Ballot Reform as Suffrage Restriction: Evidence from
Brazil’s Second Republic*
Daniel W. Gingerich, University of Virginia
PO Box 400787
Charlottesville, VA 22904-4787
434-243-4923
[email protected]
September 20, 2016
Few innovations in democratic institutional design are considered as fundamental as the introduction
of voting through the use of a uniform, official, and secret ballot. One account claims that the official
ballot liberates dependent voters from the dictates of local elites, thereby enhancing democratic competition. Another argues that in contexts of widespread illiteracy its adoption may be tantamount to
a suffrage restriction. This paper adjudicates between these views by drawing upon an original dataset
of municipal-level voting returns from Brazil’s Second Republic (1945-1964). The unique staggered
rollout of the official ballot during this period permits one to assess its impact with unprecedented accuracy. The paper finds that the primary consequence of ballot reform was suffrage restriction. Rather
than liberating poor and dependent voters, the official ballot made it exceedingly difficult to have their
votes counted. Moreover, archival sources indicate that this was an anticipated and intended effect of
the reform.
*This material is based upon work supported by the National Science Foundation under Grant No. 1119908. The author
also gratefully acknowledges the support of the Bankard Fund for Political Economy (University of Virginia). Thanks go to David
Samuels for kindly sharing electoral returns for federal deputy for the state of São Paulo and for helpful comments on an earlier draft
of this paper. In addition, the author would like to thank Michael Albertus, Barry Ames, Kate Baldwin, Samuel Berlinski, William
Roberts Clark, Alberto Dı́az-Cayeros, David V. Fleischer, Anna Gryzmala-Busse, Daniel Hidalgo, Jeff Jenkins, Karen Jusko, David
Leblang, Staffan I. Lindberg, Fabiana Machado, Shaun McGirr, Sid Milkis, Scott Morgenstern, Kevin Morrison, Silvia A. Otero,
Jonathan Rodden, Miguel Rueda, Carlos Scartascini, Kenneth Scheve, Todd Sechser, Alberto Simpser, David Stasavage, Ernesto
Stein, and Eric Wibbels for their comments on this work, as well as seminar participants at Duke University, the Inter-American
Development Bank, the Juan March Institute, Stanford University, the University of Chicago, the University of Illinois, UrbanaChampaign, the University of Michigan, Ann Arbor, the University of Pittsburgh, the University of São Paulo, and the University
of Virginia.
I.
Introduction
The ballot is the fundamental instrument of democracy in the modern world. Perhaps more
than any other device utilized in contemporary elections, it has come to embody the process
by which voter preferences determine the makeup of the ruling elite. Yet many citizens of
democratic polities are unaware of the series of institutional innovations in the dissemination
and composition of the ballot that led to its enshrinement as the central object of democratic
contestation. Among such innovations, it was a specific packet of reforms in ballot preparation
and distribution, the so-called Australian ballot (AB ) reforms, which were responsible for one
of the most crucial turning points in the history of democracy—the transition from de jure to
de facto secrecy in vote choice.
What did the adoption of the AB entail? In many countries, the adoption of the AB was
preceded by a system in which nominally secret voting was enacted through privacy guarantees
provided at the moment of suffrage, usually in the form of closed voting booths and a requirement that ballots not have markings that could identify individual voters. In such contexts,
parties and/or candidates printed ballot papers themselves and distributed them, via campaign
workers, to potential voters. The ballot papers utilized in such circumstances were specific to
the candidate or party, and represented a vote only for that individual or organization. This
state of affairs allowed candidates or campaign workers to condition any benefits (or punishments) targeted to individual voters on their receipt of the candidate or party printed ballot.
Moreover, in such settings political actors often organized the transportation of voters to the
polls and allocated the candidate or party printed ballots to voters immediately before the
latter entered the polling station, thereby reinforcing the actors’ capacity to infer whether or
not voters had voted their way. In an AB system, by contrast, all ballots—printed by the state
at public expense—have an identical format, permit the selection of any registered candidate,
and are typically allocated to voters inside the polling station by electoral authorities. As such,
the capacity to monitor individual vote choice is much reduced by the adoption of the AB. At
the same time, the AB also imposes new educational requirements on voting, as the ability to
1
read the printed word may suddenly become a prerequisite for casting a valid ballot.
This paper provides an empirical examination of how the adoption of the AB, and ipso
facto, the transition from the nominal to effective secret vote, shapes the nature of political
representation. It does so by drawing from the historical experience of Brazil before and after
the introduction of the AB in this country in the 1950s and early 1960s. Engaging the claims of
both an emerging comparative politics literature on institutions and clientelism in democratic
development as well as an established literature on the consequences of the AB in the United
States, the paper assesses the impact of the Australian ballot on two outcomes: 1) the degree of
electoral control enjoyed by local vote brokers (coronéis as they are called in the historiography
of Brazilian elections); and 2) the capacity of citizens to meaningfully participate in the electoral
process.
In exploring how the AB shapes these outcomes, the paper contributes to an evolving scholarly debate about how effective vote secrecy shapes the political representation of voters. The
dominant vein of scholarship on this topic envisions the effect of vote secrecy as being primarily
one of voter emancipation. In other words, the adoption of secrecy enhancing electoral institutions such as the AB is viewed first and foremost as a means of insulating voters from acts of
economic and physical coercion that might otherwise prevent them from manifesting their true
preferences at the ballot box. In this perspective, institutions such as the AB open up politics
to a wider set of viewpoints and political parties than would have been countenanced under the
previous set of electoral arrangements. A contending branch of scholarship on this issue views
the impact of secrecy enhancing institutions like the AB in a nearly diametrically opposed way.
Pointing to the accessibility challenges associated with the introduction of an official ballot
in contexts of low literacy, this work views the AB as a vehicle for voter disenfranchisement.
Scholarship in this vein views the effect, and quite often the intent, of the introduction of the
AB to be to narrow the franchise in a pre-determined way, thereby purging the electorate of
voters deemed to be undesirable. According to this view, the specific mechanism by which
disenfranchisement is attained is via the erosion of meaningful political participation by less
2
educated voters.
To adjudicate between these perspectives, this paper brings to bear a painstaking assembled
dataset containing municipal-level electoral returns for federal deputy and senate contests in
the electoral cycles immediately prior to and after the AB was introduced for federal deputy
elections in Brazil. As will be described in detail below, the unusual staggered and geographically and institutionally targeted fashion in which the AB was introduced in Brazil provides
the study with an unprecedented opportunity to measure with precision the causal impact of
this important democratic innovation. There are three sources of variation that undergird the
empirical analysis: 1) spatial (differential adoption of the AB across states); 2) temporal (differential adoption of the AB over time); 3) institutional (differential timing of the adoption of
the AB for different types of elected offices). Taking maximal advantage of these three sources
of variation, the paper variously utilizes both the difference-in-differences technique and a triple
differences method to estimate the impact of the AB on control by local elites and meaningful
political participation. Moreover, in order to better understand the factors that drove electoral
reform, the paper supplements its statistical analysis with a detailed discussion of parliamentary debates and public and private pronouncements on the AB in the period leading up to its
adoption.
The paper’s findings indicate that the primary and immediate consequence of ballot reform
in Brazil was suffrage restriction. Rather than liberating poor and dependent voters, the official
ballot made it exceedingly difficult to have their votes counted. In particular, the paper finds
that the introduction of the AB led to an enormous increase in the percentage of wasted votes
cast by voters. The increase in wasted votes caused by the introduction of the AB was especially
pronounced in those areas where levels of illiteracy were highest, indicating that vote wastage
reflected an inability of lesser educated voters to properly utilize the new ballot. In contrast, the
paper finds no evidence in favor of the proposition that the introduction of the AB weakened
the ability of local elites to guide or coerce the voting decisions of voters still able to make their
votes counted at the polls.
3
The disenfranchising impact of the AB identified by the paper appears to have been an
anticipated and intended consequence of its introduction. Consultation of the historical record
reveals that the architects of ballot reform explicitly sought to neutralize the electoral influence
of functional illiterates through the introduction of the AB, and that opponents of reform
argued against the AB for fear of the same outcome. The Brazilian experience with the AB
thus offers a cautionary tale for social scientists wishing to interpret the political and historical
significance of secrecy-enhancing ballot reform. Depending on the nature of extant electoral
institutions and the socio-demographic profile of the electorate, there may be a fundamental
trade-off between the secrecy of the ballot and its accessibility. Indeed, Brazil’s experience with
the AB reveals that there are contexts in which the accessibility effects of ballot reform, at least
in a short to intermediate time horizon, may significantly swamp the downstream consequences
of enhanced secrecy.
II.
Theses about the Consequences of the
Australian Ballot
Issues of ballot design have been the focus of considerable scholarly production in political science in recent years. Most publications have concentrated on identifying the ramifications of variation in ballot composition and layout. Work on ballot composition—who gets
on the ballot—has examined in detail issues such as the electoral and policy implications of
candidate nomination control vested in party leaders versus primary voters (cf. Carey and
Polga-Hecimovich 2006; Hirano, Snyder, and Ting 2009; Serra 2011) and the impact of gender
and ethnic quotas on representational equity (cf. Bhavnani 2009; Jones 2009; Reynolds 2005).
The literature on ballot layout—the organization and presentation of choices on the ballot—
encompasses a vast research program on the manner in which the capacity of voters to manifest
preferences over candidates contained within party lists (in proportional representation systems) affects party discipline (cf. Carey 2009; Desposato 2006; Morgenstern 2004), the content
and targeting of public goods (cf. Ames 1995; Crisp et al. 2004), and more distal outcomes
4
such as corruption (cf. Gingerich 2013; Golden and Chang 2001; Persson, Tabellini, and Trebbi
2003). It has also spawned several targeted research programs examining the electoral impact of
the order of candidates on the ballot (cf. Ho and Imai 2008; Miller and Krosnick 1998; Lijphart
and Lopez-Pintor 1988), the use of color and symbols (Reynolds and Steenbergen 2006), and
overall ballot complexity (Niemi and Herrnson 2003; Wand et al. 2001).
This upsurge of research on often relatively fine-grained nuances of ballot design in contemporary democracies has inspired social scientists to take a step back and revisit a question
which has long intrigued historians of democratic practice: How did the transition from de jure
secret voting—the use of the ballot box and closed voting booths—to de facto secret voting—
the introduction of the AB —affect local power structures and electoral outcomes? Two main
responses have been proffered to these questions.
Firstly, scholars have argued for many years that the AB undermines the effectiveness of
using bribes and/or sanctions to manipulate the voting decisions of individual voters, and, as a
consequence, erodes the power of political actors who rely on these tactics as a central means
of bringing out the vote. The logic undergirding this claim rests on the enforceability of vote
contracts with and without the AB. In the absence of the AB, voters generally cannot get away
with accepting the ballot and bribe offered by a political operative then voting their conscience
anyway, nor do they usually have the wherewithal to resist the demands of an employer or
patron without paying a steep price. This is because the reliance of voters on candidate or
party printed ballots makes shirking on a deal or the failure to comply with orders easily
verifiable by the concerned political actor. Of course, in theory voters may be able to accept
the ballot from one political operative then switch it later for another. However, to do so the
voter must solicit the ballot from a different operative; this is a semi-public act which may
very well be observed by the operative the voter has betrayed. Moreover, to the degree that
political operatives control transportation to the polls, the opportunities for such switching
may be practically nonexistent.
In the mid-19th century United States and 1940s Minas Gerais, Brazil, for example, nearly
5
identical institutions were developed to minimize ballot switching. Referred to as “ cooping”
in the former and the use of “ electoral barracks” in the latter, campaign workers would funnel
voters into a building or warehouse, provide them with drink and entertainment until the day of
the election, then carefully guard the voters’ approach to the polling place so that no competing
ballots could ever reach their hands (Bensel 2004, pp.179-184, Carvalho 1958, pp.30-32). In late
19th century Prussia, ballot switching was (literally) policed by the state: police officers would
often detain voters carrying opposition ballots and arrest campaign workers distributing them
(Mares 2015, pp.52-59). In such environments of high vote monitorability, voting contracts are
enforceable by political actors at a reasonable cost. As a consequence, it is typically argued
that without the AB, a market in votes—negotiated either directly between candidates and
voters or indirectly between candidates and brokers who control a captive electorate—is likely
to flourish.
By contrast, it has been argued that the AB makes it virtually costless for a voter to accept
a bribe or countenance a threat then subsequently vote as his heart desires. Individual-level
voting contracts in this context are largely unenforceable, which eliminates the possibility of an
efficiently operating market in votes. Consequently, political actors who enjoy a comparative
advantage in bribery and intimidation are likely to see their political clout decline as a result
of the adoption of the AB. As famously put by the American reformer John Wigmore:
By compelling the dishonest man to vote in secrecy, [the AB ]
renders it impossible for him to prove his dishonesty, and thus deprives
him of the market for it. By compelling the honest man to vote in secrecy
it relieves him not merely from the grosser forms of intimidation but from
more subtle and perhaps pernicious coercion of every sort (1889, p.32).
This logic leads to what one might dub the Wigmore hypothesis: the adoption of the secret
ballot fosters the independence of voters, allowing them to vote according to their conscience
instead of the dictates of employers, state agents, or local bosses.
6
A number of studies have presented evidence consistent with this hypothesis. Drawing
upon data from pre- and post-reform Chile (pre/post 1958), Baland and Robinson (2008; 2012)
show that the electoral strength of Chilean landowners’ favored political parties lessened after
the introduction of the AB, as did the prices of their land. They interpret these results as
indicating the weakening of the coercive electoral power of local elites after reform. In a similar
vein, Mares (2015) draws upon data from Germany before and after the enactment of secrecy
enhancing reforms (pre/post 1903), finding that reform increased support for the left-of-center
Social Democratic party, especially in areas with many low skilled workers (who might otherwise
have been beholden to right-leaning employers). Analyzing US House elections from 1860-1930,
Kuo and Teorell (2016) find that the adoption of the AB by the states reduced the incidence
of contested election results based on claims of vote buying and voter intimidation. For the
case of Costa Rica, Lehoucq and Molina (2002) also present evidence that the introduction of
vote secrecy reduced reports of vote buying and intimidation. Using reports from contested
elections in Britain during the 1820-1906 period, Kam (2016) finds that bribe prices declined
with the introduction of the secret ballot, which is also consistent with the logic of the Wigmore
hypothesis. Finally, in a survey of voters conducted in contemporary Argentina, Stokes (2005)
finds that individuals who received their ballots directly from party agents were more likely
to indicate they had sold their vote, a result suggestive of the potential of the AB to liberate
voters from the influence of vote buyers in that country.
The Wigmore hypothesis is one whose implications should be taken seriously in the context
of Brazilian political development. Rural Brazil has had arguably the most politically relevant
and longest lived experience with vote brokerage by local elites of any polity in the Americas.
A vast set of scholarly contributions from historians, anthropologists, and sociologists describes
the activities of the country’s so-called coronéis, backland bosses who routinely negotiated the
electoral support of their dependents in exchange for control over patronage resources within
their municipalities of influence (cf. Leal 1949; Vilaça and Albuqueque 2003 [1965]; Pang
1973). These individuals were notorious and highly politically active during the time of the
7
introduction of the AB in Brazil. Thus, if the AB has an immediate and powerfully reductive
effect on coercion employed by local elites over the electorate, one would expect to detect this
effect in the Brazilian case.
A second argument about the consequences of the AB concerns its ramifications for political participation. Most extant scholarship on the topic argues that the AB erodes citizen
participation in elections, particularly as measured by voter turnout (cf. Aidt and Jensen 2016;
Heckelman 1995, 2000; Schaffer 2002). One strand of this literature argues that in certain
instances it may be appropriate to think of the AB as a form of suffrage restriction, i.e. a legal
impediment to voting instituted so as to keep certain social groups from successfully participating in the electoral process (Kousser 1973, 1974; Crowley 2006). The claim centers on how the
move to the AB affects the voting process for illiterates. Under a party or candidate printed
ballot system, the mechanics of voting require nothing more than to slip the ballot into the
urn (or into the official envelope and then the urn, as was the case in Brazil), a task requiring
no special educational background. In order to participate effectively with the AB in place,
however, the voter must typically have sufficient literacy skills to recognize the written name
of his favored candidate and be able denote his preference appropriately. If these skills are not
widely held nor uniformly distributed, then meaningful political participation will plumment
among those social groups for whom functional literacy is low.
Modern historiography and social science scholarship shows that this is what occurred in
a number of states in the late 19th century American South (Kousser 1974; Perman 2001).
Democratic party leaders realized that an AB system would effectively disenfranchise AfricanAmericans (whose literacy rates at that time were quite low), thereby giving them an electoral
advantage vis-a-vis their Republican rivals (who depended upon African-Americans as a crucial
part of their support coalition). Indeed, the epoch known as the so-called “ one party South”–
due to the electoral dominance of the Democratic party–followed soon after the introduction of
the AB (as well as a host of other, more explicit, suffrage restrictions). A stanza from a racist
Democratic campaign song in Arkansas from the era illustrates the perception among party
8
members that the AB would disenfranchise African Americans, thereby benefiting the party:
The Australian Ballot works like a charm/It makes them think and scratch/And
when a negro gets a ballot/He has certainly met his match (quoted in Kousser 1974,
p.54)
This type of stanza was not a one off: extensive review of state legislative debates and newspaper
editorials of the time make it abundantly clear that the motivation behind the adoption of the
AB was to disenfranchise African-Americans, albeit to do so in a manner in keeping with the
letter (if not the spirit) of the Constitution (Perman 2001, pp.48-69). Moreover, ecological
analysis of state-level electoral data from the time period strongly suggests that the adoption
of the AB depressed turnout among African-Americans, thereby accomplishing this objective
(Kousser 1974, pp.104-138).
In this way, the experience of the US South leads to what one might refer to as the Dixie
hypothesis: In settings where functional illiteracy is widespread, the adoption of the secret
ballot may operate as a de facto franchise restriction, eliminating the prospect of meaningful
political participation by less educated voters.
So understood, the Dixie hypothesis also needs to be taken seriously for the case of Brazil.
The 1960 demographic census, which took place around the time of the institutional changes
studied here, reported that 40.3% of Brazilians above the age of 18 (the legal age of suffrage) did
not know how to read and write (Instituto Brasileiro de Geografia e Estatı́stica 1967). Thus, a
very large contingent of voters stood to be affected by ballot reform that necessitated familiarity
with the written word. Now, according to the letter of the law, this actually should not have
mattered all that much since illiterates had been legally barred from voting in Brazil since the
Constitution of 1891 (a right they only regained almost a century later with the Constitution of
1988). As a matter of practice, however, illiterates regularly participated in Brazilian elections
throughout the 20th century (Love 1970). This state of affairs was due largely to lax registration
requirements: the proof of literacy required to register to vote entailed no more than signing
one’s name in the presence of an electoral judge. As a consequence, coronéis were known to
9
coach functional illiterates on how to sign their names so that these individuals could later
be brought before a judge and inscribed as registered voters (Blondel 1957, pp.74-77). These
voters would subsequently become part of a coronel ’s clientele, voting the ballots that he or
his underlings allocated to them. The role that these local elites played in mediating the
registration process was so great that their efforts in this regard became popularly known as
“making voters” (fazendo eleitores) (Limongi 2015).
III.
Institutional Background
Brazil’s experience in rolling out the AB was both unusual and fortuitous from the standpoint of the social scientist. The AB was first introduced for presidential and vice-presidential
elections in August 1955 (law 2582). The following year, the AB was extended to all other
elected offices attained by a plurality vote (law 2982, art. 9): governor and vice-governor, senators and their substitutes (suplentes), mayors and vice-mayors, and justices of the peace. Subsequently, in July 1962, law 4109 extended the AB to offices filled by proportional representation—
federal deputy, state deputy, and municipal councilor—but only for select jurisdictions. According to the law, the AB only applied to those proportional contests taking place in the state of
São Paulo, the state of Guanabara (comprised solely of the city of Rio de Janeiro), and state
capitals. A new electoral code promulgated in July 1965 (law 4737) was ostensibly to be the
final step in the process, extending the AB to all electoral contests in all jurisdictions in Brazil.
However, the part of the code dealing with the AB did not actually take effect until 1970.1
In order to get leverage on the impact of the AB, this study examines municipal-level
vote returns for federal deputy and senate contests in the 1958-1962 period, the last set of
legislative contests to take place before the onset of military rule in 1964. During this time,
1. The reason was that, in an attempt to placate the country’s rural bosses, Brazil’s military government
(which took the reins of executive power in 1964) emitted a decree in August 1966 (complementary act 20) which
maintained for the legislative elections of that year the system of candidate printed ballots for proportional
contests held in municipalities with less than one hundred thousand inhabitants and which were not included
in the jurisdictions that received the AB in 1962.
10
general legislative elections—which put seats for both the Chamber of Deputies and Senate
simultaneously up for grabs—were held in 1958 and 1962. In 1958, the AB was in effect
throughout all of Brazil for senatorial contests, whereas candidate printed ballots were utilized
throughout all of Brazil for the deputy contests. In 1962, the AB remained in effect for all
of Brazil for the senatorial contests but was also extended to deputy contests in the state of
São Paulo, the city of Rio de Janeiro, and state capitals. Figure 1 provides a timeline of these
events.
Figure 1: Timeline of Implementation of AB in Legislative Elections in Brazil, 1958-1962
In municipalities where the AB was in effect for both types of contests, voters received
an official ballot from the presiding officer in the polling station divided into two columns:
one allowing the voter to denote her preferences for the plurality contests (which included the
Senate) and another allowing the voter to denote her preferences for the proportional contests
(which included federal deputy). In municipalities where the AB was in effect for the plurality
contests but not the proportional contests, voters received an official ballot from the presiding
officer for the plurality contests only. In this case, the voter voted in two stages. First, she
entered the closed voting cabin and marked her preferences on the official ballot for the plurality
contests, subsequently folding it and dropping it into the urn. Next, she returned to the closed
voting cabin, deposited the candidate printed ballot for the proportional contests into an official
envelope provided by the presiding officer, and subsequently dropped the envelope into the urn.
For the plurality contests, the AB listed the names of each candidate running for each office
(senator, governor, etc.), requiring the voter to denote her preference by making a mark next
to the preferred candidate. For the proportional contests (federal deputy, state deputy, etc.),
11
where the number of candidates running was typically very large, the AB provided the voter
with two lines where she could write in the name or candidate number of her favored candidate
for a given office. The full lists of candidates for these offices was not included on the ballot
itself, although said lists were required to be posted in a visible location within the polling place
(TSE Resolution No. 7018, September 4, 1962).
In the locales where candidate printed ballots were utilized for proportional contests, there
were a series of requirements these had to satisfy. They all had to be white, rectangular in
shape (approximately 7 × 10 cm), and sufficiently flexible to be folded into the official envelopes
furnished by the electoral authorities. The name of the candidate, his party or coalition, and
the office being contested all had to be typed onto the ballot. No other signs, slogans, or
illustrations were permitted on the ballot. These rules were uniform: the regulations governing
the layout of official ballots and candidate printed ballots applied equally to all municipalities
in the territory of Brazil.
Figure 2 displays two of the types of ballots Brazilian voters used during this period. The
panel on the left-hand side of the figure presents an example of an individual ballot. This
particular ballot was used by Octavio de Abreu Sampaio, a candidate for federal deputy in the
1958 legislative election in São Paulo. The panel on the right-hand side of the figure presents
the model of the official ballot used in 1962 in the states of São Paulo and Guanabara as well
as the state capitals. Voters denoted their choices for the plurality contests on the left-hand
side of the ballot and their choices for the proportional contests on the right-hand side of the
ballot.
A virtue of the peculiar mode of adoption of the AB in Brazil is that it permits the current
study to improve upon existing analyses of the impact of the AB in several fundamental ways.
First, Brazil’s adoption pattern permits the first set of analyses of the AB outside of the US that
are able to exploit both cross-sectional and temporal (not to mention institutional) variation in
its adoption. This naturally enhances leverage for making causal inferences from the data and
strengthens the credibility of conclusions about the impact of the AB. Second, the Brazilian
12
case permits a somewhat cleaner analysis of the impact of the AB than studies focusing on
the US case, where both temporal and cross-sectional variation in the adoption of the AB were
also present.2
Figure 2: Types of Ballots used in Brazilian Elections, 1958-1962
Note: The left-hand side panel displays a candidate-printed ballot utilized by Octavio de Abreu Sampaio in the Federal Deputy
contest in São Paulo state in 1958 (source: Estado de São Paulo, October 3, 1958, p.9). The right-hand side panel displays a model
of the official ballot (AB ) employed in select jurisdictions in 1962 (source: Jornal do Brasil, July 27, 1962, p.4).
The inferential challenge in studying the AB in the US context is that the internal organization of the official ballot differed drastically by state, as did the nature of the party printed
ballots that preceded the official ballot. As a consequence, it is difficult to speak of the impact of “the” AB in the US, since no such single entity existed.3 This is not an issue for the
Brazilian case. As described above, Brazil’s electoral legislation carefully specified the characteristics that both the official ballot and the candidate printed ballots had to satisfy. These
were nationwide standards strictly enforced by the regional electoral tribunals that make up
the country’s centralized electoral justice system. Consequently, the presence or absence of the
2. Scholars examining the introduction of the AB in the United States have been particularly concerned with
its impact on the strength of the party system, especially as manifested in split ticket voting. See, inter alia,
Rusk (1970; 1974), Converse (1972), Burnham (1974), and Reynolds and McCormick (1986).
3. Put in the language of program evaluation, the myriad and subtle ways in which the AB was implemented
across states in the US make it probable that analyses based on the US case will suffer from biases due to
“unrepresented versions of the treatment” (Rubin 1986).
13
AB for a given office in one state meant the same thing as it would in any other state.
IV.
Data and Measurement
The primary data upon which this study draws are municipal-level vote returns, broken
down by candidate and party, for federal deputy and senate contests held in 1958 and 1962.
These returns are contained in municipal acts (atas municipais) which are housed in the archives
of regional electoral tribunals located in the capital of each Brazilian state. A team of research
assistants, composed of Brazilian nationals, was paid to visit these electoral tribunals and
provide a digitized copy of the returns. Upon receipt of the digitized returns, these were then
coded into electoral format by a team of graduate and undergraduate research assistants at
the author’s home institution. In total, municipal-level returns from this time period which
included both federal deputy and senate elections were obtained for eight states: São Paulo
(the treatment state), Bahia, Paraı́ba, Paraná, Pernambuco, Piauı́, Rio Grande do Sul, and
Rio de Janeiro state (the control states). The secondary data upon which this paper relies are
demographic and socio-economic indicators from Brazil’s census of 1960. Such indicators include
population figures, literacy rates, employment broken down by sector (agriculture, industry),
and several measures of material wealth.
The sample utilized in this study consists of interior municipalities only (non-state capitals)
within the states indicated. This is because the experience of interior municipalities provides
unique leverage in assessing the impact of the AB. In 1958, all such municipalities had the AB
in place for the senate contest but not the federal deputy contest, providing a common baseline
with which to compare the experience of the São Paulo municipalities and the control state
municipalities. In 1962, all the interior municipalities of São Paulo adopted the AB for the
federal deputy contest, whereas as none of the interior municipalities of the control states did
so, permitting one to assess the impact of the AB from observed over time changes in electoral
outcomes across municipalities in São Paulo and the control states.
14
Outcomes. The dependent variables utilized in the analysis correspond to the two hypothesized consequences of the AB discussed above. In order to measure the coercive power
and influence of Brazil’s local vote brokers, the coronéis, we use as our dependent variable
the Herfindahl-Hirschman index of vote concentration for contests for federal deputy. For a
municipality m in a given election, the vote concentration index is equal to:
N
X
V CIm =
s2i,m ,
i
where si,m is the vote share of candidate i in municipality m and N is the total number of
candidates. The V CI is bounded between 0 and 1, with higher values of the variable denoting
a greater concentration of votes on a small number of candidates.
The motivation for using vote concentration in federal deputy contests as a proxy for broker
control stems from Brazil’s unique high magnitude open list proportional representation system
in force for such contests. Federal deputies in Brazil run for election in districts which consist
of entire states. Given the large number of seats up for grabs in most states, it is customary
for dozens if not hundreds of candidates to run for election at a given time, which makes
the potential dispersion of the vote across candidates quite high. This theoretical potential
for dispersion notwithstanding, a political anthropology literature has documented that where
powerful coronéis were present, a municipality’s votes tended to be highly concentrated on the
single candidate who had been endorsed by the reigning broker (or otherwise split between two
candidates if there happens to be dueling brokers) (Alves 2006, cf. Greenfield 1977, and Gross
1973). For this reason, vote concentration serves as a reasonable approximation of the electoral
control of Brazil’s local brokers.4
In interpreting this index, it is important to understand that the recommendations coronéis
gave to their voters were absolutely explicit, as were the favors for votes arrangements the
4. Since Brazil’s open list system of proportional representation allows voters to vote for one of many different
candidates running under the same party label, instances of high vote concentration are not explicable in terms
of ideological sympathies for particular party platforms. For discussion of the link between vote concentration
and oligarchical politics in Brazil, see Ames (2001) and Soares (2001).
15
coronéis sealed with politicians. In a classic study of four Northeastern coronéis, Vilaça and
Albuquerque (2003 [1965]) described the voting behavior of coronel -dominated towns as follows:
They are votes ceded to deputies and senators, sometimes as a consequence of a
simple telegram sent to the coronel on the eve of the election, when the state party
bosses perceived, at the last minute, that one or another candidacy of interest was
weak. Remote-control votes (votos teleguiados)...The dominion of the coronel over
his electoral college was effectively absolute during the golden era of his political
power (pp.60-61).
Thus, one would expect that where coronéis are active and powerful, votes should concentrate
heavily on the particular candidate with whom a deal had been struck. Where they are absent
or weak, less concentration would be expected. The Wigmore hypothesis would lead one to
expect a decline in vote concentration with the adoption of the AB, since the ability of coronéis
to monitor individual voting behavior declines with vote secrecy.
It should be noted that the vote concentration index is a relevant measure only for federal
deputy contests. The use of such an index in senate contests is not particularly useful, as those
contests employed either single member districts in entire states (1958) or two member districts in the same (1962) during the period studied here, thereby sharply limiting the potential
dispersion of the vote.
In order to capture (the inverse of) effective political participation, the paper utilizes as
an outcome variable the proportion of wasted votes (blank votes plus null votes) among all
votes cast. This is a particular appropriate indicator of the quality of political participation in
Brazil due to the fact that voting was and remains obligatory for registered voters, with fines
and even a prohibition against future public employment assessed against violators. Given the
costs associated with sitting out the election, one would expect that if the AB indeed made
election procedures unintelligible to some voters, the evidence for this would show up primarily
in wasted votes and less in related outcomes such as a decline in turnout. Thus, the paper
16
evaluates the Dixie hypothesis by examining the impact of the AB on the percentage of wasted
votes cast by voters.
Municipal “ Emancipation”. The primary unit of analysis for this study is the Brazilian
municipality, a political unit roughly akin to a county in the United States. During the period
from 1958-1962, some of the municipalities in the study sample underwent a so-called process
of emancipation, meaning that they broke up into two or more new municipalities. In instances
where this occurred, newly created municipalities in the second electoral period (1962) were
grouped in such a way as to ensure intertemporal comparability with municipalities in the
first electoral period (1958). For example, in February 1959, the district of Colômbia, São
Paulo, broke away from the municipality of Barretos to become its own municipality. Thus,
the electoral and demographic information for the municipalities of Barretos and Colômbia in
1962 was combined to provide comparability with the municipality of Barretos in 1958. Other
instances of municipal liberation were handled in the same way. All told, our sample consists
of 422 municipalities in the interior of the state of São Paulo and 714 municipalities located in
the interior of the control states.
V.
Identification and Estimation
The causal estimand of interest in the paper is the average treatment effect for the treatment (ATT ), the average effect of the AB for the outcome in the federal deputy contest for
those municipalities that utilized it for said office (i.e. the municipalities in the state of São
Paulo). This paper utilizes two approaches for estimating this quantity. In order to study the
impact of the AB on the vote concentration index, which varies across space and time but not
types of office contested, the difference-in-differences technique is employed. This consists of
comparing the average pre/post reform (1958/1962) change in vote concentration in the interior municipalities in the state of São Paulo to the average over time change in the interior
municipalities of the other, non-reform, states during the same time period. In studying the
17
impact of the AB on the proportion of wasted votes cast, which varies across time, space, and
office contested, both the difference-in-difference technique and the triple difference technique
are utilized. Here the difference-in-differences analysis simply utilizes the percentage of wasted
votes in the federal deputy contest as the relevant outcome but otherwise approaches the analysis in the same way. The triple differences technique, on the other hand, makes use of the data
from the senate contests. In particular, it compares the average pre/post reform (1958/1962)
change in the difference in the proportion of wasted votes across federal deputy versus senate
contests in the interior municipalities in the state of São Paulo to the average over time change
in the difference in the proportion of wasted votes across federal deputy versus senate contests
in the interior municipalities of the non-reform states.
These two estimation strategies rely on somewhat different identification assumptions. The
use of difference-in-differences assumes parallel paths between the municipalities in São Paulo
and those in the non-reform states. This means that the average change in a given electoral
outcome in the federal deputy contest would have been the same for the municipalities in the
state of São Paulo as it was for the municipalities in the other states if the municipalities in São
Paulo had not (counterfactually) received the AB in 1962. In essence, the assumption requires
that there are no variables omitted from the analysis that may be driving differences in the
trajectories of electoral outcomes in São Paulo versus the control states.
The triple differences technique relies on a weaker parallel paths assumption. In particular,
it assumes that the average over time difference in the difference in wasted votes across offices
for the São Paulo municipalities would have been the same as the average over time difference in
the difference in wasted votes across offices for the control state municipalities had the São Paulo
municipalities not utilized the AB for the federal deputy contest. This very weak assumption
allows for arbitrary baseline differences between São Paulo and control state municipalities due
to (time invariant) unobserved heterogeneity, the existence of distinct time trends for federal
deputy and senate contests, and the possibility that sources of unobserved heterogeneity affect
18
outcomes for federal deputy contests and senate contests in different ways.5
The triple differences technique also requires making a second identifying assumption. It is
that, on average, the ballot technology adopted for the federal deputy contest did not affect the
proportion of wasted votes in the senate contest for the municipalities that employed the AB.
This assumption would seem quite plausible on its face since all municipalities employed the
AB for the senatorial contest–irrespective of whether or not they did so for the federal deputy
contest. Given that the ballot technologies for the two contests were effectively delinked, there
is no compelling reason to believe that the presence or absence of the AB for the federal deputy
contest would have systematically affected patterns of wasted votes for the senate.
In conditioning on baseline characteristics of municipalities, the statistical analysis below
utilizes several fairly standard strategies for estimating the ATT, including nearest neighbor
matching (Rosenbaum and Rubin 1985), local linear matching (Heckman, Ichimura, and Todd
1997), and coarsened exact matching (Iacus, King, and Porro 2011). In the interest of assessing
the robustness of the findings, the ATT is also estimated using such methods as propensity
score weighting (Hirano and Imbens 2001) and ordinary least squares regression.
VI.
Findings
Raw Data. I begin by examining the spatial, temporal, and institutional differences in
outcomes without conditioning on the demographic features of municipalities. First consider
the impact of the AB on the proxy for broker control, the vote concentration index (VCI).
It is important to recognize that this variable is constructed based upon the valid votes cast
for all candidates within a given municipality. As a consequence, it can be thought of as a
measure of broker influence over the vote choices of individuals who cast their votes correctly,
5. Such flexibility is substantively important. For instance, the weak identifying assumption of the triple
differences procedure would allay potential concerns that unmeasured aspects of the socio-economic development
of São Paulo could be driving differences in the rate of change in wasted votes, a situation that could pose a
challenge to traditional difference-in-differences.
19
i.e. individuals who were able to cast votes that pasted muster with electoral authorities and
thus counted towards the final totals that determined the election of candidates.
Figure 3: The Estimated Impact of the AB on Municipal Vote Concentration, (Interior
Municipalities, 1958-1962) Raw Estimates (No Conditioning on Covariates)
Note: The upper panels depict box-and-whisker plots and jitterplots of the municipality-level vote concentration index in the
federal deputy contest occurring in the year indicated. The bottom panel shows that municipalities in the interior of São Paulo,
which had the AB in place for the federal deputy contest in 1962 but not in 1958, experienced a slightly greater decline in the vote
concentration index than did municipalities in the interior of the control states (Bahia, Paraı́ba, Paraná, Pernambuco, Piauı́, Rio
de Janeiro, Rio Grande do Sul), which did not have the AB in place for federal deputy during either year.
Figure 3 displays the manner in which this variable changed over time in the municipalities
in São Paulo and the control states, respectively. The uppermost panels in the figure present
box-and-whisker plots and jitterplots of the distribution of the VCI in the 1958 and 1962
elections. There was a slight decline in municipal-level vote concentration in both São Paulo
20
and the control states, indicating a marginal increase in the competitiveness of elections at the
local level. However, the magnitude of the change was small in both instances.
The bottom panel of the figure provides a differences-in-differences diagram which illustrates this explicitly. For municipalities in the state of São Paulo, the average value of the
VCI declined from 0.15 in 1958 to 0.12 in 1962. For municipalities in the control states, the
average value of the VCI declined from 0.21 to 0.18 during the same period. Thus, the naı̈ve
(without conditioning) difference-in-differences estimate of the ATT is equal to -0.01 (rounding
to the second decimal), with a 95% confidence interval equal to [-0.02,0.00] and a p-value of
0.28. What these results seem to indicate is that there was no radical reduction in the capacity
of Brazil’s coronéis to control the (valid) vote as a consequence of the introduction of the AB.
Municipalities which conducted federal deputy elections using the AB did not experience a
significantly larger shift towards more competitive elections at the local level than did municipalities which conducted said elections using the old candidate printed ballots. The first cut
at the data does not support the contentions of the Wigmore hypothesis.
Now turn to Figure 4, which presents difference-in-difference graphs depicting how the
proportion of wasted votes in federal deputy contests evolved as a function of the introduction
of the AB. The figure shows that there was an enormous increase in wasted votes cast in
federal deputy elections in the municipalities in São Paulo from the pre-reform election to the
post-reform election. In 1958, prior to the adoption of the AB, the average proportion of
wasted votes cast in São Paulo municipalities was 0.11. In 1962, after the adoption of the AB,
the average proportion of wasted votes cast more than tripled , increasing to 0.34. Over the
same time period, the change in the proportion of wasted votes cast in the municipalities of the
control states was relatively modest, increasing from 0.08 to 0.12. Following the same procedure
described above, the naı̈ve ATT estimate for the effect of the AB on the proportion of wasted
votes cast is equal to 0.19, with a 95% confidence interval of [0.18, 0.20]. This suggests that
the AB indeed eroded the capacity for meaningful political participation for a large segment of
the electorate, confirming the expectations of the Dixie hypothesis.
21
Figure 4: The Estimated Impact of the AB on Wasted Votes, Difference-in-Differences
(Interior Municipalities, 1958-1962) Raw Estimates (No Conditioning on Covariates)
Note: The upper panels depict box-and-whisker plots and jitterplots of the proportion of wasted votes in the federal deputy
contest occurring in the year indicated. Due to missing data, the state of Pernambuco was omitted from the analysis. The bottom
panel shows that municipalities in the interior of São Paulo, which had the AB in place for the federal deputy contest in 1962
but not in 1958, experienced a much greater increase in wasted votes in the federal deputy contest than did municipalities in the
interior of the control states (Bahia, Paraı́ba, Paraná, Piauı́, Rio de Janeiro, Rio Grande do Sul), which did not have the AB in
place for federal deputy during either year.
Figure 5 presents a triple differences analysis to gauge the impact of the AB on the proportion of wasted votes cast. Displayed in the figure are over time changes in the difference in the
proportion of wasted votes cast in federal deputy versus senate elections.
22
Figure 5: The Estimated Impact of the AB on Wasted Votes, Triple Differences (Interior
Municipalities, 1958-1962) Raw Estimates (No Conditioning on Covariates)
Note: The upper panels depict box-and-whisker plots and jitterplots of the difference in the proportion of wasted votes cast in the
federal deputy versus senate contest occurring in the year indicated. Due to missing data, the state of Pernambuco was omitted
from the analysis. The bottom panel shows that municipalities in the interior of São Paulo, which had the AB in place for the
federal deputy contest in 1962 but not in 1958, experienced a significant increase in the cross-office difference in proportions of
wasted votes, whereas the municipalities in the interior of the control states (Bahia, Paraı́ba, Paraná, Piauı́, Rio de Janeiro, Rio
Grande do Sul), which did not have the AB in place for federal deputy during either year, experienced a decrease in the same.
As above, in the municipalities within São Paulo, one observes a remarkable shift. In 1958,
when the AB was in place for senate elections but not federal deputy elections, on average the
proportion of wasted votes in senate contests was substantially greater than that encountered
for federal deputy contests (0.24 vs. 0.11). In 1962, when the AB was in place for both types
of contests, on average the proportion of wasted votes in the two types of contests was nearly
identical (0.33 vs. 0.34). In the control state municipalities, on the other hand, one finds a
different pattern. In 1958, as in the case of São Paulo, on average the proportion of wasted votes
23
in senate contests was substantially greater than that encountered for federal deputy contests
(0.16 vs. 0.08). In 1962, with the institutional rules in these municipalities staying fixed, one
continues to observe a substantially greater proportion of wasted votes for senate contests than
for federal deputy contests (0.28 vs. 0.12). It would seem extraordinarily difficult to explain
away this peculiar pattern of a large relative increase in wasted votes in the São Paulo federal
deputy elections other than to accept the hypothesis that the AB led substantial challenges in
accessing the ballot.
By comparing the over time difference in the cross-office differences in wasted votes across
São Paulo and control municipalities, one calculates a naı̈ve estimate of the ATT equal to 0.22.
This quantity is very precisely estimated, having a confidence interval of [0.21,0.23] and a pvalue which for all intents and purposes is zero. It is also astounding large. On the basis of
this estimate, one would conclude that the introduction of the AB on average increased wasted
votes as a percentage of total votes cast in federal deputy elections in municipalities in the
interior of São Paulo by 22 percentage points. To put this number in perspective, the top vote
getting list in the entire state of São Paulo in 1962 only received 17% of total votes cast in
the federal deputy election. In other words, the AB increased the percentage of wasted votes
cast by an amount greater than the total percentage of votes being tallied by the country’s
major party organizations! These results suggest that the AB almost certainly had a major
disenfranchising effect on the Brazilian electorate.
Estimates with Covariate Conditioning. The state of São Paulo is different from other
states in Brazil in a number of important ways. It is the country’s most populous state and
it has the largest economy of any Brazilian state by far. Moreover, São Paulo is among the
country’s most educated states, with an extensive university system, a large industrial base,
and a well developed service sector. Of course, São Paulo, like all Brazilian states, is highly
diverse: it contains municipalities with extremely high levels of wealth and human capital and
municipalities with substantially lower levels of the same. In order to make certain that the
estimates of the causal effect of the AB are not being driven by differences in the baseline
24
characteristics of municipalities in São Paulo and the control states, this section of the paper
presents ATT estimates based upon explicitly conditioning on such characteristics. Since the
paper’s approaches to estimation should alleviate most concerns about confounding, covariates
were employed in a limited and judicious manner. Specifically, the analysis conditions on
four key demographic and socioeconomic indicators (all measured in 1960) along with one
relevant political indicator: 1) the population size of the municipality; 2) the literacy rate in
the municipality; 3) the proportion of permanent homes with a radio (a proxy for access to
political information); 4) the proportion of the working age population employed in industry;
5) the proportion of the vote won by the Social Democratic Party (PSD) party in the federal
deputy elections of 1958.
The economic indicators tap into the level of development and human capital. PSD vote, in
turn, reflects baseline levels of clientelistic politics. The PSD was Brazil’s largest political party
of the time. Contrary to its name, it was largely a conservative organization within which the
country’s backland bosses enjoyed outsized influence. Thus, the pre-reform electoral strength
of the PSD is likely to tap into exiting bases of coronelismo.
Table 1 depicts the means of these indicators across our treatment and control groups. There
are indeed some stark differences across the groups. For instance, the average literacy rate for
the São Paulo municipalities is substantially higher than that for the control municipalities
(60% vs. 42%), as is the average percentage of homes with a radio (41% vs. 18%). Support
for the PSD was substantially lower (27% vs. 38%). Fortunately, the table also shows that the
conditioning strategies adopted by the paper were able to produce matched control groups which
approximated fairly closely the characteristics of the treatment municipalities. By selectively
weighting municipalities within the treatment group, these algorithms help ensure that the
paper’s estimates of causal impact are based upon a comparison of units that are reasonably
similar.
25
Table 1: Covariate Means Across Treatment and Control Groups (Before and After Conditioning)
nearest
local
p-score
neighbor
linear
coarsened exact
raw data
weighting matching matching
matching
covariate
treated
control
control
control
control
treated
control
population
21306
32049
22565
25522
23330
47615
40876
literacy rate
59.7%
42.3%
68.3%
62.7%
63.5%
61.7%
60.4%
homes w/ radio
40.5%
17.8%
50.2%
41.0%
41.5%
46.5%
42.8%
working age
pop. in industry
4.1%
2.2%
5.1%
3.4%
3.9%
6.9%
5.1%
PSD vote in 1958
27.1%
38.1%
29.4%
30.6%
30.6%
27.9%
25.8%
Figure 6 presents dotplots of the ATT estimates and their 95% confidence intervals for each
outcome examined in the paper along with each covariate conditioning method. The basic story
told by the estimates in the figure is consistent with the findings based on an examination of the
raw data. The plots provide little evidence that the introduction of the AB had an impact on
the concentration of the vote. Although the negative sign of the estimated ATT was consistent
with the Wigmore hypothesis, in no case could the null that the estimated impact of the AB
on vote concentration was zero be definitively rejected. In contrast, the findings on wasted
votes remained exceptionally stark. The introduction of the AB clearly engendered a massive
increase in the proportion of wasted votes cast, with point estimates of the effect ranging from
0.16 to 0.19 utilizing the difference-in-differences approach and from 0.17 to 0.22 utilizing triple
differences. As before, these estimated effects had extremely narrow confidence intervals. In
sum, the AB disenfranchised a sizeable segment of the electorate but it did not appear to have
any immediate effect on the persuasive power of local vote brokers.
26
Figure 6: Estimated Impact of the AB on Vote Concentration and Wasted Votes (Interior
Municipalities) Conditioning on Covariates Using the Method Indicated
Note: ATT point estimates are denoted by black circles; 95% confidence intervals are denoted by black lines.
Subgroup Effects of the AB. In order to further probe the degree to which the data
are consistent with the expectations of the Wigmore and Dixie hypotheses, this section reports the results of a series of analyses examining the impact of the AB for distinct subsets
of municipalities. Firstly, in order to maximize the possibilities of detecting an erosion of the
electoral influence of Brazil’s rural bosses due to the introduction of the AB, subgroup analyses
were conducted in which the impact of the AB on vote concentration was estimated for subsamples based on the terciles of population size, literacy rate, and pre-reform PSD vote share,
respectively. Given that the literature on clientelism emphasizes that vote brokerage thrives
in small towns and in settings of low human capital, one would expect the impact of the AB
to be most acute in sparsely populated municipalities and/or municipalities with low levels of
literacy. Similarly, given that the PSD was associated with rural bossism, one might expect the
impact of the AB to have been most pronounced in the strongholds of the party.
Figure 7 presents the results of these subgroup analyses. Given the limited number of
observations per subgroup, ATT estimates were calculated without conditioning on covariates
or by conditioning using ordinary least squares. For no subgroup of the data was the impact of
the AB on vote concentration statistically significant by conventional standards. The impact
of the AB came closest to statistical significance in the municipalities in the upper tercile of
27
pre-reform PSD support, but even here the null of no effect could not be rejected. Population
size had no apparent effect in mediating the impact of the AB on concentration and neither
was there any clear mediating influence of the literacy rate.
Figure 7: Subgroup Impact Estimates of the AB on Vote Concentration (Subgroups Defined
by Terciles of the Indicated Covariate)
Note: ATT point estimates are denoted by black circles; 95% confidence intervals are denoted by black lines.
Secondly, in order to detect if the underlying mechanism of the Dixie hypothesis is correct,
i.e. that the AB specifically disenfranchises functional illiterates, subgroup analyses were conducted to examine whether or not the AB had a differential effect on wasted votes depending
on the prevailing level of literacy. Figure 8 presents the results of these analysis. The figure
indeed reveals an extremely strong differential effect of the AB. This is the case both when
difference-in-differences is utilized as well as when triple differences are employed. The largest
disenfranchizing effects of the AB were encountered in those municipalities with the highest
rates of illiteracy, followed by municipalities with intermediate rates of illiteracy, with the most
28
literate municipalities, in turn, suffering the smallest disenfranchizing effects of the AB. Using
the triple differences procedure, in municipalities where the literacy rate was the lowest (less
than 45% literate), the estimated impact of the AB on wasted votes was approximately twice as
large as the estimated impact of the AB in municipalities where the literacy rate was the highest
(greater than 60% literate). The absolute magnitude of the impact of the AB on wasted votes
within the least literate subsample was truly striking: in these municipalities, approximately
a third of the otherwise correctly voting public cast a wasted vote due to the introduction of
the AB. All told, the results of the subgroup analyses provide strong support for the Dixie
hypothesis but are unable to detect evidence in favor of the Wigmore hypothesis.
Figure 8: Subgroup Impact Estimates of the AB on Wasted Votes (Subgroups Defined by
Terciles of the Literacy Rate)
Note: ATT point estimates are denoted by black circles; 95% confidence intervals are denoted by black lines.
29
VII.
Suffrage Restriction as an Intended Effect of
Reform
The results presented above on the disenfranchising effects of the AB may be surprising
to some political scientists, but they would not have been to the proponents of reform. The
role that the introduction of the AB would likely play in disenfranchising illiterate voters in
Brazil was understood from the very beginning. In fact, the architects of ballot reform made
it exceedingly clear that they believed purging the electorate of illiterates was a necessary and
salutary step for the moralization of the vote in the country.
The earliest advocates of electoral reform laid out their case based on the perceived inability
of the uneducated rural voter to thoughtfully participate in the electoral process. Since such
voters would always be easy prey for local bosses and vote buyers, went the thinking, the
best path forward for Brazilian democracy would be to simply eliminate any prospect of their
participating meaningfully in elections. For instance, Ismar de Góes, PSD senator from Alagoas
and Minister of the Superior Military Tribunal, remarked in a January 18,1954 pronouncement
on the Senate floor that “the depoliticized interior voter...votes as if paying a tribute and obeys
the most powerful” (Boletim Eleitoral, February 1955, p.322). Consequently, electoral reform
required “the adoption of a list of names, with the goal that the voter reads and insodoing votes
for the candidate or party of his preference.” Góes concluded: “This would already be a form
of selection of the electorate; and selection is indispensible for future elections” (ibid). Picking
up on this last point, Olavo Oliveira, PSP senator from Ceará, immediately responded: “The
vice is in the root (de origem), it is in the composition of the electorate. As long as we do not
modify it, nothing will improve” (ibid).
The most critical phase in the adoption of the AB was the elaboration and presentation of
project law n◦ 94 (1955), presented to the Chamber of Deputies on March 17, 1955. Written by
Edgard Costa, president of Brazil’s Supreme Electoral Tribunal (TSE), the document states in
its “justifications” section that among the advantages of the AB is the fact that it “demands
from the voter a minimum of reflection and discernment at the moment of casting his vote,
30
thereby requiring that he knows, at least, how to read easily” (Diário do Congresso Nacional
I, March 22, p. 1339). The specific ballot layout proposed in the law was justified on the
grounds that it would “eliminate the possibility of the vote by the illiterate voter” (ibid.). Lest
there be any uncertainty about the matter, Alexandre Marcondes Filho, the Minister of Justice,
provided an exposition praising the project law (which also promised to strengthen registration
requirements) in which he extolled the fact that it would “verify the literacy of the voter” (ibid.,
p.1340).
Reactions to the potentially disenfranchising consequences of the AB were immediate and
forceful on both sides of the issue. Addressing the floor of the Chamber on March 19, 1955,
Antonio Bruzzi de Mendoça, then PRT federal deputy from the Federal District (city of Rio
de Janeiro), railed against the project. Speaking in particular about the prospect of using
candidate numbers to vote in proportional representation contests, he argued this amounted
to “nothing more and nothing less than a masked attempt to impede a good part of the
population, exactly that part referred to as less educated, that part of the population which is
less literate, from expressing themselves in electoral contests” (Diário do Congresso Nacional
I, March 19, 1955, p.1319). Pushed by Deputy Arnaldo Cerdeira (PSP-SP) to explain how he
came to the conclusion that the AB was part of a concerted attempt by elites to deprive the
poor and uneducated of the vote, Bruzzi de Mendoça described an earlier conversation with
Deputy Oscar Correia (UDN-MG), in which Correia purportedly stated: “This is precisely the
objective. I believe that it is preferable to reduce the Brazilian electorate to half its current
size, so long as it becomes ever more educated” (ibid., p.1320).
Other deputies expressed their admiration for the franchise-restricting properties of the AB
more openly. Herbert Levy, a UDN party leader from São Paulo, argued that the adoption of
the AB was a crucial component of any electoral reform since it would make voting impossible
for “the incapable voter” and also “the illiterate, he who habitually follows the guidance of
ward heelers (cabos eleitorais) in the choice of his candidates” (speech to the Chamber of
Deputies, May 12, 1955; printed in Boletim Eleitoral, June 1955, pp.527-528). Carlos Lacerda,
31
the infamous journalist and UDN deputy from the Federal District, lauded the AB ’s narrowing
of the franchise on the grounds that “although it is clear that it is in the interest of democracy
to have voters, those voters should be conscious, capable, and not voters whose incompetent
and uninformed opinion will destroy the collective patrimony (acêrvo) of the electorate that
is already conscious, politicized, and capable of assuming the responsibility of choosing its
leaders” (speech to the Chamber of Deputies, May 26, 1955; printed in Boletim Eleitoral, June
1955, pp.542). Tenório Cavalcanti, UDN federal deputy from Rio de Janeiro, put it bluntly: “If
we vote for this law, we are at least going to improve the quality of the Brazilian electorate”
(ibid.,p.547).
Opponents of the AB similarly emphasized how reform would affect the ability of less
educated voters to vote and, ipso facto, the scope of the franchise. Gustavo Capanema, PSD
federal deputy from Minas Gerais and majority leader in the Chamber, expressed his opposition
to the AB in terms of accessibility concerns, stating, “I am persuaded...that the great electoral
mass of the interior, which I know, since I was born and raised in it, that great electoral mass of
men and women of all ages and conditions is not able to enter the voting booth and vote with all
the complications of the [AB ]” (ibid.,p.543). Reacting to accusations by deputy Adauto Cardoso
(DF-UDN) that “the vote for illiterates is what is being revindicated here–it is a new coup,”
Capanema ironically responded that the AB might as well be included in a broader packet of
reforms that “definitively tears apart the principal of universal suffrage” (ibid.). In a Chamber
debate on the extension of the AB to proportional representation contests, Aurélio Viana (PSBAL) commented that the proposed reform “instead of simplifying the voting process, makes it
more difficult, principally for the uneducated, semi-literate voter.” What was needed instead,
he argued, was “a process that allows every Brazilian of voting age to realize democracy through
their vote” (Anais de Camara dos Deputados, July 2, 1962, p.364).
Debates in the Senate followed very similar lines. In a heated exchange, Paulo Fernandes
(PSD-RJ) and Fernandes Távora (UDN-CE) debated the virtues and demerits of the AB explicitly in terms of the inclusion or exclusion of functionally illiterate voters in the electoral
32
process. Responding to a statement by Fernandes that the AB would not be a panacea for
all that ailed Brazilian democracy, Távora shot back back: “At least one of the principle evils
would be resolved, as illiterates would not be able to vote” (Diário do Congresso Nacional II,
August 19, 1955, p. 2012). He continued, “With the [AB] system, the illiterate voter will no
longer vote with the ballot given to him by his employer or the ward heeler (cabo eleitoral ). If
he does not know how to write, he will not vote” (ibid.). Reacting to this line of argument,
Fernandes worried aloud that “the implantation of the [AB ] will result in a radical modification
of regime. What we would be instituting, if by chance it was adopted, would be a qualified vote,
a qualified democracy” (ibid., p. 2013). Távora curtly replied, “This alone would represent
a great improvement in the voting process” (ibid.). Fernandes then proceeded to sum up his
opposition by retorting: “If, by unfortunate circumstance, we are a poorly educated people,
then the votes should reflect this deficiency. That is a real democracy: one reflecting the will
of the man who produces and has a right to citizenship” (ibid.). Távora remained implacable:
“Real democracy is one with a conscious vote. Any individual who votes unconsciously is not
a voter, he is a human rag (trapo humano)” (ibid.).
Edgard Costa himself, the primary architect of reform, was publicly quite outspoken in his
view that removing functional illiterates from the electorate would be a positive outcome of
the adoption of the AB. In an interview with the Estado de São Paulo, he reiterated the point
that an attractive feature of the AB is that it “requires that the voter at least know how to
read comfortably, eliminating semi-illiterates from the electorate (those who do not know how
to mark a name)” (O Estado de São Paulo, May 11, 1955, p.3). Responding to critics of his
proposal, he expressed the opinion that purging the electorate of illiterate voters was a crucial
step in modernizing Brazil’s democracy and in creating a more ethical form of politics:
Those who are against the [AB ] argue that its adoption would result in a significant reduction of the electorate, principally in the interior [countryside] of the
country. This position is an admission that the interior electorate is in fact composed of a great number of illiterate and thoughtless voters, over which is most
33
easily exercised the coercion, bribery, and domination that so delegitimate electoral
results. As stated by the Estado de Sao Paulo in a recent editorial, giving these
voters the vote in such conditions is tantamount to sacrificing the modern populations (populações adiantadas) to the social norms (conveniências) of illiterates in
the interior (ibid.).
Doubling down on the notion that a purification of the electorate was required, Costa continued:
The lack of preparation of the electorate to use the [AB ] is another argument used
by those who oppose it, one which is divorced from any evidence. I am convinced
that even a schoolboy would be capable of filling it out with ease and precision...This
argument [against the AB ] is another admission that today’s electorate, at least I’d
say 60% of it, has a mental capacity inferior to that of a schoolboy. And if so,
it is impossible for things to continue as they are now, for that would place the
democratic system on a bad path (ibid.).
Costa’s remarks reflect a view of reform in which the Wigmore result would be acheived
via the Dixie path: liberate the electorate from the coercion of local bossism by taking the
ballot out of the hands of voters most easily influenced by local bosses. Political modernization
achieved by the disenfranchisement of the poorest and least educated.
These arguments appear to have been part of a more general philosophical stance against
the participation of illiterates in politics. In a text on electoral legislation written subsequent to
his tenure as TSE president, Costa argued that legally recognizing the right of illiterates to vote
would also enable them to run for elected office, thereby sullying the dignity of public office.
“Democracy is not an expression of number, but a consecration of quality,” he concluded. “And
without this qualitative character, that gives the best individuals (os melhores) the possibility
of interfering more decisively, democracy would descend into a vortex of demagoguery” (Costa
1964, p.301).
34
Costa’s strong objections to the participation of illiterates in politics were shared by other
members of Brazil’s Supreme Electoral Tribunal. In an article published in the Jornal do
Comercio on May 10, 1955, newly elected TSE judge José Duarte advocated in favor of the
Costa-authored AB reform bill largely on the grounds of its franchise-restricting properties.
“There is a class of individuals who are not able to benefit from citizenship and are not considered apt to participate in suffrage and the governing of the Nation,” he asserted. Being more
specific, Duarte continued:
It is indispensable to impede that the semi-literates, those that only know how
to sign or trace their names, persons that do not have discernment, the mentally
retarded, maladjusted, crassly ignorant, in the name of a wrongheaded concept
of democratic equality, be included among the ranks of voters who are educated,
conscious, and capable...the vote is nothing but a reflection of this sentiment of independence and the educated opinion that a citizen possesses when he is not subject
to the chains of political bossism and knows how to choose (Boletim Eleitoral, May
1955, pp.490-491).
The historical parallels between the US South and what was transpiring in Brazil were not lost
to opponents of ballot reform. Among the most vociferous critics of the proposed adoption
of the AB was Ulysses Guimarães, one of the founders of the PSD and at that time federal
deputy from the state of São Paulo. In a blistering dissent written for the Chamber of Deputies’
Commission on the Constitution and Justice, Guimarães inveighed against the AB based largely
on the argument that the AB, by complicating the mechanics of voting, would operate in effect
as a form of suffrage restriction for uneducated and/or illiterate voters. In making his case,
Guimarães explicitly compared the anticipated effects of the AB in Brazil to the operation of
Jim Crow in the American South:
The [AB] is an oblique or indirect way of violating the spirit of universal suffrage,
enshrined in article 134 of the Federal Constitution...The [AB] would remove from
35
the polls or prevent from voting—quite consciously—thousands of Brazilians legitimately enabled to do so. In the United States, similar subterfuges were invented
and legal constructions arranged in order to facilitate discrimination, notably on
the basis of color...The Southern States, in order to deprive the Negro of his right to
vote, conspired to frustrate the principle of universal and equal suffrage proclaimed
by the American Constitution and confirmed by the Supreme Court (Diário do
Congresso Nacional, August 18, 1955, p.5060).
Guimarães’ was well aware that his line of critique ran up against the criticism that, nods
to universalism in the Brazilian constitution notwithstanding, suffrage was explicitly restricted
to literates. So if the AB would restrict the franchise, it would be doing so principally for a
population that did not legally have the right to vote in the first place. With this point in mind,
he emphasized in other writings that the AB would not address the issue of illegally registered
illiterate voters by purging them from voter lists, but would rather result in these same voters
wasting their votes. That is to say, he anticipated the exact mechanism of disenfranchisement
that this paper has identified. In a letter to Ernâni do Amaral Peixoto, then governor of the
Federal District of Rio de Janeiro and PSD president, Guimarães put the point as follows:
The [AB] does not exclude or impede the poorly educated or even illiterate
voter who has been illegally registered from voting. This is because such a voter
will still turn out to vote, but he will vote incorrectly or randomly...The practical
consequence of the [AB] is that the voter knows for whom he wants to vote, but
does not know how to vote, given the complexity of the system [emphasis in original]
(CP-DOC EAP an 1955.01.27).
This particular argument about the AB was expressed repeatedly by opponents of reform in
parliamentary debates. Echoing nearly exactly the sentiment of Guimarães, Paulo Fernandes
claimed that with the AB system, the illiterate voter would still go to the polls but once he
found himself in the closed voting booth would “have to make a choice at random” (Diário
36
do Congresso Nacional II, August 19, 1955, p. 2012). Describing the likely effect of the
application of the AB to proportional representation contests, Francisco Pereira da Silva, a
PSD federal deputy from Amazonas, predicted that, given the large number of candidates
running in these contests, “there would be enormous confusion, the election results nullified,
completely vitiated...The voter [...] would not know how to vote for the person he had chosen”
(Diário do Congresso Nacional I, June 27, 1958, p.3767). In the Senate, Carlos Gomes de
Oliveira (PTB-SC) criticized the use of the AB in the proportional representation contests
on the grounds that for the “uneducated voter” (eleitor simples) the proposed system would
“generate space for confusion that could greatly alter the true will of the voter” (Diário do
Congresso Nacional II, April 15, 1959, p.584).
Even more to the point, in the days before the vote to implement the AB in proportional
contests, a long debate raged in the Chamber of Deputies in which multiple deputies cited the
large number of wasted votes in recently held majoritarian contests (under the AB ) as a reason
to delay or modify the passage of the proposed electoral law. Mario Palmério, defender of
the AB and PTB deputy from Minas Gerais, responded to these concerns in a manner wholly
consistent with that of previous proponents of reform:
Let us defend, above all, the educated voter; let us counteract the argument of
defending the uneducated voter with the argument of defending the literate voter
who is, yes, being impeded from voting by the current process...Let us worry a little
less about the bad voter. The corrupt voter, the venal voter, the voter without education, the voter who is little interested in elections and only thinks about pleasing
his political boss or ward heeler (Anais da Camara dos Deputados, July 2, 1962,
p.450-451).
In sum, the franchise-restricting properties of the AB were not a peripheral part of the
debates surrounding its adoption. Underlying these debates were fundamental disagreements
about how the Brazilian electorate should be constituted, at least in practical terms if not
according to the letter of the Constitution. Such disagreements notwithstanding, there was an
37
important point of consensus among participants in this debate. Whether or not the disenfranchisement of functional illiterates was viewed as a salutary development for Brazilian democracy, nearly all speakers were of the mindset that, were the AB to be instituted, the prospect
of effective political participation by functional illiterates would be greatly diminished.
VIII.
Conclusion
With a few important exceptions, discussions of the impact of the AB have concentrated on
the downstream implications of the degree of effective vote secrecy it provides but have largely
ignored the manner in which it affects the accessibility of the vote. This study’s examination of
Brazil’s historical experience with the AB suggests that taking accessibility issues for granted is
a mistake. The introduction of the AB created a massive increase in the proportion of wasted
votes, one that was especially acute in zones of high functional illiteracy. This was not an
unexpected or undesired byproduct of the secrecy-enhancing properties of the AB. Rather, it
was an explicitly stated, and much debated, goal of ballot reform. In its disenfranchising effects
and intent, the Brazilian experience with the AB shares much in common with that of the US
South.
Given the extant evidence on the franchise-restricting properties of the AB in the American
historical experience and that presented here for Brazil, future work on the political relevance
of the secret vote would do well to give careful consideration to how ballot reform shapes the
electorate. In two of the world’s largest democracies, the motivation behind and immediate consequence of adopting the AB –the effective secret vote–was to purge the electorate of individuals
deemed to be undesirable voters. There is thus a potential dark side to reform. Vote secrecy,
typically considered a milestone in the representation of the interests of society’s most downtrodden, may in certain contexts be implemented in such a way as to neutralize the political
participation of these very same individuals.
Whether or not this occurs often rests with the specific details of how the official ballot
38
is adopted. In the case of Brazil, the AB required voters to write in the name or candidate
number of the candidate of their preference for the proportional representation contests, a
particularly daunting task for voters with low literacy. In many other countries around the
developing world, official ballots structure vote choice through the use of party symbols, colors,
and photos, thereby placing much less strain on voters with low levels of education. It is
thus not necessarily the fact of secrecy itself that produces disenfranchisement, but rather the
particular way in which electoral engineers choose to organize the secret vote for voters given
the electoral system in place. In the Brazilian historical experience, the franchise-restricting
aims of electoral engineers and the inherent complexity of the country’s large magnitude open
list proportional representation system combined to make the secret ballot highly inaccessible
to many voters. It is difficult to interpret this as a milestone of democratic liberation. A truly
liberating ballot is one that combines the privacy of vote choice with a voting process that is
nearly universally intelligible to adult citizens in the polity undergoing reform.
39
References
Aidt, T.S. and P.S. Jensen. 2016. “From Open to Secret Ballot: Vote Buying and
Modernization.” Forthcoming in Comparative Political Studies, DOI: 0010414016628268.
Alves. A.C. 2006. “Clientelismo Eleitoral e Coronelismo Polı́tico: Estudo de un Pequeno
Municı́pio Paranaense.” In AN Codato and FJ dos Santos, orgs., Partidos e Eleições no
Paraná. Curitiba: Tribunal Regional Eleitoral do Paraná, pp. 149-169.
Ames, B. 1995. “Electoral Strategy under Open-list Proportional Representation.” American
Journal of Political Science 39 (2): 406-433.
. 2001. The Deadlock of Democracy in Brazil. Ann Arbor: University of Michigan Press.
Baland, J.-M. and J.A. Robinson. 2008. “Land and Power: Theory and Evidence from Chile.”
American Economic Review 98 (5): 1737-1765.
. 2012. “The political value of land: political reform and land prices in Chile.” American
Journal of Political Science, 56(3): 601-619..
Bensel, R.F. 2004. The American Ballot Box in the Mid-Nineteenth Century. Cambridge:
Cambridge University Press.
Bhavnani, R.R. 2009. “Do Electoral Quotas Work after They Are Withdrawn? Evidence from
a Natural Experiment in India.” American Political Science Review 103 (1): 23-36.
Blondel, J. 1957. As Condições da Vida Polı́tica no Estado da Paraı́ba. Rio de Janeiro:
Instituto de Direito Público e Ciencia Polı́tica.
Burnham, W.D. 1974. “Theory and Voting Research: Some Reflections on Converse’s ‘Change
in the American Electorate.’” American Political Science Review 68: 1002-1023.
Carey, J.M. 2009. Legislative Voting and Accountability. Cambridge: Cambridge University
Press.
and J Polga-Hecimovich. 2006. “Primary Elections and Candidate Strength in Latin
America.” Journal of Politics 68(3):530-543.
Carvalho, O.M. 1958. Ensaios de Sociologia Eleitoral. Belo Horizonte: Universidade Federal
de Minas Gerais.
Converse, P.E. 1972. “Change in the American Electorate.” In A. Campbell and P.E.
Converse, eds. The Human Meaning of Social Change. New York: Russell Sage, 263-337.
Costa, E. 1964. A Legislação Eleitoral Brasileira. Imprensa Nacional.
Crisp, BF, MC Escobar Lemmon, BS Jones, MP Jones, and MM Taylor Robinson. 2004.
“Vote-Seeking Incentives and Legislative Representation in Six Presidential Democracies.”
Journal of Politics 66 (3): 823-846.
Crowley, J. 2006. ”Uses and Abuses of the Secret Ballot in the American Age of Reform.” In
R Bertrand, JL Briquet, and P Pels, eds. The Hidden History of the Secret Ballot.
Bloomington: Indiana University Press, pp.43-68.
Gingerich, D.W., 2013. Political institutions and party-directed corruption in South America:
Stealing for the team. New York: Cambridge University Press.
Golden, M.A. and E.C.C. Chang. 2001. “Competitive Corruption: Factional Conflict and
Political Malfeasance in Postwar Italian Christian Democracy.” World Politics 53:
588-622.
40
Greenfield, S.M. 1977. “Patronage, Politics, and the Articulation of Local Community and
National Society in Pre-1968 Brazil.” Journal of Inter-American Studies and World
Affairs 19 (2): 139-172.
Gross, D.R. 1973. “Factionalism and Local Level Politics in Rural Brazil.” Journal of
Anthropological Research 29 (2): 123-144.
Heckelman, J.C. 1995. “The Effect of the Secret Ballot on Voter Turnout Rates.” Public
Choice 82: 107-124.
. 2000. ”Revisiting the Relationship between Secret Ballots and Turnout.” American
Politics Quarterly 28 (2): 194-215.
Heckman, J.J., H. Ichimura, and P. Todd. 1997. “Matching as an Econometric Evaluation
Estimator: Evidence from Evaluating a Job Training Programme.” Review of Economic
Studies 64: 605-654.
Hirano, K. and G. Imbens. 2001. “Estimation of Causal Effects using Propensity Score
Weighting: An Application to Data on Right Heart Catherterization.” Health Services and
Outcomes Research Methodology 2: 259-278.
Hirano, S., J.M. Snyder Jr, and M. Ting. 2009. “Distributive Politics with Primaries.” Journal
of Politics 71: 1467-1480.
Ho, D. and K. Imai. 2008. “Estimating Causal Effects of Ballot Order from a Randomized
Natural Experiment: The California Alphabet Lottery, 1978-2002.” Public Opinion
Quarterly 72 (2): 216-240.
Iacus, S.M., G. King, and G. Porro. 2011. “Multivariate Matching Methods That Are
Monotonic Imbalance Bounding.” Journal of the American Statistical Association 106
(493): 345-361.
1967. Instituto Brasileiro de Geografia e Estatı́stica (IBGE), Brazil. Censo Demográfico de
1960. Rio de Janeiro: IBGE.
Jones, M.P. 2009. “Gender Quotas, Electoral Laws, and the Election of Women: Evidence
From the Latin American Vanguard.” Comparative Political Studies 42 (1): 56-81.
Kam, C., 2016. “The Secret Ballot and the Market for Votes at 19th-century British
Elections. Forthcoming in Comparative Political Studies, DOI: 0010414015626451.
Kousser, J.M. 1973. “Post-Reconstruction Suffrage Restrictions in Tennessee: A New Look at
the VO Key Thesis.” Political Science Quarterly 88 (4): 655-683.
. 1974. The Shaping of Southern Politics: Suffrage Restriction and the Establishment of the
One Party South. New Haven: Yale University Press.
Kuo, D. and Teorell, J., 2016. “Illicit Tactics as Substitutes: Election Fraud, Ballot
Reform, and Contested Congressional Elections in the United States, 1860-1930.”
Forthcoming in Comparative Political Studies, DOI: 0010414016649481.
Leal, V.N. 1949. Coronelismo, Enxada e Voto. Rio de Janeiro: Revista Forense.
Lehoucq, F.E. and Molina, I., 2002. Stuffing the ballot box: fraud, electoral reform, and
democratization in Costa Rica. New York: Cambridge University Press.
Lijphart, A. and R. Lopez-Pintor. 1988. “Alphabetic Bias in Partisan Elections: Patterns of
Voting for the Spanish Senate, 1982 and 1986.” Electoral Studies 7 (3): 225-231.
Limongi, F., 2015. “Fazendo Eleitores e Eleições: Mobilização Polı́tica e Democracia no Brasil
Pós-Estado Novo.” Revista Dados, 58 (2): 27-54.
Love, J.L. 1970. “Political Participation in Brazil, 1881-1969.” Luso-Brazilian Review 7 (2):
3-24.
41
Mares, I., 2015. From open secrets to secret voting: Democratic electoral reforms and voter
autonomy. New York: Cambridge University Press.
Miller, J.M. and J.A. Krosnick. 1998. “The Impact of Candidate Name Order on Election
Outcomes.” Public Opinion Quarterly 62: 291-330.
Morgenstern, S. 2004. Patterns of Legislative Politics. Cambridge: Cambridge University
Press.
Niemi, R.G. and P.S. Herrnson. 2003. ”Beyond the Butterfly: The Complexity of U.S.
Ballots.” Perspectives on Politics 1: 317-26.
Pang, E-S. 1973. “Coronelismo in Northeast Brazil.” In R. Kern, ed. The Caciques:
Oligarchical Politics and the System of Caciquismo in the Luso-Hispanic World.
Albuqueque: University of New Mexico Press, pp.65-88.
Perman, M. 2001. Struggle for Mastery: Disfranchisement in the South, 1888-1908. Chapel
Hill: UNC Press.
Persson, D, G. Tabellini and F. Trebbi. 2003. “Electoral Rules and Corruption.” Journal of
the European Economic Association 1, 4: 958-989.
Reynolds, A. 2005. “Reserved Seats in National Legislatures: A Research Note.” Legislative
Studies Quarterly 30 (2): 301-310.
. and M. Steenbergen. 2006. “How the World Votes: The Political Consequences of Ballot
Design, Innovation, and Manipulation.” Electoral Studies 25: 570-598.
Reynolds, J.F. and R.L. McCormick. 1986. “Outlawing ‘Treachery’: Split Tickets and Ballot
Laws in New York and New Jersey, 1880-1910.” Journal of American History 72 (4):
835-858.
Rosenbaum, P.R. and D.B. Rubin. 1985. “Constructing a Control Group Using Multivariate
Matched Sampling Methods That Incorporate the Propensity Score.” American
Statistician 39: 33-38.
Rubin, D.B. 1986. “Comment: Which Ifs Have Causal Answers.” Journal of the American
Statistical Association 81 (396): 961-962.
Rusk, J.G. 1970. “The Effect of the Australian Ballot Reform on Split Ticket Voting:
1876-1908.” American Political Science Review 64: 1220-1238.
. 1974. ”Comment: The American Electoral Universe: Speculation and Evidence.”
American Political Science Review 68: 1028-1049.
Serra, G. 2011. “Why Primaries? The Party’s Trade Off Between Policy and Valence.”
Journal of Theoretical Politics 23 (1): 21-51.
Schaffer, F.C. 2002. “Might Cleaning Up Elections Keep People Away from the Polls?
Historical and Comparative Perspectives.” International Political Science Review 23 (1):
69-84.
Soares, G.A.D. 2001. A Democracia Interrompida. Rio de Janeiro: FGV.
Stokes, S.C. 2005. “Perverse Accountability: A Formal Model of Machine Politics with
Evidence from Argentina.” American Political Science Review 99: 315-325.
Vilaça, M.V. and R.C. de Albuquerque. 2003 [1965]. Coronel, Coronéis. Rio de Janeiro:
Bertrand Brasil.
Wand, J.N., KW Shotts, JS Sekhon, WR Mebane Jr, MC Herron, HE and Brady. 2001. “The
Butterfly Did It: The Aberrant Vote for Buchanan in Palm Beach County, Florida.”
American Political Science Review 95:793-810.
42
Wigmore, J.H. 1889. The Australian Ballot System as Embodied in the Legislation of Various
Countries. Boston: Boston Book Company.
43