Electoral Clientelism or Relational Clientelism?

Healthcare and Sterilization in Brazil∗

Simeon Nichter

Harvard University &

University of California, San Diego
August 30, 2011

Abstract

In many countries, clientelist parties (or political machines) distribute selec-

tive benefits in contingent exchange for political support. Studies of clientelism
should distinguish between substantively different patterns of machine politics. A
fundamental but often overlooked distinction lies between “electoral clientelism”
and what I term “relational clientelism.” Electoral clientelism delivers all pay-
offs to citizens before voting, and involves the threat of opportunistic defection
by citizens. By contrast, relational clientelism continues to deliver benefits to
citizens after voting, and involves the threat of opportunistic defection by both
citizens and politicians. Building on fieldwork and regression analysis, this study
explores these distinct patterns of clientelism in the context of healthcare and
female sterilization in Brazil.

∗
Paper prepared for presentation at the Annual Meeting of the American Political Science Association in
Seattle, Washington, September 1-4, 2011. The author expresses appreciation for support by the National
Science Foundation; the Center on Democracy, Development, and the Rule of Law at Stanford University;
the Harvard Academy for International and Area Studies, and the Center for Global Development. This
draft is preliminary and any comments would be appreciated (snichter@wcfia.harvard.edu). Any errors or
omissions are the responsibility of the author.

Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1919567
http://ssrn.com/abstract=1919567
Introduction

While many of the world’s political parties rely on policy platforms and ideological ap-
peals to attract voters, other parties offer contingent benefits in exchange for political sup-
port. Such clientelist parties (or political machines) typically focus on distributing benefits
to the poor (Scott 1969: 1150; Stokes 2005: 315, 321–2), who may be especially responsive
to material inducements in contexts where the state fails to provide a social safety net.
For many years, scholars viewed clientelism as a characteristic of traditional societies
that would evolve and eventually decline with modernization (Kitschelt & Wilkinson 2007).
Much early research on clientelism, based on anthropological case studies, emphasized the
role of traditional patterns of deference as fundamental to patron-client relations (e.g., Bois-
sevain 1966; Mair 1961). In a seminal article, James Scott (1969: 1145–7) argued that
modernization would erode such patterns of deference, leading clientelism to rely more heav-
ily on “concrete, short-run, material inducements.” Scott depicted such clientelism as only
an interim “phase,” which would eventually give way to class-based, programmatic appeals
as countries developed (Scott 1969: 1146–7). Various other studies similarly discussed the
transitional role of clientelism (e.g., Weingrod 1968: 381–5; Powell 1970: 422).
In sharp contrast with such research, many recent studies view clientelism as a political
strategy (e.g., Shefter 1994; Kitschelt 2000; Piattoni 2001). Shefter’s (1994: Ch 2) ground-
breaking work argues that parties adopt a clientelist rather than programmatic strategy
under two conditions: (1) the party “enjoyed access to patronage at the time it first un-
dertook to mobilize a popular base”; and (2) at that time, a “constituency for bureaucratic
autonomy” had not yet formed to oppose clientelism (27–28). Along a similar vein, Kitschelt
(2000) explains that political elites explicitly choose whether or not to engage in clientelism
when competing for electoral support, and examines how factors such as socioeconomic
modernization, electoral laws and ethnocultural cleavages influence this decision (see also
Kitschelt & Wilkinson 2007). Levitsky (2003), who also views clientelism from a strategic
perspective, finds that the Argentine Peronist party shifted from labor-based politics towards

Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1919567
http://ssrn.com/abstract=1919567
clientelism in order to maintain working-class and lower-class support while engaging in sub-
stantial economic reforms that were in many ways unfavorable for its traditional trade-union
base.
Although the strategic perspective marks a significant advance in scholarly research, I
argue that studies adopting this approach have an unfortunate tendency to conflate dis-
tinct strategies of clientelism. Unlike previous research that emphasizes the heterogeneity of
clientelism across time and space (e.g., Scott 1969; Lemarchand & Legg 1972), more recent
studies adopting the strategic perspective tend to be far more reductionist. Analysts should
distinguish between substantively different patterns of machine politics. With the goal of
motivating further analytical differentiation of clientelism, I first emphasize the fundamental
— but frequently overlooked — distinction between “electoral clientelism” and what I term
“relational clientelism” (Nichter 2010). The study then explores these distinct patterns of
machine politics in the context of healthcare and female sterilization in Brazil.

1 Electoral vs. Relational Clientelism

For many years, scholars focused almost entirely on relational clientelism, which involves
ongoing relationships of mutual (albeit asymmetric) support and dependence (e.g., Banfield
& Wilson 1963; Scott 1969; Powell 1970; Kitschelt 2000; Auyero 2001; Levitsky 2003). In
recent years, however, some researchers have begun to study electoral clientelism, which
involves elite payoffs to citizens exclusively during electoral campaigns (e.g., Callahan &
McCargo 1996; Hicken 2002; Stokes 2005; Lehoucq 2007; Schaffer & Schedler 2007; Dunning
& Stokes 2009; Gans-Morse et al 2010). It is crucial to emphasize and clarify the distinction
between these two patterns of machine politics. Electoral clientelism, such as vote buying
and turnout buying, delivers all benefits to citizens before voting, and involves the threat of
opportunistic defection by citizens.1 By contrast, relational clientelism continues to deliver
1
Additional forms of electoral clientelism include negative turnout buying (rewarding opposing voters for
staying at home) and double persuasion (rewarding nonvoters for showing up at the polls and voting against

Electronic copy available at: https://ssrn.com/abstract=1919567

benefits to citizens after voting, and involves the threat of opportunistic defection by both
citizens and elites.
In order to clarify the distinction between electoral and relational clientelism, it is im-
portant to examine the credibility problems involved in each phenomenon. As a first step,
Figure 1 emphasizes that unlike programmatic politics, all forms of clientelism involve issues
of citizen credibility. In contrast with programmatic politics, clientelism involves contingent
exchange in which voters promise to provide political support in exchange for selective ben-
efits (Kitschelt & Wilkinson 2007: 10; see also Kitschelt 2000: 849–50; Robinson & Verdier
2003: 1). Politicians are thus concerned about the threat of opportunistic defection by cit-
izens when engaging in clientelist strategies. But while citizen credibility is a key concern
for clientelism, it is not for programmatic politics. As Kitschelt & Wilkinson (2007: 22)
point out: “Programmatic politicians do not engage in contingent exchange and therefore
do not try to monitor and enforce conformity of voters with certain party preferences, while
clientelist politicians most definitely engage in such practices.”
In order to distinguish between electoral and relational clientelism, Figure 2 disaggregates
the top cell of Figure 1. As shown, only relational clientelism — and not electoral clientelism
— involves issues of elite credibility.2 This key difference arises because all benefits are
received before voting with electoral clientelism, whereas at least some benefits are received
after voting with relational clientelism. In the case of electoral clientelism, citizens face no
risk of opportunistic defection, because politicians deliver all benefits before the citizen votes.
Thus, citizen actions do not depend on the credibility of elite promises. By contrast, with
relational clientelism, voters look forward to assess the likelihood that a given politician
will actually follow through with promises to deliver selective benefits. Given the risk of
opportunistic defection, citizen actions depend on the credibility of elite promises.
their preferences). For further discussion and a typology, see Nichter (2008).
2
To the best of my knowledge, no other study employs issues of credibility to distinguish between strategies
of clientelism. On the other hand, numerous studies on clientelism mention issues of credibility. Most
exclusively mention citizen credibility (e.g., Stokes 2005; Kitschelt & Wilkinson 2007) or elite credibility
(e.g., Keefer 2007). However, Robinson & Verdier (2003: 1) and Finan & Schechter (2010: 1) mention the
“double” credibility problem of both citizens and elites.

Electronic copy available at: https://ssrn.com/abstract=1919567

Figure 1: Credibility Problem of Citizens: Clientelism vs. Programmatic Politics

Clientelism

Yes How Elites Distribute

Benefits Depends on
Credibility of Promises
of Support by Citizens
Does a Citizen’s
Credibility Affect
Actions of Elites? Programmatic
Politics

How Elites Distribute

No
Benefits Does Not
Depend on Credibility
of Promises of
Support by Citizens

Figure 2: Credibility Problem of Elites: Relational vs. Electoral Clientelism

Does an Elite’s Credibility

Affect Actions of Citizens?
Yes No

Relational Electoral
Clientelism Clientelism

Does a Citizen’s How Citizens Act How Citizens Act

Credibility Affect Yes Depends Does Not Depend
Actions of Elites? on Credibility of on Credibility of
Promises by Elites Promises by Elites

Electronic copy available at: https://ssrn.com/abstract=1919567

 Figure 3: Citizens and Elites: Interaction of Credibility Problems

Does an Elite’s Credibility

Affect Actions of Citizens?
Yes No

Relational Electoral
Clientelism Clientelism
Yes
Dual Citizen
Commitment Commitment
Problem Problem
Does a Citizen’s
Credibility Affect
Actions of Elites?
Prospective Retrospective
Programmatic Programmatic
No Politics Politics

Elite No
Commitment Commitment
Problem Problem

Based on this discussion of citizen and elite credibility, Figure 3 provides a typology of
commitment problems across different types of elite-citizen linkages. The row labels match
that of Figure 1 (issues of citizen credibility) and the column labels match that of Figure 2
(issues of elite credibility). Electoral clientelism involves only a citizen commitment problem.
That is, citizen credibility affects elite actions, but elite credibility does not affect citizen
actions. By contrast, relational clientelism involves a dual commitment problem. Not only
does citizen credibility affect the elite actions, but also elite credibility affects citizen actions.
While not the focus of the present study, Figure 3 also contrasts the differences of credi-
bility problems involved with different forms of clientelist and programmatic linkages. With
prospective programmatic politics, citizens vote partly on the basis of campaign promises
(e.g., Kitschelt 2000). There is thus an elite commitment problem: the credibility of elites
affects whether citizens believe they will actually follow through with such promises. With

Electronic copy available at: https://ssrn.com/abstract=1919567

retrospective programmatic politics, it is assumed that citizens rationally ignore campaign
promises (e.g., Fiorina 1978; Alesina 1988). Given that citizens make electoral decisions
solely based on past performance, retrospective programmatic politics involves no commit-
ment problem.

2 The Case of Healthcare in Brazil

Evidence on healthcare in Brazil draws further attention to the important distinction

between electoral and relational clientelism. The present study builds on 18 months of
fieldwork in Brazil. During the 2008 municipal elections, I conducted a total of 110 formal
interviews on clientelism across the state of Bahia.3 Bahia is the most populous state in
Northeast Brazil, the poorest region of the country (IBGE 2009). The formal interviews
included 55 interviews of community members and 55 interviews of elites. In addition, I
lived in a rural municipality of 10,000 citizens for five months, and conducted an additional
350 informal interviews of citizens and elites. This fieldwork focused on small municipalities,
as defined by those with 100,000 citizens or fewer. Even though 49 percent of Brazilians live
in such municipalities, most scholarly research pays relatively little attention to the role of
clientelism in small communities.
Although few studies on clientelism focus on the provision of healthcare, evidence suggests
that politicians in some Brazilian municipalities deliver medicine and health-related services
in contingent exchange for political support. Two characteristics of healthcare in Brazil
heighten its attractiveness to local political machines — discretion over public healthcare
expenditures and substantial unmet needs. Politicians at the municipal level have consid-
erable discretion over public healthcare expenditures and are thus able to employ political
criteria when allocating scarce resources. Brazil’s 1988 Constitution established a public
health system intended to be universal and comprehensive, which is commonly referred to
3
The appendix provides a more detailed description of this fieldwork. Specific details about interview
respondents are disguised to protect anonymity.

Electronic copy available at: https://ssrn.com/abstract=1919567

as SUS (Sistema Único de Saúde, or Unified Health System). SUS accounts for 46 percent
of the country’s total health expenditures. Given the country’s high level of political and
fiscal decentralization, officials at the municipal level have substantial discretion over SUS
expenditures. Whereas 91 percent of the federal health budget was centralized in 1994, the
federal share fell to only 33 percent by 2005 (Falleti 2010: 182). Meanwhile, the municipal
share increased markedly from 2 to 40 percent, providing local politicians with substantially
more discretion over resource allocation (Falleti 2010: 182).
Healthcare not only provides local machines with discretion over how resources are al-
located, but also involves high demand for benefits. Approximately 75 percent of Brazil’s
population receives health care exclusively through the public health system (Pan American
Health Organization 2008). Even though SUS was designed to be universal and compre-
hensive, many citizens remain in desperate need of medical care. Health outcomes such
as life expectancy and infant mortality vary dramatically across Brazil, and continue to be
particularly low in impoverished rural areas of the Northeast region. For example, although
infant mortality has reduced substantially in recent decades, it remains over twice as high
in the Northeast (33.2 per 1000 live births) than in the South and Southeast (15.1 and 16.6,
respectively).4 And despite significant advances in healthcare provision, there continue to be
inadequate personnel, facilities and supplies in many municipalities. For example, in Bahian
municipalities with up to 20,000 citizens, only 40 percent have ultrasound machines and
44 percent have X-ray machines (SUS 2010).5 Many citizens are particularly vulnerable to
medical emergencies, as ambulance service remains unreliable or downright unavailable in
many areas. The launch of SAMU/192, a national ambulance service, has improved access
since 2004, but the service still serves only 61 percent of Brazilians and reaches 1,502 of the
country’s 5,564 municipalities.6
4
Sı́ntese de Indicadores Sociais 2010, Instituto Brasileiro de Geografia e Estatı́stica
5
In Bahia, municipalities of this size comprise 60 percent of all municipalities and 22 percent of the state’s
overall population (IBGE 2010).
6
http://portal.saude.gov.br/portal/saude/area.cfm?id area=1787

Electronic copy available at: https://ssrn.com/abstract=1919567

2.1 Healthcare as Electoral Clientelism

Given that healthcare involves considerable local discretion and unmet needs, it provides
a valuable source of clientelist benefits in some Brazilian municipalities. With respect to
electoral clientelism, evidence suggests that some candidates deliver medical care and services
as payoffs in order to buy votes during campaigns. Operatives dole out a broad range of
contingent healthcare benefits, including medicines, health exams, dentures, wheelchairs and
orthopedic boots, as well as various non-health related goods such as building materials and
cash. The prevalence of such handouts during campaigns led over one million citizens to sign
a petition in 1999 calling for stricter legislation, resulting in the country’s first law passed
by popular initiative (Law 9840). Under this law, at least 660 politicians were prosecuted
in Brazil between 2000 and 2008 for distributing selective benefits during campaigns.7 In
a recent national survey, over 13 percent of respondents admitted voting for candidate in
exchange for a benefit.8 During my interviews, citizens often gave examples of vote buying
during campaigns, and 87 percent (47 of 54 citizens responding) reported that the practice
happens in their municipality.9
Court records from Law 9840 prosecutions provide insight about how healthcare bene-
fits are used for electoral clientelism. For example, in the Bahian municipality of Itambe,
Eleuzete Lima Santos Rodrigues was convicted of using medicines to buy votes during her
city council campaign.10 Authorities seized 124 boxes of medicines from her home, as well
as numerous citizens’ prescriptions and voter registration records. One witness testified that
somebody in the municipal health office referred her to Rodrigues — with whom she re-
portedly did not have an ongoing relationship — to obtain help in filling a prescription.
7
Movimento de Combate á Corrupção Eleitoral (http://www.mcce.org.br), 2009.
8
Survey conducted by research firm Datafolha in August 2009 included 2,133 respondents across 150
municipalities.
9
Another form of electoral clientelism, turnout buying, is less common in Brazil due to stringent enforce-
ment of compulsory voting (see Gans-Morse, Mazzuca & Nichter 2010). During interviews, only 14 percent
(7 of 51 citizens responding) reported that turnout buying occurs.
10
Information from this case provided by the state electoral court of Bahia (TRE-BA), Document Infor-
mation: Acórdão Numero 3.198/2004 (01.12.2004), Proceso 7.227.

Electronic copy available at: https://ssrn.com/abstract=1919567

Rodrigues agreed to supply the witness with three medicines (including Amoxicillin), but
explained that she was running for office and thus wanted “a little support” (uma forcinha)
in exchange. A similar vote-buying scandal resulted in the conviction of the mayor and
vice-mayor of Condado in the Northeast state of Paraiba.11 Electoral authorities raided the
mayor’s house during the 2004 campaign and discovered sacks of pills that were proven to
be distributed in exchange for votes. And in Cabedelo, another Paraiba municipality, a city
councilman was convicted for buying votes using dentures as payoffs. Lùcio Josè do Nasci-
mento Araùjo distributed signed vouchers for dental prostheses, which numerous citizens
redeemed in exchange for their votes.12
These convictions provide examples of electoral clientelism, to the extent that they ex-
clusively involve payoffs before voting, rather than ongoing benefits. Because electoral clien-
telism by definition does not involve continued benefits after the campaign, politicians are
often particularly keen to develop innovative monitoring strategies to ensure that recipients
follow through with their side of the bargain. For example, on the eve of the 2008 municipal
election, I met with several political operatives who explained that they use an unusual test
to check whether recipients of cash actually voted as instructed. Brazil has fully electronic
voting, and the polling machine shows a candidate’s picture only if selected as the vote choice.
Therefore, the operatives ask recipients what color shirt the candidate wore in the displayed
picture: “If he voted for the guy, he knows. White shirt, striped.”13 Another strategy —
lending camera phones to capture photos of vote choices — was previously used in Brazil (as
in Italy and Mexico), but electoral officials quickly caught on and banned cellular devices.14
Politicians will likely continue to adapt to develop effective monitoring mechanisms, which
are especially important with electoral clientelism.
11
Movimento de Combate á Corrupção Eleitoral (http://www.mcce.org.br), 2009.
12
Movimento de Combate á Corrupção Eleitoral (http://www.mcce.org.br), 2009.
13
Interview conducted by author in a Bahia municipality with 10,000 citizens on October 4, 2008. For
similar evidence, see also Desposato (2001).
14
TSE Resolução Numero 23.218.

10

Electronic copy available at: https://ssrn.com/abstract=1919567

2.2 Healthcare as Relational Clientelism

Although studies suggest that such electoral clientelism is prevalent in Brazil (e.g., Speck
& Abramo 2001), it is by no means the only form of machine politics. Many Brazilian
politicians engage in relational clientelism, developing long-term relationships with citizens
in which they provide ongoing selective benefits in exchange for political support. The state
often fails to provide a basic social safety net, and as a result citizens frequently rely on
relationships with politicians to meet their dire needs. For example, medicines are often
out of stock or otherwise unavailable at the public pharmacy. Many interviewees indicated
that politicians buy medicine for their supporters at private pharmacies, either using public
funds (especially mayors) or out of their own salaries (especially city councilmen). A city
councilman emphasized his provision of medical benefits to supporters when explaining: “If
you spend four years in office giving this help to a voter, then you can be more or less certain
that he’s not going to deny you in the hour that you also need him.”15
Various mechanisms facilitate the ongoing exchange relations of relational clientelism.
For instance, reciprocity often plays a particularly important role. Throughout the world,
political machines often prefer to distribute benefits to individuals who feel a “personal
obligation” to reciprocate with electoral support (Schaffer & Schedler 2007: 33; Lawson 2009;
Finan & Schechter 2010). In addition, Nichter (2009) emphasizes the strategic role of citizens
in “declared support,” a pattern of relational clientelism observed in Northeast Brazil. As
mentioned above, Brazilian politicians must overcome relatively difficult challenges when
monitoring how citizens vote, due to electronic voting. But despite ballot secrecy, the poor
often choose to signal their votes to politicians in order to obtain continued access to goods
and services. Voters can publicly declare their support before an election through actions
such as campaigning on behalf of a candidate or placing banners on their homes. Yet
declaring one’s support is risky. If a voter declares support for a candidate who wins, she
may expect to receive priority access to local government services, including medicines and
15
Interview conducted by author in a Bahia municipality with 50,000 citizens on November 13, 2008.

11

Electronic copy available at: https://ssrn.com/abstract=1919567

health exams, after the election. But if a voter declares support for a candidate who loses, she
may be disfavored by elected candidates. Thus, many voters choose to remain undeclared,
keeping silent about their preferred candidates during an electoral campaign.
Fieldwork in Bahia suggests that declared support, a form of relational clientelism, pro-
vides some citizens with ongoing preferential access to healthcare. For example, a vice-mayor
revealed that declared supporters receive priority access to the municipality’s only ambu-
lance.16 In another part of the state, a 50-year-old mason placed a large banner on the
outside of his house to “declare my vote before voting,” and therefore felt “trust” that his
politicians would reciprocate if he needed a costly medical procedure.17
However, the obvious flip side of this form of relational clientelism is that some citizens
receive disproportionately less access to scarce resources. Many interviewees complained
that citizens who declare support “incorrectly” — that is, for a losing candidate — are
discriminated against with regards to healthcare and other social services. A 32-year old
teacher explained that such a person “is persecuted, and for that reason isn’t able to get a
medical treatment in Salvador [the distant state capital].”18 Elsewhere, a 40-year old cook
indicated that citizens who declared for losing candidates would be sent away if they asked
elected officials for help obtaining medical procedures. They would be told, she explained:
“Oh, go ask for help from that person, go ask for a doctor, you were at his rally last year.”19
Numerous politicians also pointed to the dependence of healthcare on declared support. For
example, a city councilman explained that “people who are allied have preference ... you’re
on the side of who’s in power, you have greater facility, definitely. Definitely! And others
have to deal with lines ... and deal with all the difficulties ... This is the reality that we live
in here in Bahia.”20
16
Interview conducted in a municipality with 60,000 citizens on November 5, 2008.
17
Interview conducted by author in a municipality with 80,000 citizens on November 21, 2008.
18
Interview conducted by author in a municipality with 60,000 citizens on November 3, 2008.
19
Interview conducted by author in a municipality with 10,000 citizens on October 1, 2008. In another
neighborhood of the same municipality, a 37-year old restaurant worker noted that he doesn’t declare his
vote during campaigns because of the potential consequences: “if he falls sick, he goes to look for him [a
politician] and he will say ‘you didn’t vote for me’” (interview conducted on October 16, 2008).
20
Interview conducted by author in a municipality with 90,000 citizens on December 18, 2008.

12

Electronic copy available at: https://ssrn.com/abstract=1919567

 The maintenance of ongoing patterns of preferential access and political discrimination
depend on extensive social networks. As such, relational clientelism is likely to be more
prevalent in contexts where parties or candidates have invested considerable time and re-
sources to become socially proximate to citizens.21 For example, interviewees suggest that
operatives working at admission desks in local clinics serve as gatekeepers, and thus play
a central role in providing preferential access to declared supporters. In fact, some respon-
dents suggest that the political discrimination they face is entirely unbeknownst to doctors
and specialists, who are often transferred temporarily from distant cities and not actively
involved in local politics. A 24-year woman explained that “just because you’re from the
other party, sometimes receptionists — when we go with a child or otherwise seek a doctor
— invent things, saying the doctor isn’t there.”22 Similarly, a 59-year old maid complained
that these employees often “spite” citizens who supported opposition candidates, leaving
them “without treatment, they say the doctor’s not there — that the doctor left — while
the doctor is there. It’s not the doctor’s fault because he’s in his office and doesn’t know
what’s going on.”23 In effect, relational clientelism in the context of healthcare is particularly
dependent on party embeddedness through social networks, but does not necessarily rely on
the involvement of all medical professionals.

2.3 Healthcare, Clientelism and Credibility

Overall, these healthcare examples from Brazil underscore the need to distinguish between
different forms of machine politics. On the one hand, politicians engage in a particular form
of electoral clientelism — vote buying — when they provide medicines and health services
before the election as payment in exchange for vote choices. Citizens who receive such payoffs
are not embedded in ongoing relationships with the politicians who deliver them, in that part
of the bargain is not continued access to goods and services beyond election campaigns. This
21
For more on how the role of social proximity and party organizational structures affect the use of
clientelism more generally, see Kitschelt & Rozenas (2011) and Stokes (2009).
22
Interview conducted by author in a municipality with 10,000 citizens on October 16, 2008.
23
Interview conducted by author in a municipality with 10,000 citizens on October 2, 2008.

13

Electronic copy available at: https://ssrn.com/abstract=1919567

pattern of electoral clientelism contrasts sharply with another form of particularism observed
in Brazil — the ongoing contingency of healthcare benefits through relational clientelism.
Our discussion of healthcare examples in Brazil provides additional insight about the role
of credibility issues in clientelism. Regardless whether politicians use healthcare benefits to
engage in electoral or relational clientelism, they face uncertainty that citizens will actually
follow through with their promises of political support. But each form of clientelism tends to
rely on different ways of alleviating such uncertainty. On the one hand, political operatives
buying votes on Election Day look towards monitoring mechanisms such as testing recipients
about shirt colors to ensure they follow through with their end of the bargain. With relational
clientelism, such monitoring may at times also be employed, but politicians indicate that
long-term relationships help to ameliorate issues of credibility. As suggested by the city
councilman’s quote above, politicians tend to believe that a citizen receiving ongoing help
will reciprocate by providing electoral support. A mayor further elaborated that such a
citizen “stays loyal” because when “you make a real bond of trust with a certain candidate
— city councilman or mayor — you think he can solve all of your problems.”24
While qualitative evidence provides insights about how citizen credibility issues affect
both forms of clientelism, it also confirms the intuition that only relational clientelism in-
volves elite credibility issues (see Figure 2). Citizens involved in vote-buying transactions (a
form of electoral clientelism) receive their pills, wheelchairs or dentures during the campaign,
and thus are not concerned with elite credibility. By contrast, citizens pay close attention to
elite credibility in the context of relational clientelism, often evaluating past relationships to
gauge whether politicians will follow through on their promises of future support. Candidates
with a proven track record of providing assistance are deemed far more likely to help in the
future. For example, 82 percent of citizens (41 of 50 responding) indicated that they would
vote for a candidate who had helped them in small ways over a longer period, rather than
a candidate who helped them substantially during a campaign. The latter candidate was
24
Interview conducted by author in a Bahia municipality with 30,000 citizens on December 1, 2008.

14

Electronic copy available at: https://ssrn.com/abstract=1919567

viewed as an unreliable source of future assistance. A citizen role-played how this candidate
would respond if asked for help once elected: “[The politician] places his hand on his pocket
and says: ‘I already paid.’ — [Citizen:] ‘But I voted for you!’ — [Politician:] ‘Ah, but
I already paid for that vote!’”25 Elite credibility is thus a pressing concern with relational
clientelism, but not with electoral clientelism.

3 Sterilization: Electoral or Relational Clientelism?

Thus far, the present study has argued that increased analytical differentiation of clien-
telism is crucial. Electoral and relational clientelism are distinct and often coexisting phe-
nomena, in which politicians engage in different patterns of distributing contingent benefits.
In order to emphasize the importance of making this distinction, we now turn to the case of
female sterilization in Brazil. Regressions of panel data provide evidence that the incidence
of tubal ligations (i.e., female sterilization) increases during election years, and fieldwork
suggests that politicians use these surgeries as a form of health patronage. Findings are con-
sistent with the clientelist distribution of healthcare benefits discussed in Section 2. However,
a drawback of the analysis of sterilization below is that like much recent scholarly work, it
fails to clarify whether politicians provide sterilization as a form of electoral clientelism, rela-
tional clientelism, or both. Additionally, as is discussed below, it remains crucial to ascertain
whether increased sterilization during campaigns actually constitutes clientelism at all, or
if instead more tubal ligations are performed during electoral periods simply to generate
goodwill among voters in a non-contingent manner.

3.1 Profile of Female Sterilization

Across the world, female sterilization is the most frequently used form of birth control
(see Figure 4a). Over 20 percent of married women of reproductive age in the world are
25
Interview conducted by author in a Bahia municipality with 80,000 citizens on November 21, 2008.

15

Electronic copy available at: https://ssrn.com/abstract=1919567

 Figure 4: Female Sterilization vs. Other Contraception Methods (World)
(a) (b)

Contraceptive Use, World in 2007 Contraceptive Use,

(% of Married Women of Reproductive Age) by Level of Development

4.4
Condom
16.1
Other Modern
Method, 7%
15.1
Condom, 6% Not Using IUD
a Method, 37% 9.1

Pill, 9%
7.2
Pill
18.1

Female 22.3
Female Sterilization 8.1
Sterilization,
20% Any Traditional
Method, 7% Any Modern 55.7
IUD, 14%
Method 58.6

Less Developed Countries

More Developed Countries

Source: UN Population Division, World Contraceptive Use 2009.

sterilized, compared to only 14 percent using IUDs and 9 percent using the pill (the next
two leading contraceptive methods). Yet there is considerable heterogeneity in the profile
of usage patterns. As shown in Figure 4b, although the usage rate of modern contraceptive
methods is comparable across level of development, the types of methods employed vary
dramatically. Women in richer countries tend to use temporary forms of birth control such
as condoms and the pill, whereas women in poorer countries tend to use more permanent
methods such as sterilization and IUDs.
A 2009 study by the United Nations Population Division suggests that Brazil has the
second-highest rate of female sterilization in the world, following the Dominican Republic (see
Figure 5a). Although data on sterilization rates fluctuate, with more recent data suggesting
lower prevalence in Brazil, the country’s sterilization rate remains high given its level of

16

Electronic copy available at: https://ssrn.com/abstract=1919567

 Figure 5: Female Sterilization (Brazil and World)
(a) (b)

Top 5 Countries in Use of

Female Sterilization Contraceptive Use in Brazil
(2009 UN Study) (% of Married Women of Reproductive Age)
47.4%
Method 1986 1996 2006
40.1%
38.5%
Any Method 66.2 76.7 80.6
37.3%
33.1%
Tubal Ligation 26.8 40.1 29.1
Pill 25.7 22.0 28.7
Condom 1.7 4.4 12.2
Vasectomy 0.8 2.6 5.1
IUD 1.0 1.1 1.9
Traditional 9.0 6.1 3.2
Other 1.2 0.4 0.4
Dominican Brazil 1996 Puerto Rico India China 2001
Republic 2002 2005/06
2007

Source: UN Population Division, World Contraceptive Use 2009; Amorim & Bonifacio (2010)
based on DHS data.

development.26 The data show that tubal ligation is the leading form of birth control in
Brazil, though the extent to which it leads the next most common method, the pill, varies
substantially across the years.27 Research suggests that one factor inhibiting greater pill
usage is an inconsistent supply of birth control kits, with 28 percent of municipalities in a
recent survey complaining that the Ministry of Health was “slow to respond to demand”
(Osis et al 2006).
Male sterilization through vasectomies, though less invasive than female sterilization and
not requiring general anesthesia, has traditionally been far less common in Brazil partly due
to machismo and fear of inhibiting sexual performance. However, this pattern is changing:
whereas there were 7.8 tubal ligations performed for every vasectomy in 1999, within a
26
The 2009 UN ranking employed data from Brazil’s 1996 Demographic and Health Survey (DHS), but as
shown in Figure 5b, more recent DHS data from 2006 show a substantial decline in sterilization rates.
27
It should be noted that Caetano (2010) argues that the apparent recent decline in sterilization rates is
actually just a statistical artifact of the DHS sample’s specific age composition.

17

Electronic copy available at: https://ssrn.com/abstract=1919567

 Figure 6: Female Sterilization and Fertility Across Brazil
(a) (b)

Total Fertility Rate (2000)

Female Sterilization in Brazil, by Region Across Municipalities
(% of Married Women of Reproductive Age)

Method 1986 1996 2006

North - 51.3 41.0
Northeast 24.6 43.9 36.9
Southeast 29.9 37.8 23.9
South 18.1 29.0 19.2
Center-West 42.0 59.5 38.9
Total 26.8 40.1 29.1

Source: Amorim & Bonifacio (2010) based on DHS data; Atlas do Desenvolvimento Humano
no Brasil.

decade this ratio fell to 1.8 (61,874 tubal ligations vs. 35,015 vasectomies in 2008) (SUS
2010). Condom use continues to increase sharply (1.7 percent in 1986 to 12.2 percent in
2006), but consistent usage is relatively expensive for the poor even though millions are
distributed for free at specific events such as Carnaval.28 Tubal ligations are most prevalent
in poorer regions, which also tend to have greater fertility rates (see Figures 6a and 6b).
Over the past 15 years, sterilization in Brazil has transformed from an illegal (albeit
prevalent) method to a publicly funded method of contraception. Until 1997, tubal ligations
and vasectomies were illegal and banned by medical ethics code without“precise indication
approved by two medical doctors.” Though illegal, sterilizations were not criminalized, and
doctors commonly employed successive cesareans as a rationale for performing tubal ligations
28
For example, 1.6 million condoms were distributed during Carnaval in Salvador in 2011.
(http://www.teiadenoticias.com.br/noticia/cidade/prefeitura-distribui-1-6-milh-o-de-preservativos-no-
carnaval)

18

Electronic copy available at: https://ssrn.com/abstract=1919567

(Caetano & Potter 2004). Conducting tubal ligations during cesarean births also helped to
cover costs, as women would already be under general anesthesia even though public funds
could not be directly expended on sterilization (Caetano 2000). In a radical departure
from previous policy, and after substantial effort from civil society organizations, Law 9263
legalized and provided public funding for both female and male sterilization. Recipients
must be at least 25 years old or already have two living children, and must undergo a 60-
day waiting period with required consultations.29 After the passing of Law 9263, doctors
performed 452,413 publicly funded tubal ligations between 1998 and 2010 (SUS 2010). The
number of female sterilizations performed by SUS is expanding rapidly, and by 2006 nearly 64
percent of all tubal ligations in Brazil were performed by the public health system. Despite
this expansion, demand for publicly funded tubal ligations remains high.30 There are an
estimated one million illegal abortions annually in Brazil,31 and women most frequently
cite their economic situation when asked why they seek sterilization (Caetano 2000: 165).
Given scarce resources, women must endure long waiting periods (3 to 18 months) and many
reportedly give up without receiving desired tubal ligations.32

3.2 Female Sterilization and Politics

Evidence suggests a link between female sterilization and politics. In a 2011 newspaper
article, ex-federal deputy Raul Jungmann (from Pernambuco in Northeast Brazil) explains
that many politicians “have highly vulnerable voters, who depend on clientelist actions ...
In general it’s an individualized good: I give you a tubal ligation, a job, and you give me a
vote.”33 In addition, an excellent study by Caetano & Potter (2004) argues that politicians in
Northeast Brazil offer sterilizations to poor women as a form of patronage. The study does
29
As of 2001, 74 countries similarly have laws explicitly allowing sterilization as a contraceptive meth-
ods, of which 24 have age or parity requirements. See: http://www.engenderhealth.org/files/pubs/family-
planning/factbook contents.pdf
30
Demand for contraception has grown in Brazil, contributing to a rapid decline in fertility rates in recent
decades (the total fertility rate decreased from 6.2 births per woman in 1960 to 1.8 in 2009).
31
http://oglobo.globo.com/pais/mat/2007/05/30/295957896.asp
32
http://www.cemicamp.org.br/noticia 0022.htm; http://jornalsg.com.br
33
“Transparência é mau negócio para deputados governistas,” O Estado de São Paulo, January 30, 2011.

19

Electronic copy available at: https://ssrn.com/abstract=1919567

 Figure 7: Electoral Cycles of Female Sterilization (from 1996 DHS Sample)

Tubal
Ligations

Source: Reproduced from Caetano & Potter (2004), based on DHS 1996 Data. Numbers are
frequencies of sample of 2,012 women who reported sterilizations in years shown, adjusted
for sampling weights.

not present econometric evidence but provides intriguing descriptive and qualitative analyses
in support of its argument. Using Demographic and Health Survey (DHS) data, Caetano &
Potter find electoral cycles in the incidence of female sterilization. As shown in Figure 7,
more women reported receiving tubal ligations in election years than in non-election years.
While these descriptive data are only suggestive, and do not control for potential confounding
variables, the electoral pattern deserves further analysis and is corroborated below.
Moreover, Caetano (2000) provides qualitative evidence of the link between politics and
tubal ligations. In Pernambuco, an ex-mayor operating as a surgeon provided this frank
assessment: “I don’t know if it elects a candidate, but it makes a big difference. For instance,

20

Electronic copy available at: https://ssrn.com/abstract=1919567

a candidate for the city council here is elected with 1,000 votes. If you are a physician and
you perform 300 tubal ligations, don’t you have chance of being elected? Because she votes,
her husband votes, the grown children vote ... You can get elected” (Caetano 2000: 230).
Citizens interviewed also indicated a greater supply of tubal ligations during campaigns.
As one woman reported, “it was election time and during these periods many doctors and
politicians do this and that for free. Tubal ligations, especially, were at the peak! The day
I went to have my surgery there were about 12 women who would have tubal ligations”
(Caetano 2000: 191). Furthermore, in survey of 281 sterilized women in four municipalities,
nearly 32 percent of respondents reported that politically affiliated doctors arranged or paid
for their sterilizations, while another 9 percent directly attributed politicians (Caetano 2000:
204).34
Building on such findings, the present study investigates the relationship between elec-
tions and female sterilization. Regressions in the present paper analyze a more recent,
extensive dataset and similarly provide evidence of electoral cycles in female sterilization.
As shown in Figure 8, using monthly SUS data on tubal ligations between 1998 and 2010,
there appears to be pattern of increased surgeries during election years. Specifications below
provide more rigorous evidence of an electoral cycle in tubal ligations, employing a 13-year
panel dataset of all 5,564 Brazilian municipalities.

3.3 Data

3.3.1 Dependent Variable

Regressions employ municipal-level data on female sterilizations from the Unified Health
System (Sistema Único de Saúde, or SUS) of Brazil’s Ministry of Health. The data consist of
the number of tubal ligations per year funded by the public health system in each Brazilian
municipality between 1998 and 2010. The dataset provides a large sample, covering 452,413
34
Note that this survey was taken only shortly after the public funding of sterilization became legal through
SUS in 1997, so responses might well be different if surveying women today.

21

Electronic copy available at: https://ssrn.com/abstract=1919567

 Figure 8: Publicly Funded Female Sterilizations in Brazil (Monthly, 1998-2010)
Publicly Funded Female Sterilization in Brazil, 1998-2010

Municipal
8000 Election

National
National Election
Election
6000
Municipal
Tubal Election
Ligations
Per
Month National
by
4000 Election
SUS
Municipal
Election
2000 National
Election

0
Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan-
98 99 00 01 02 03 04 05 06 07 08 09 10

Source: Datasus, Procedimentos hospitalares por local de residência (internações).

publicly funded tubal ligations across Brazil’s 5,564 municipalities. More specifically, the
data reflect the number of tubal ligations received by women who reside in each municipality,
regardless of where the surgeries were actually performed. Thus, specifications capture the
incidence of female sterilization in each municipality, even though patients in small or rural
municipalities must often travel to other localities for advanced medical care. The data
employed are the number of Hospital Admission Authorizations (Autorizações de Internação
Hospitalar, or AIHs) for tubal ligations. AIHs are important documents that hospitals must
file when admitting patients in order to obtain reimbursement from the public health system.

3.3.2 Independent Variables

Given that specifications explore the relationship between female sterilization and the
electoral cycle, the primary independent variable of interest is whether or not an election

22

Electronic copy available at: https://ssrn.com/abstract=1919567

occurred in a given year. During the 13 years spanned by the dataset, Brazil held seven
elections. Legislative and presidential elections occurred in 1998, 2002, 2006 and 2010,
while municipal elections occurred in 2000, 2004 and 2008. To explore the possibility that
the incidence of sterilization varies across each type of elections, categorical variables were
employed for both national and municipal elections.
Various controls come from the SUS database of the Ministry of Health. Annual data on
the following SUS variables for each municipality are employed: total births, hospital births,
births by cesarean section, births by women 30 years or older, and births by single women. In
addition, SUS data are used to control for the total number of AIHs (i.e., hospital admission
authorizations for all medical procedures) for residents in each municipality. Annual data
on population and GDP per capita are from the Institute for Applied Economic Research
(Instituto de Pesquisa Economica Aplicada, IPEA), a Brazilian government agency. Data
on inequality in each municipality (measured by the Theil index) are also provided by IPEA,
based on the 2000 national census.
Numerous controls come directly from the 2000 national census, which is administered
by the Instituto Brasileiro de Geografia e Estatı́stica (IBGE). These controls include the per-
centage of each municipality’s population that is women of reproductive age (15-49 years).
Specific characteristics about women of reproductive age in each municipality are also em-
ployed: the percentage married, economically active, illiterate, and with children. In addi-
tion, given the potential role of religion in decisions about sterilization, census data on the
share of the population that is Catholic are also included. Overall descriptive statistics for
these variables are provided in Table 1.

3.3.3 Analysis

To estimate the relationship between tubal ligations and elections, the following specifi-
cation is first used:

23

Electronic copy available at: https://ssrn.com/abstract=1919567

 Table 1: Descriptive Statistics

Variable N Mean SD
Tubal Ligations 72,696 4.245 43.25
Population 72,150 32,326.14 192,677.80
GDP per Capita (R$ 2000) 55,510 4,630.65 5,308.40
Inequality (Theil, in 2000) 71,591 0.521 0.109
% Catholic (in 2000) 71,591 0.819 0.117
% Female 15-49 years (in 2000) 71,591 0.251 0.021
% w/ Kids (in 2000) 71,591 0.667 0.050
% Married (in 2000) 71,591 0.441 0.114
% Working (in 2000) 71,591 0.508 0.130
% Illiterate (in 2000) 71,370 0.121 0.089
Crude Birth Rate 66,586 15.90 5.238
% of Births in Hospitals 66,539 0.951 0.131
% of Births by C-Section 66,539 0.390 0.187
% of Births by Women 30yrs+ 66,536 0.219 0.083
Births 66,779 546.94 3,277.25
Births in Hospitals 66,779 530.11 3,248.79
Births by C-Section 66,779 229.08 1,588.10
Births by Women 30yrs+ 66,776 129.43 990.91
Births by Single Women 55,648 269.78 1,676.19
Hospital Authorizations 72,079 2,145.64 10,233.86

Note: Each observation refers to a municipality-year. Percentages in this table are expressed
in terms of proportions.

lnLit = φ1 Mt + φ2 Nt + Xit β + φ3 t + γi + ǫit (1)

where Lit is the number of tubal ligations in municipality i during year t, Mt is a dummy
variable equal to 1 in municipal election years and 0 otherwise; Nt is a dummy variable
equal to 1 in national election years and 0 otherwise; and Xit is a matrix of the covariates
described above. The number of tubal ligations, GDP per capita and population are logged.
State fixed effects γi are introduced to control for state-level factors that are assumed to be
time-invariant. In addition, a time trend variable (t) is employed to control for temporal
factors, such as the sharp overall increase in sterilizations over the last decade. It should
be noted that additional specifications (discussed below) employ year and municipal fixed

24

Electronic copy available at: https://ssrn.com/abstract=1919567

 Table 2: OLS Estimations: Female Sterilization & Elections

(1) (2) (3) (4) (5) (6)

ln Tubal ln Tubal ln Tubal ln Tubal ln Tubal ln Tubal
Municipal Election 0.108∗∗ 0.107∗∗ 0.107∗∗ 0.100∗∗ 0.101∗∗ 0.104∗∗
(0.0346) (0.0346) (0.0345) (0.0323) (0.0326) (0.0318)
National Election 0.0527+ 0.0514+ 0.0520+ 0.0730∗ 0.0722∗ 0.0672+
(0.0267) (0.0268) (0.0270) (0.0358) (0.0360) (0.0360)
ln Population 0.742∗∗∗ 0.723∗∗∗ 0.813∗∗∗ 0.822∗∗∗ 0.762∗∗∗ 0.735∗∗∗
(0.0968) (0.0958) (0.0766) (0.0819) (0.0857) (0.0723)
ln GDP per Capita 0.0808 0.0409 0.0811
(0.0834) (0.0937) (0.0961)
Inequality -1.002∗∗ -0.919∗∗ -0.851∗
(0.331) (0.343) (0.341)
% Catholic -1.279∗ -1.346∗∗
(0.532) (0.522)
% Female 15-49 Yrs 0.318 6.114∗
(3.215) (3.117)
% w/ Kids -0.419 0.900
(1.408) (1.175)
% Married -0.932∗ -0.829∗
(0.436) (0.422)
% Working -0.248 -0.341
(0.222) (0.215)
% Illiterate -1.279∗ -1.182∗
(0.546) (0.527)
Crude Birth Rate -0.0241∗∗∗
(0.00648)
% Births: Hospital 0.0683
(0.189)
% Births: Csec -1.397∗∗∗
(0.356)
% Births: 30+ yrs 2.257∗∗∗
(0.602)
Year 0.173∗∗∗ 0.172∗∗∗ 0.177∗∗∗ 0.179∗∗∗ 0.187∗∗∗
(0.0264) (0.0265) (0.0312) (0.0310) (0.0330)
Constant -10.55∗∗∗ -357.8∗∗∗ -356.7∗∗∗ -364.7∗∗∗ -366.3∗∗∗ -383.9∗∗∗
(0.782) (53.35) (53.66) (63.24) (62.54) (66.91)
State Fixed Effects NO NO YES YES YES YES
N 72150 72150 72150 55070 54900 54894
R2 0.118 0.188 0.283 0.273 0.277 0.285
Standard errors shown in parentheses and clustered at the state level.
+ p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

effects, which provide additional rigor but prevent usage of some controls. Standard errors
are clustered at the state level.
Table 2 shows regression estimates for several variants of Equation 1. Column (1) includes

25

Electronic copy available at: https://ssrn.com/abstract=1919567

only logged population as a control; Column (2) introduces a time trend, Column (3) adds
state fixed effects; Column (4) introduces economic controls (logged GDP per capita and
inequality); Column (5) adds demographic controls; and Column (6) introduces controls
related to births. Across all specifications, municipal election years are associated with a 10
to 11 percent increase in tubal ligations. This estimate is significant at the 99 percent level
of confidence in all specifications. National elections are also associated with an increase in
female sterilization, but the effect is smaller (ranging from a 5 to 7 percent increase) and
significant at the 90 to 95 percent level.
In order to reduce potential endogeneity, first-differenced specifications were also exam-
ined using the following model:

∆lnLit = φ1 Mt + φ2 Nt + Xit β + τt + µi + ǫit (2)

where ∆lnLit is the change in the (logged) number of tubal ligations in municipality i
between years t − 1 and t, Mt is a dummy variable equal to 1 in municipal election years and
0 otherwise; Nt is a dummy variable equal to 1 in national election years and 0 otherwise;
Xit is a matrix of first-differenced covariates; τt reflects year fixed effects; and µi represents
municipal fixed effects. Standard errors are clustered at the municipal level.
Table 3 shows regression estimates for several variants of Equation 2. Column (1) presents
the base model with election dummies; Column (2) introduces year fixed effects; and Col-
umn (3) adds municipal fixed effects and population indicators.35 Across these specifications,
municipal election years are associated with an 18 to 27 percent increase in tubal ligations,
whereas national election years are associated with a 14 percent increase in female steril-
ization. These estimates are significant at the 99.9 percent level of confidence. Column (4)
presents a specification with numerous first-differenced covariates and year fixed effects, to
which Column (5) adds municipal fixed effects and population indicators. In these two spec-
ifications, municipal election years are associated with an 11 to 13 percent increase in female
35
Population indicators are dummies constructed for each quartile of municipal population size.

26

Electronic copy available at: https://ssrn.com/abstract=1919567

 Table 3: OLS Estimations: Female Sterilization & Elections (First-Differences)

(1) (2) (3) (4) (5)

∆ ln Tubal ∆ ln Tubal ∆ ln Tubal ∆ ln Tubal ∆ ln Tubal
Municipal Election 0.183∗∗∗ 0.267∗∗∗ 0.195∗∗∗ 0.132∗∗∗ 0.112∗∗
(0.0199) (0.0276) (0.0248) (0.0359) (0.0363)
National Election 0.141∗∗∗ 0.143∗∗∗ 0.139∗∗∗ 0.0775∗ 0.150∗∗∗
(0.0205) (0.0292) (0.0309) (0.0346) (0.0346)
∆ ln GDP per Capita -0.175∗∗ -0.140∗
(0.0539) (0.0608)
∆ ln Hospital Authorizations 0.204∗∗∗ 0.217∗∗∗
(0.0305) (0.0354)
∆ ln Births -0.0955+ -0.0872
(0.0558) (0.0617)
∆ ln Hospital Births 0.0350 0.0396
(0.0256) (0.0280)
∆ ln Csec Births 0.0323 0.0419+
(0.0211) (0.0236)
∆ ln Births: 30+ yrs -0.0179 -0.0230
(0.0258) (0.0281)
∆ ln Births: Single Mom -0.00569 -0.0128
(0.0148) (0.0163)
Constant 0.0902∗∗∗ -1.58e-12 -0.0842∗ 0.159∗∗∗ 0.138+
(0.00870) (0.00000) (0.0388) (0.0221) (0.0782)
Year Fixed Effects NO YES YES YES YES
Municipal Fixed Effects NO NO YES NO YES
Population Indicators NO NO YES NO YES
N 67104 67104 66643 44374 44374
R2 0.002 0.004 0.026 0.003 0.045
Standard errors shown in parentheses and clustered at the municipal level.
+ p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

27

Electronic copy available at: https://ssrn.com/abstract=1919567

sterilization (significant at the 99 percent to 99.9 percent level), whereas national election
years are associated with an 8 to 15 percent increase in tubal ligations (significant at the 95
to 99 percent level).
Regressions also explore whether mayoral reelection incentives affect the relationship
between female sterilization and the electoral cycle. Since a 1998 constitutional amendment,
incumbent mayors in Brazil are permitted to stand for reelection for one additional term.
Employing data from the Supreme Electoral Court (Tribunal Superior Eleitoral, or TSE).
we construct a dummy variable Dit equal to 1 if the mayor is a “lame duck” due to previous
reelection, and 0 if the mayor is eligible for reelection. In addition, Dit (lame duck dummy)
is interacted with Mt (municipal election dummy) to indicate instances in which a mayor is
ineligible for reelection during a municipal election year. Adding these two terms to Equation
2 yields the following model for OLS estimation:

∆lnLit = φ1 Mt + φ2 Nt + φ3 Dit + φ4 Mt ∗ Dit + Xit β + τt + µi + ǫit (3)

Table 4 shows findings for several variations of Equation 3. Column (1) presents the base
model with dummies for election years and lame duck mayors; Column (2) introduces year
fixed effects; and Column (3) adds municipal fixed effects and population indicators. The
relationship between the electoral cycle and female sterilization continues to be significant at
the 99.9 percent level of significance: municipal election years are associated with an 18 to
36 percent increase in tubal ligations, while national election years are associated with a 14
to 33 percent increase. These specifications suggest a moderate effect of mayoral reelection
incentives — the presence of a lame duck mayor is associated with an approximately 3 percent
decrease in female sterilization (significant at the 95 percent level). With the inclusion of
numerous first-differenced covariates in Column (4), the significance level of the lame duck
dummy falls to 90 percent. Column (5) includes the interaction between the lame duck and
municipal election dummies, and the coefficient on this interaction term is negative but not
statistically significant. An F-test suggests that this interaction term and the lame duck

28

Electronic copy available at: https://ssrn.com/abstract=1919567

dummy are jointly significant at the 90 percent level. Findings of an electoral cycle in female
sterilization remain significant at the 99.9 percent level across all specifications. In the most
inclusive specifications in Table 4 (Column 4 and 5), municipal elections are associated with
an 18 percent increase in tubal ligations, and national elections are associated with a 14
percent increase.
Altogether, the specifications in Tables 2-4 provide evidence that female sterilizations
increase significantly during elections in Brazil. Such findings are consistent with qualitative
evidence presented above suggesting that politicians deliver healthcare benefits in contingent
exchange for political support. However, this quantitative analysis, like many recent studies
of clientelism, does not provide insight as to whether tubal ligations constitute electoral
clientelism (e.g., vote buying) or ongoing patterns of relational clientelism (e.g., declared
support). To distinguish between these different phenomena, it is necessary to conduct
interviews of both citizens and politicians to probe the underlying logic of the exchange
relations observed. In addition, it would important to conduct quantitative analyses of the
broader range of benefits that citizens receive over time (e.g., using individual-level panel
data on all social services received by 2000 citizens over 10 years).
Another important issue that deserves emphasis is the need to examine whether politi-
cians are in fact providing surgeries contingent on political support. Because if increased
sterilization is provided without strings attached in an effort to boost goodwill among vot-
ers, then it does not entail the contingency that is central to the definition of clientelism.
Numerous politicians commented that incumbent politicians boost healthcare temporarily
during elections. For example, a city councilman explained: “Generally, the mayor, during
election time, does many programs for the community — assistance for everyone who needs
36
healthcare.” Likewise, a party leader explained that “healthcare, when it’s election time,
always has better service. Sometimes they place more doctors, they treat people better,
for fear of losing votes.”37 Citizens also observed a pattern of increased health services, in-
36
Interview conducted by author in a municipality with 10,000 citizens on November 26, 2008.
37
Interview conducted by author in a municipality with 30,000 citizens on December 4, 2008.

29

Electronic copy available at: https://ssrn.com/abstract=1919567

 Table 4: OLS Estimations: Female Sterilization & Elections (First-Differences/Mayor)

(1) (2) (3) (4) (5)

∆ ln Tubal ∆ ln Tubal ∆ ln Tubal ∆ ln Tubal ∆ ln Tubal
Municipal Election 0.182∗∗∗ 0.357∗∗∗ 0.208∗∗∗ 0.176∗∗∗ 0.183∗∗∗
(0.0200) (0.0203) (0.0256) (0.0327) (0.0329)
National Election 0.142∗∗∗ 0.326∗∗∗ 0.324∗∗∗ 0.137∗∗∗ 0.139∗∗∗
(0.0207) (0.0266) (0.0279) (0.0343) (0.0344)
Lame Duck Mayor -0.0304∗ -0.0269∗ -0.0362∗ -0.0333+ -0.0134
(0.0124) (0.0132) (0.0160) (0.0177) (0.0192)
Lame Duck*Mun Elec -0.0791
(0.0492)
∆ ln GDP per Capita -0.106+ -0.105+
(0.0566) (0.0566)
∆ ln Hospital Authorizations 0.166∗∗∗ 0.166∗∗∗
(0.0306) (0.0306)
∆ ln Births -0.0998∗ -0.0995∗
(0.0431) (0.0431)
∆ ln Hospital Births 0.0365+ 0.0359
(0.0221) (0.0221)
∆ ln Csec Births 0.0275 0.0277
(0.0169) (0.0169)
∆ ln Births: 30+ yrs -0.0137 -0.0136
(0.0228) (0.0228)
Constant 0.0993∗∗∗ -1.55e-13 -0.0908∗ 0.0742 0.0676
(0.00912) (1.38e-08) (0.0385) (0.0639) (0.0641)
Year Fixed Effects NO YES YES YES YES
Municipal Fixed Effects NO NO YES YES YES
Population Indicators NO NO YES YES YES
N 66536 66536 66410 49663 49663
R2 0.002 0.004 0.026 0.043 0.043
Standard errors shown in parentheses and clustered at the municipal level.
+ p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

30

Electronic copy available at: https://ssrn.com/abstract=1919567

cluding transportation to distant cities for treatment, during campaigns. As one interviewee
explained: “Everything that we want, during the election, we get ... If you need a doctor,
if you need transportation, they give it. Now, after the elections, they suspend all the cars
they had.”38 The important task of evaluating the contingency of benefits can be complex,
because both clientelist and programmatic strategies often simultaneously entail ramping up
the delivery of output during campaigns.

4 Discussion

The present study emphasizes that a fundamental yet frequently overlooked distinction
lies between electoral and relational clientelism. Whereas electoral clientelism provides pay-
offs during campaigns, relational clientelism provides ongoing access to goods and services.
These two forms of clientelism involve different risks of opportunistic defection. Electoral
clientelism only involves a citizen credibility problem: citizens receive all benefits before vot-
ing, and politicians are unsure whether recipients will comply and deliver electoral support.
By contrast, relational clientelism involves citizen and elite credibility problems: citizens are
uncertain whether politicians will deliver future selective benefits, and politicians are unsure
whether recipients will provide electoral support.
Evidence on healthcare suggests that both forms of clientelism coexist in Brazil. On
the one hand, politicians deliver medicines and dentures to buy the votes of citizens with
whom they do not have ongoing relationships. Such transactions do not involve continued
access to healthcare services after Election Day. But in the same contexts, politicians also
engage in relational clientelism, providing citizens ongoing preferential access to ambulances
and medical treatments. Findings also suggest that female sterilization is used to obtain
support during elections, yet it remains unclear whether increased surgeries reflect electoral
clientelism, relational clientelism, or both. Further analyses will need to distinguish between
these forms of clientelism, and also determine whether tubal ligations are indeed contingent
38
Interview conducted by author in a municipality with 10,000 citizens on October 16, 2008.

31

Electronic copy available at: https://ssrn.com/abstract=1919567

on political support.
Failing to distinguish between electoral and relational clientelism may result in scholarly
confusion. To see why, consider Stokes’s (2005: 316) assertion that many clientelist practices
“make a mockery out of democratic accountability.” At first glance, it might appear that
Kitschelt (2000: 851) counters Stokes when he argues that it is “imprecise or even misleading”
to argue that “clientelist politics undercut democratic accountability, whereas programmatic
politics creates it.” Yet actually, these scholars’ viewpoints need not conflict. Although
neither Kitschelt nor Stokes makes this distinction explicit, they focus on entirely different
strategies of clientelism. On the one hand, Stokes focuses on a specific form of electoral
clientelism — vote buying — that rewards citizens for voting against their preferences during
a given election. This focus leads Stokes to argue that clientelism holds citizens, rather than
politicians, accountable for their actions: “perverse accountability — the ability of parties to
monitor constituents’ votes, reward them for their support and punish them for defection —
is what sustains machine politics” (325). By contrast, Kitschelt examines ongoing patterns of
relational clientelism, which he argues involve elite accountability: “Politicians who refuse to
be responsive to their constituents’ demands for selective incentives will be held accountable
by them and no longer receive votes and material contributions” (2000: 852; see also Kitschelt
et al. 2010: 292). More broadly, to ensure that researchers do not talk past each other, we
must be sure to distinguish between electoral and relational clientelism.
Although the present study focuses on distinction between these two forms of machine
politics, an important avenue for further research involves deeper analysis of the different
mechanisms that sustain them. Recent research has made considerably more headway in
unpacking the mechanisms underlying electoral clientelism than those underlying relational
clientelism. Given that strategies of relational clientelism involve credibility problems of both
elites and citizens, further investigation is needed to understand such ongoing relationships.
For example, what are the various ways in which politicians build trust among citizens
that they will actually follow through with promises of benefits after the election? Further

32

Electronic copy available at: https://ssrn.com/abstract=1919567

formal and empirical research would provide valuable insights. Another key issue is the
link between relational clientelism and electoral clientelism. For example, does relational
clientelism typically represent a substitute for, or a complement to, electoral clientelism?
Given that parties may engage in a portfolio of strategies when vying for votes (e.g., Estévez,
Magaloni & Diaz-Cayeros 2002; Calvo & Murillo 2010: 5–6), researchers should analyze
the factors that influence how parties allocate resources between longer-term relationships
and payoffs during elections. Overall, distinguishing between two fundamentally distinct
strategies — electoral clientelism and relational clientelism — may deepen our understanding
of both electoral practices.

Appendix: Description of Fieldwork in Brazil

While conducting my dissertation research, I lived in Brazil for more than 18 months.
Prior to and after the October 2008 municipal elections, I conducted a total of 110 formal
interviews on clientelism in the state of Bahia. These formal interviews included 55 interviews
of community members and 55 interviews of elites. Each of these interviews was conducted in
Portuguese, and lasted an average of 70 minutes. Each interview was taped and transcribed,
totaling over 4,500 pages of typed transcripts. In addition, I conducted informal interviews
of another 350 citizens and elites, as well as three focus groups of citizens.
My dissertation research focuses on small municipalities, as defined by those with 100,000
citizens or fewer. While much academic research on clientelism focuses on large metropoli-
tan areas, there are relatively few studies on smaller communities. This lack of research
is particularly unfortunate in Brazil, given that so much of its population lives in small
municipalities. In Brazil, 49 percent of the population lives in municipalities with 100,000
citizens or fewer. In addition, 96 percent of Brazilian municipalities are this size (IBGE
2000). Interviews were conducted in Bahia, the most populous state in the Northeast region
of Brazil with nearly 15 million citizens (IBGE 2009). The Northeast is the poorest region
of Brazil and one of the most unequal regions in the world. Approximately 41 percent of

33

Electronic copy available at: https://ssrn.com/abstract=1919567

Bahia’s population lives in households below the poverty level, and the state has amongst
the lowest social indicators in Brazil (World Bank 2004). The Gini coefficient of income
inequality in Bahia is 0.61 (World Bank 2004).
In order to identify potential themes, develop interview questions, and field test my
citizen and elite interview protocols, I began qualitative research in a municipality of 10,000
citizens in central Bahia, where I lived for approximately five months. During this time,
I selected a stratified random sample of six additional municipalities to conduct further
interviews. Overall, the municipalities spanned each of Bahia’s seven “mesoregions,” which
are defined by Brazil’s national census bureau (IBGE) as areas that share common geographic
characteristics. Figure 9 shows the geographical distribution of interview locations across the
state of Bahia. The population sizes of the seven municipalities selected were approximately:
10,000; 15,000; 30,000; 45,000; 60,000; 80,000, and nearly 100,000.
Within each selected municipality, individuals for community member interviews were
selected randomly using stratified sampling. Inclusion / exclusion criteria for individuals
included the following: (1) at least sixteen years of age (the voting age in Brazil), (2) had
lived in the municipality since the previous mayoral election in 2004, and (3) not a member of
the same household as any other interviewee. The sample was stratified to ensure balanced
representation across gender, age, and urban/rural mix.
With respect to elites, a range of perspectives were obtained by interviewing ten mayors
and former mayors, 28 city councilmen (vereadores), three vice-mayors, six party heads,
five heads of social services, and several other elites. These interviews were balanced to
include a combination of elites both allied and opposed to the current administration. Given
that mayors likely face different incentives if ineligible for reelection, the random sample of
municipalities was stratified to include several municipalities with second-term mayors.
Informed consent was obtained from all community members and elites before initiat-
ing each interview. The citizen and elite interview protocols consisted of both open-ended
and closed-ended questions. An iterative research design was employed; pertinent themes

34

Electronic copy available at: https://ssrn.com/abstract=1919567

 Figure 9: Map of Research Sites in Bahia (Northeast Brazil)

emerging during thematic analysis were investigated during ongoing interviews. While the
original, core questions in the interview protocols were asked of all respondents, probes about
emerging themes were included in later interviews.

References

Alesina, Alberto. 1988. “Credibility and Policy Convergence in a Two-Party System with
Rational Voters.” American Economic Review 78(4): 796–805.

Auyero, Javier. 2001. Poor People’s Politics: Peronist Survival Networks and the Legacy of
Evita. Durham: Duke University Press.

35

Electronic copy available at: https://ssrn.com/abstract=1919567

Banfield, Edward C. and James Q. Wilson. 1963. City Politics. Cambridge: Harvard Uni-
versity Press.

Boissevain, Jeremy. 1966. “Patronage in Sicily.” Man 1(1): 18–33.

Caetano, André Junqueira. 2000. “Sterilization for Votes in the Brazilian Northeast: The
Case of Pernambuco.” Ph.D. Dissertation, University of Texas at Austin.

Caetano, André Junqueira. 2010. “Ascensão e Queda da Laqueadura Tubária no Brasil?

Uma Avaliação das Pesquisas de Demografia e Saúde de 1986, 1996 e 2006.” PPGCS-PUC
Minas e Cedeplar-UFMG.

Caetano, André Junqueira and Joseph E. Potter. 2004. “Politics and Female Sterilization in
Northeast Brazil.” Population and Development Review 30(1): 79–108.

Cox, Gary W. and J. Morgan Kousser. 1981. “Turnout and Rural Corruption: New York as
a Test Case.” American Journal of Political Science 25(4): 646–663.

Desposato, Scott W. 2001. “Institutional Theories, Societal Realities, and Party Politics in
Brazil.” Ph.D. Dissertation, University of California, Los Angeles.

Division, UN Population. 2009. World Contraceptive Use 2009. New York: United Nations.

Dunning, Thad and Susan C. Stokes. 2008. “Persuasion vs. Mobilization.” Paper presented
at the American Political Science Association Conference, Boston, MA.

Estévez, Federico, Alberto Diaz-Cayeros and Beatriz Magaloni. 2002. “A Portfolio Diversi-
fication Model of Electoral Investment.” Typescript, Stanford University.

Falleti, Tulia G. 2010. Decentralization and Subnational Politics in Latin America. New
York: Cambridge University Press.

Finan, Frederico and Laura Schechter. 2010. “Vote-Buying and Reciprocity.” Typescript.

36

Electronic copy available at: https://ssrn.com/abstract=1919567

Fiorina, Morris P. 1978. “Economic Retrospective Voting in American National Elections:
A Micro-Analysis.” American Journal of Political Science 22(2): 426–443.

Gans-Morse, Jordan, Sebastian Mazzuca and Simeon Nichter. 2010. “Varieties of Clientelism:
Machine Politics During Elections.” CDDRL Working Paper, Stanford University.

Hicken, Allen D. 2002. “Parties, Pork and Policy: Policymaking in Developing Democracies.”
Ph.D. Dissertation, University of California San Diego.

IBGE. 2009. “Instituto Brasileiro de Geografia e Estatstica.” (http://www.ibge.gov.br).

Keefer, Philip. 2007. “Clientelism, Credibility, and the Policy Choices of Young Democra-
cies.” American Journal of Political Science 51(4): 804–821.

Kitschelt, Herbert, Kirk A. Hawkins, Juan Pablo Luna, Guillermo Rosas and Elizabeth J.
Zechmeister. 2010. Latin American Party Systems. New York: Cambridge University
Press.

Kitschelt, Herbert P. 2000. “Linkages Between Citizens and Politicians in Democratic Poli-
ties.” Comparative Political Studies 33: 845–879.

Kitschelt, Herbert P. and Steven I. Wilkinson. 2007. Patrons, Clients, and Policies: Patterns
of Democratic Accountability and Political Competition. New York: Cambridge University
Press.

Lawson, Chappell H. 2009. “The Politics of Reciprocity: Trading Selective Benefits for
Popular Support.” Typescript, Massachusetts Institute of Technology.

Lehoucq, Fabrice. 2003. “Electoral Fraud: Causes, Types, and Consequences.” Annual
Review of Political Science 6(1): 233–256.

Lehoucq, Fabrice E. 2007. When Does a Market for Votes Emerge? Historical and Theoretical
Perspectives. In Elections for Sale: The Causes and Consequences of Vote Buying, ed.
Frederic C. Schaffer. Boulder, CO: Lynne Rienner Publishers.

37

Electronic copy available at: https://ssrn.com/abstract=1919567

Lemarchand, Rene and Keith Legg. 1972. “Political Clientelism and Development a Prelim-
inary Analysis.” Comparative Politics 4(2): 149–178.

Levitsky, Steven. 2003. Transforming Labor-Based Parties in Latin America: Argentine

Peronism in Comparative Perspective. New York: Cambridge University Press.

Mair, Lucy P. 1961. “Clientship in East Africa.” Cahiers d’Études Africaines 2(6): 315–325.

Nichter, Simeon. 2008. “Vote Buying or Turnout Buying? Machine Politics and the Secret
Ballot.” American Political Science Review 102(1): 19–31.

Nichter, Simeon. 2009. “Declared Choice: Citizen Strategies and Dual Commitment Prob-
lems in Clientelism.” Typescript.

Nichter, Simeon. 2010. “Politics and Poverty: Electoral Clientelism in Latin America.” Ph.D.
Dissertation, University of California, Berkeley.

Organization, Pan American Health. 2008. Health Systems and Services Profile. Brasilia:
World Health Organization.

Osis, Maria Jos Duarte, Anı́bal Faúndes, Maria Yolanda Makuch, Maeve de Brito Mello,
Maria Helena de Sousa and Maria José de Oliveira Araújo. 2006. “Atenção ao Planeja-
mento Familiar no Brasil Hoje: Reflexões sobre os Resultados de uma Pesquisa.” Caderno
Saúde Pública 22(11): 2481–2490.

Piattoni, Simona, ed. 2001. Clientelism, Interests, and Democratic Representation: The
European Experience in Historical Perspective. Cambridge: Cambridge University Press.

Powell, John Duncan. 1970. “Peasant Society and Clientelist Politics.” The American Po-
litical Science Review 64(2): 411–425.

Robinson, James A. and Thierry Verdier. 2003. “The Political Economy of Clientelism.”
Typescript.

38

Electronic copy available at: https://ssrn.com/abstract=1919567

Schaffer, Frederic C. 2007. Elections for Sale: The Causes and Consequences of Vote Buying.
Boulder, CO: Lynne Rienner Publishers.

Schaffer, Frederic C. and Andreas Schedler. 2007. What is Vote Buying? In Elections for
Sale: The Causes and Consequences of Vote Buying, ed. Frederic C. Schaffer. Boulder,
CO: Lynne Rienner Publishers.

Scott, James C. 1969. “Corruption, Machine Politics, and Political Change.” American
Political Science Review 63(4): 1142–1158.

Scott, James C. 1972. “Patron-Client Politics and Political Change in Southeast Asia.”
American Political Science Review 66(1): 91–113.

Shefter, Martin. 1994. Political Parties and the State. Princeton: Princeton University Press.

Speck, Bruno Wilhelm and Cláudio Weber Abramo. 2001. “Transparency International
Brazil / Ibope Survey – Summary Report.”.

Stokes, Susan C. 2005. “Perverse Accountability: A Formal Model of Machine Politics with
Evidence from Argentina.” American Political Science Review 99(3): 315–325.

Stokes, Susan C. 2009. “Pork, by Any Other Name ... Building a Conceptual Scheme of
Distributive Politics.” Paper presented at the American Political Science Association Con-
ference, Toronto, Canada.

Tribunal Superior Eleitoral (TSE). 2009. (http://www.tse.gov.br).

Tribunal Superior Eleitoral (TSE). 2010. (http://www.tse.gov.br).

Weingrod, Alex. 1968. “Patrons, Patronage, and Political Parties.” Comparative Studies in
Society and History 10(4): 377–400.

39

Electronic copy available at: https://ssrn.com/abstract=1919567