected. This implies that within the set of fifteen regions dividing the state of São Paulo, the
candidates for congressional representative do not obtain votes across the entire state, which is in line
with the findings of Avelino et al. (2011). If there is a degree of concentration of votes in a regional
sphere larger than the municipality, it does not affect the state as a whole. For the purposes of this
paper, the analysis turns towards verifying the concentration of these areas within the zones of
influence established by the hierarchy of cities.
Therefore, the question is if knowledge of a candidate’s reference area alters the value of Z.
In order to evaluate this connection, the relationship of the Z value shall be calculated with the
following explanatory variables: 1) a dummy that indicates the region for each candidate’s reference
state. This is the most relevant explanatory variable, since by knowing the candidate’s place of origin
we expect the difference between votes obtained by this candidate and the average of other
candidates in that area to be positive. The association between region and candidate is done in two
ways: the candidate's voter registration site and the city in which the politician has already run for
mayor or city council. These two variables were combined into a third and more general variable: the
area of a candidate is defined when those two areas are equal or when only one of them is available.
If there is a divergence, it is left blank; 2) the total number of votes obtained by candidates in each
region per year. It is expected that the greater the number of votes, on average, the more disperse
they will be in the territory and therefore the lower the value of Z; 3) the G index for assessing the
general level of voting dispersion per state. The more dispersed the votes, the lower the value of Z;
and lastly, 4) the population for each region within the state
14
. It is expected that regions with the
highest population will have a greater value for Z, since the applicant could reach a larger number of
voters in a smaller geographic area, not having, therefore, to campaign in other areas of the territory.
There are other control variables, such as the number of previous elections for Congress the
candidate has partaken and if the candidate has ever been elected for mayor or city council in a
municipality within the state of São Paulo. The data refers to all candidates whose candidatures were
deferred by the TSE
15
. The first table below shows the estimated results
16
.
14
For the construction of this population variable in election years, the following correspondence was made: for the 1994
election, we used the 1991 census data; for the 1998 election, data from the 1996 population count; for the 2002 election,
2000 Census data; for the 2006 election, the 2007 count; and for the 2010 election, we used the 2010 Census.
15
For the year of 1994, we worked only with data on those elected, since information regarding place of affiliation of the
candidates not elected and their voter’s registration, used to establish the connection between candidates across elections,
are not complete. For the 1998 elections, from 681 candidates we have 659 deferred applications; for 2002, from 793
applications, 703 are the deferred; as for in 2006, from a total of 1098 candidatures, 952 are deferred, and in 2010, from
1276 applications, 1030 are deferred.
16
It is important to inform that for cases in which individuals were candidates once again during the study period, we
only considered the case presented with the highest G index, indicating the greater de-concentration of votes.
23
Table 7 – Estimated Parameters for the Regression on Z
(1) (2) (3) (4) (5) (6) (7) (8)
z z z z z z z z
Dummy_region 4,634 4,646 4,714 5,376 4,150 4,706
(0,0246)*** (0,0232)*** (0,0414)*** (0,0569)*** (0,0515)*** (0,0424)***
Dummy_region_filiation 4,589
(0,0262)***
Dummy_region_candidate 4,370
(0,0292)***
G -1,554 -1,549 -2,028 -1,441 -1,781
(0,0357)*** (0,0404)*** (0,0793)*** (0,0487)*** (0,0435)***
Dummy_elected_deputy 0,086 -0,026 -0,048 0,014 -0,032
(0,0478)* (0,0522) (0,0601) (0,0861) (0,0517)
Dummy_cand_mun -0,080 -0,028 -0,177 -0,104
(0,0189)*** -0,025 (0,0294)*** (0,0189)***
Dummy_elected_mun -0,078 -0,234 -0,027 -0,077
(0,0114)*** (0,0161)*** (0,0154)* (0,0112)***
Population (ln) 0,092 0,158 0,043 0,036
(0,0276)*** (0,0567)*** (0,0315) (0,0278)
Total of votes in the region 8,74E-06 4,47E-06 2,74E-05 8,78E-06
(2,62e-06)*** (1,50e-06)*** (4,02e-06)*** (2,62e-06)***
Dummy_Araraquara 0,417
(0,1040)***
Dummy_Bauru 0,551
(0,0677)***
Dummy_Botucatu 0,409
(0,1033)***
Dummy_Campinas 0,385
(0,0300)***
Dummy_Jundi 0,157
(0,0828)*
Dummy_Marília 0,475
(0,0804)***
Dummy_Ourinhos 0,352
(0,1973)*
Dummy_Presidente Prudente 0,254
(0,0966)***
Dummy_Ribeirão Preto 0,356
(0,0502)***
Dummy_Santos 0,001
(0,0443)
Dummy_São Carlos 0,303
(0,1320)**
Dummy_São José do Rio Preto 0,343
(0,0474)***
Dummy_São José dos Campos 0,109
(0,0406)***
Dummy_Sorocaba 0,241
(0,0463)***
Constant -3,033 -3,021 -2,919 -2,614 -1,495 0,650 -2,836 -1,533
(0,0090)*** (0,0091)*** (0,0096)*** (0,0142)*** (0,1564)*** (0,2238)*** (0,2106)*** (0,1546)***
Observations 41.415 41.415 41.415 41.415 41.415 19.605 21.810 41.415
Adjusted R2 0,29 0,28 0,17 0,32 0,32 0,37 0,29 0,33
Robust Standard Errors in parenthesis
* significant at 10%; ** significant at 5%; *** significant at 1%
24
In the first three regressions, only the information regarding the candidate’s region was
considered. The first model considers the region synthesizing the candidature location and the
affiliation of the candidates, the second considers the region from the candidate's affiliation locale,
and the third considers the region for the candidacy for city council or mayor. The three variables
show to be statistically significant and, more importantly, an estimated value for a very significant
parameter. A beta estimated around, respectively, 4.63, 4.59, and 4.37, shows the high impact over
the Z variable. Since both the estimated value is higher in the first case and the value of the adjusted
R
2
is slightly higher, we decided to maintain this broader definition of the candidate’s region in the
other models.
In the fourth regression, two other explanatory variables are included: the G index, calculated
for the municipal level
17
, and a dummy indicating whether the politician was elected in that election.
What confirms the previously described results is that the value for the estimated parameter for this
variable is very low, 0.086, indicating that the observed result does not vary between elected and
non-elected. With the inclusion of other controls in other models, the estimated value decreases even
further, no longer distinguishing from zero. As for the estimated parameter for the G variable, it
presents a negative sign, as expected, and significant for all considered models. The value for the
estimated beta is around -1.5, quite high when compared to the other parameters. The following three
models are similar, with only one basic difference: in the models 6 and 7 candidates whose cities of
reference are, respectively, the pole-city and non-pole cities are separated for the analysis. Model
number 5 lacks this separation. The signs of the parameters and their magnitudes do not vary. The
estimated parameter in model 5 for the reg variable is quite close to the others (4.7), but increases
significantly in the model with individuals from the pole-city (5.38). In addition, it is noteworthy that
the model with only the pole-cities presents a slightly higher adjusted R
2
(0.37 against 0.32 of the
complete model). Lastly, the only variable with variation in its significance level is the dummy
variable, which indicates whether the individual has been a candidate for city council or mayor in
model 6. While the magnitude of the estimated parameters is very low for all cases, this occurrence
suggests that reference in the pole-city is sufficient to account for the varying results. As for the last
estimated model, dummies are included for 14 regions of the state, minus the metropolitan area. With
the exception of Santos, the statistical significance is high in all of them and with a positive sign
parameter. This result suggests that in the Metropolitan Region of São Paulo the difference in the
17
The G variables calculated for the micro and mesoregional levels were tested in the models. The results do not change.
For the different levels at which this index can be calculated, see Avelino et al. (2011).
25
candidates’ voting in the region is lower than the average obtained in this region when compared
with other areas of the state.
The results observed may be broken down in the comparison between candidates who have
the pole-city as their municipality of reference and those whose reference city is another city in the
same region. This comparison helps with the understanding of the dynamics not only between
regions, but also between cities within the same regions, and to better understand the differences
observed in the model above. The graph below shows the estimated beta parameter for the reg
variable for the previously presented fifth regression model. The controls are, therefore, the same.
Graph 2 – Estimated Confidence Intervals of Z in regions by pole-city and no pole-cities.
Elaborated by authors.
The average of the estimated parameters is around 4. With the exception of Marília,
Presidente Prudente, and São José do Rio Preto, in all other regions the estimated parameters are
above 3, close to the observed values for all the regions in our previously presented models.
Furthermore, the differences between the parameters for the pole-city and other cities from the same
region are almost the same, except for Jundiaí, Ourinhos, and the Metropolitan Region of São Paulo.
This result is quite interesting to show that the internal dynamics of the regions are equivalent,
regardless of the city in which the career is built.
26
Another way of studying the same phenomenon is to evaluate the average results of the
candidates in the regions neighboring their region of reference. If the dynamics occurs within each
region, we would expect to see votes concentrated in the reference regions and few votes in the
neighboring towns. Thus, we considered the same previous model and a dummy variable was
introduced to indicate the neighboring region
18
to the one of reference. The results for this dummy
variable are shown in the following graph, keeping the comparison between individuals with careers
in pole-cities and non-pole cities.
Graph 3 – Estimated Confidence Intervals of Z in referred neighbor regions by pole-city and no pole-
cities.
Elaborated by authors.
What can be observed in the graph above is that the parameter values are on average much
lower than in the previous graph. While the previous parameter had an average above 4, in this case
the average is around 1. This means that the voting difference for a candidate and the average for all
other candidates in the reference region is greater than the difference in the neighboring regions.
Moreover, in this case, the differences between the pole-city and others are not too prominent. With
the exception of São Carlos, in all other cases the differences between the groups are within the
18
Neighboring was established by the existence of borders between regions. Thus, a region is neighbor to another if there
is a border between them.
27
limits of the calculated confidence intervals. This result only strengthens the argument that regional
dynamics appear to be an important aspect for explaining the spatial distribution of votes.
VI. Final Considerations
The data presented above, yet descriptive, indicates the existence of a regional voting pattern
in São Paulo that respects the hierarchy of the cities. The votes thus appear to be subjected to a social
and economic dynamics that interferes in the election outcome. Context matters for explaining voting
dispersion. Whether due to the flow of information or the ease for credit claiming, or even by the
different socializations by which voters are subjected: the relationship between cities seems to
impose a layer with new dynamic that explains the voting dispersions observed from the
transmission of biased information.
These findings make way for an investigation thus far neglected in studies on the Brazilian
electoral system. The attention is far too focused on the incentives imposed by electoral rules on the
political system as a whole. However, the grounding theoretical construction reached an extreme
level of attention towards incentives emanating from formal rules. Understanding the incentives does
not mean to say that all individuals subjected to such incentives will behave equally. There are other
institutions, especially informal institutions, which are basically neglected by the institutional
literature. Moreover, context - as defined here - in which these incentives are built, are also not
evaluated. The basic premise is that formal electoral rules would be enough to comprehend the
national political system.
The introduction of a new approach from a political geography standpoint is promising. In the
first incursion carried out in this work, we note that the knowledge of the political history of an
individual’s region, in composition to the hierarchy of the state’s municipalities, provides enough
information to understand the dispersion of votes throughout the territory. However, further research
is necessary to advance the theoretical developments provided by this approach.
Bibliographical References
AGNEW, John A. (1987), Place and Politics: The Geographical Mediation of State and Society.
London: Allen and Unwin.
28
______. (1996), “Mapping politics: how context counts in electoral geography”. Political
Geography, vol. 15, nº 2, pp. 129-146.
AMES, Barry. (1995a), “Electoral Strategy under Open-List Proportional Representation”. American
Journal of Political Science, vol. 39, n. 2, pp. 406-33.
______. (1995b), “Electoral Rules, Constituency Pressures, and Pork Barrel: Bases of Voting in
Brazilian Congress”. The Journal of Politics, vol. 57, n. 2, pp: 324-43.
______. (2001), The Deadlock of Democracy in Brazil, The University of Michigan Press, University
of Michigan.
______. (2003), Os Entraves da Democracia no Brasil, Fundação Getúlio Vargas: Rio de Janeiro.
ARMIJO, Leslie, FAUCHER, Phillippe and DEMBINSKA, Magdalena. (2006), “Compared to
what? Assessing Brazil’s Political Institutions”. Comparative Political Studies, vol. 39, nº6, pp.
759-786.
AVELINO, George, BIDERMAN, Ciro and SILVA, Glauco P. (2011), “A concentração eleitoral nas
eleições paulistas: medidas e aplicações”. Dados, vol. 54, nº 1, pp.319-347.
BENAVID-VAL, Avrom. (1991), “Economy Composition Analysis”, in Regional and Local
Economic Analysis for Practioners, New York, Praeger, pp. 67-76.
CAREY, John M. e SHUGART, Matthew S. (1995), “Incentives to Cultivate a Personal Vote: a
Rank Ordering of Electoral Formulas”. Electoral Studies, vol. 14, nº 2, pp. 417-439.
CARVALHO, Nelson R. (2003), E no Início Eram as Bases. Geografia Política do Voto e
Comportamento Legislativo no Brasil. Rio de Janeiro, Ed. Revan.
FINGLETON, Bernard, IGLIORI, Danilo e MOORE, Barry. (2005), “Cluster Dynamics: New
Evidence and Projections for Computing Services in Great Britain”. Journal of Regional
Science, vol. 45, no. 2, pp. 283–311.
IBGE, (2008). Região de Influência das Cidades de 2007. Instituto Brasileiro de Geografia e
Estatística, Rio de Janeiro.
JOHNSTON, Ron. (2001), “Electoral Geography”, in N. Smelser e P. Baltes (orgs.), International
Encyclopedia of the Social & Behavioral Sciences, pp. 4374-4378.
KEY, Valdimer O. (1949), Southern Politics in State and Nation, New York: Alfred A. Knopf.
LAGO, Ivann C. e ROTTA, Edemar. (2014), “Conexão eleitoral e reeleição entre deputados federais
do sul do Brasil / 1998-2010”. Revista de Sociologia e Política, vol. 22, nº 49, pp. 139-156.
LAMOUNIER, Bolívar. (1992) Estrutura Institucional e governabilidade na década de 1990, in: J. P.
dos REIS VELLOSO (org.), O Brasil e as reformas políticas. Rio de Janeiro: José Olympio.
MAINWARING, Scott. (1991), “Politicians, Parties, and Electoral Systems: Brazil in Comparative
Perspective”. Comparative Politics, vol. 24, n. 1, pp.21-43.
29
MAYHEW, David (1974). Congress: The Electoral Connection. New Haven, United States: Yale
University Press.
MELO, Carlos R. (2005), “As instituições políticas brasileiras funcionam?”. Rev. Sociol. Polít.,
Curitiba, 25, pp. 199-203.
NICOLAU, Jairo. (2006), “O Sistema de Lista Aberta no Brasil”. Dados, vol. 49, no. 4, pp. 689-720.
PATTIE, Charles e JOHNSTON, Ron. (2000), “’People who talk together vote together’: An
exploration of contextual effects in Great Britain”. Annals of the Association of American
Geographers, vol. 90, nº 1, pp. 41-60.
PEREIRA, Carlos e RENNÓ, Lucio.
(2007), “O Que É que o Reeleito Tem? O Retorno: o Esboço de
uma Teoria da Reeleição no Brasil”. Revista de Economia Política, vol. 27, nº 4, pp. 664-683.
SAMUELS, David. (2002), “Pork Barreling Is Not Credit Claiming or Advertising: Campaign
Finance and the Sources of Personal Vote in Brazil”. The Journal of Politics, vol. 64, 3, pp.
845-863.
SANTOS, Fabiano. (2003), O Poder Legislativo no Presidencialismo de Coalizão. Belo Horizonte:
Editora UFMG; Rio de Janeiro: IUPERJ.
SILVA, Glauco P. e DAVIDIAN, Andreza. (2013), Identification of Areas of Vote Concentration:
Evidences from Brazil”, Brazilian Political Science Review, vol. 7, no. 2, pp. 141-156.
TERRON, Sonia L. A Composição de Territórios Eleitorais no Brasil: Uma Análise das Votações de
Lula (1989-2006). Tese de Doutorado. IUPERJ, Rio de Janeiro, 2009.
____________. (2012), “Geografia Eleitoral em Foco”. Em Debate, Belo Horizonte, vol.4, nº.2,
pp.8-18.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
In spite of the recent progress in the discussion on vote regional concentration brought by Avelino et al. (2011), there is still a lack of determination of the districts' internal areas where candidates obtained their votes. Synthetic concentration indices, as the G index, do not allow for evaluation in disaggregated levels, as municipalities, which would be relevant for the verification of the areas of a candidate's political influence. This paper aims at bridging this gap through the joint utilization of two other indicators, the Location Quotient (LQ) and the Horizontal Cluster (HC), that use different measurement units. These indices were applied to the elections of federal representatives in Brazil from 1994 to 2010, and the results were compared to those obtained by Avelino et al. (2011). The application reaches its aimed result, clearly showing the places where the candidates' votes are located in the electoral districts.
Article
Full-text available
The article deals with the electoral connection between members of the three southern states of Brazil between 1998 and 2010, covering four elections and three legislatures. The main objective of the study was to identify and analyze the associations between the profile of the parliamentary vote and his performance in Congress and between this and their chances for reelection in the subsequent election. Therefore, we use two data sets: the origin of the votes of deputies from the southern states of Brazil between 1998 and 2010 elections, and the action of these MPs, especially from the individual budget amendments. Using the model of the "reason of chance", through the index of the Electoral Impact of Amendments, we demonstrated that the allocation of resources through the budget amendments are not able to increase the chances of the deputy of obtaining more votes between the electors of the cities that received them. We concluded that the transfer of amendments cannot be taken as being among the most effective strategies to ensure reelection. Therefore, the theoretical models that take the transfer of resources must be reviewed, especially through the individual budget amendments, as being among the most effective strategies to ensure the reelection of Brazilian congressmen.
Article
Full-text available
This paper looks at the political consequences and the political origins of the Brazilian electoral system. This system has several unusual features that grant politicians nonpareil autonomy with respect to their parties. These features include a system of proportional representation that uses an open list and a mechanism known as the candidato nato, which allows politicians to get on the ticket despite the opposition of the party leadership. As a result, the electoral system reinforces the individualistic behavior of politicians and has contributed to undermining efforts to build more effective parties. Notwithstanding their frequent laments about the weakness of parties, Brazilian politicians have consistently opted for electoral systems that undermine parties. They have done so because they perceived measures that could strengthen parties as authoritarian, and also in response to their fears that executives would otherwise be able to control them ruthlessly. The extreme party weakness and individualistic patterns of representation that are reinforced by this electoral system have sustained an elitist polity.
Article
Full-text available
This article analyzes the open-list electoral system in Brazil in the elections for the Chamber of Deputies (Lower House) and compares it to other countries that employ this mechanism. The author describes the history and functioning of the prevailing electoral system; evaluates the list's impact on parties, voters, and the relationship between Deputies and their constituencies; and discusses the mechanisms offered by the electoral system for voters to either punish or reward their representatives. Linking the Brazilian case to current analyses of electoral systems, the article draws on evidence from three studies conducted with members of the Brazilian Congress and voters.
Article
Full-text available
A rich and plausible academic literature has delineated reasons to believe Brazil’s democratic political institutions—including electoral rules, the political party system, federalism, and the rules of legislative procedure—are suboptimal from the viewpoints of democratic representativeness and policy-making effectiveness. The authors concur that specific peculiarities of Brazilian political institutions likely complicate the process of solving societal collective action dilemmas. Nonetheless, Brazil’s economic and social track record since redemocratization in the mid-1980s has been reasonably good in comparative regional perspective. Perhaps Brazil’s informal political negotiating mechanisms, or even other less obvious institutional structures, provide sufficient countervailing influences to allow “governance” to proceed relatively smoothly despite the appearance of chaos and political dysfunction.
Article
Full-text available
This article compares the determinants of electoral success in two consecutive Brazilian legislative elections, 1998 and 2002. There is a clear difference between both periods that renders the comparison especially interesting. In 1998 the incumbent president was running for reelection whereas in 2002 it was an open seat contest. We hypothesize that in 1998 the proximity of the Federal Deputy with the president and the allocation of federal monies controlled by the Executive Branch played a more significant role in affecting reelection success than in 2002. Hence, if the President is himself running for reelection is an important intervening contextual variable in understanding reelection success of Federal Deputies.
Article
This paper develops a theory of candidate strategy from social-choice principles and from the workings of open-list proportional representation. The theory is used to explain the campaign behavior and the spatial patterns of vote distribution for candidates to the Brazilian Chamber of Deputies. Campaign strategy is evaluated with models that predict where deputies will offer budget amendments to benefit target municipalities whose votes they seek in subsequent elections. The choice of target municipalities is a function of the cost of erecting barriers to entry, the dominance of the deputy in the municipality, the spatial concentration of the deputy's statewide vote, the vulnerability of the municipality to invasion by outsiders, the weakness of the deputy in the last election, and the deputy's prior political career. Logistic regression of amendments to the Brazilian national budget in 1989 and 1990 and OLS regression of municipal-level electoral results for the Brazilian Chamber of Deputies in the 1990 elections. Deputies seek secure bailiwicks, search for vulnerable municipalities, and strive to overcome their own electoral weakness by delivering pork. Candidates' tactics vary, partly because political backgrounds differ and partly because the differing demographic and economic contexts of Brazilian states reward some tactics and penalize others. Candidate behavior hinders voter control over deputies, increases incentives for pork seeking, and weakens party programs and discipline.