Whither Clientelism?

Good Governance and Brazils Bolsa Famlia Program

Natasha Borges Sugiyama and Wendy Hunter

Providing social services to poor people without entangling them in patron-client

networks is a prevalent development goal. In many developing countries delivering
services free of political interference has proved exceedingly difcult. Examples of
ambitious social programs that become grist for subnational patronage mills ll the
comparative politics and development studies literatures. It is especially challenging
to ensure clean social provisioning in large countries where federal authorities oversee
programs that require local cooperation. Even programs meant to enhance poor peoples
autonomy from patronage machines can end up intensifying their dependence on them.
Given that the clean delivery of basic services is a cornerstone of the development of
social rights, it is important to learn ways to design and implement antipoverty programs
that avoid clientelism.
This rst requires identifying cases where social programs are implemented free
of political manipulation in countries with a history of patronage politics. Yet most of
the social science literature focuses on cases where this has not occurred,1 and provides few insights into successful examples, dened here as social provisioning that
conforms to rights-based bureaucratic principles rather than those of patronage politics,
whereby voters trade political support for particularistic goods and services. Also, the
positive cases analyzed often involve state governments, such as in Cear in Brazil
and Kerala in India.2
Whereas many conditional cash transfer (CCT) programs are regarded as susceptible to political machinations,3 Brazils Bolsa Famlia is often presented as a major
federal program that has avoided the clientelistic pitfalls potentially entailed by
municipal co-administration. Drawing on survey data and evidence from focus groups,
this article investigates whether in fact the Bolsa Famlia is clean. It would indeed be
noteworthy if, in a country of over 5,000 municipalities with uneven administrative
capacity, such a program managed to reach recipients without local political interference. To preview our ndings, good government in a continental-size country
43

Comparative Politics October 2013

with a history of clientelism is possible. Solid ministerial commitment, thoughtful
policy design, and strong technocratic capacity appear to be important in structuring
social provisioning in ways that avoid clientelism and empower ordinary citizens.
Brazils Bolsa Famlia is the largest CCT program in the world. Widely credited
with raising living standards, the grant represents a signicant income source for
roughly 52 million poor Brazilians.4 The Bolsas main goals are to alleviate poverty
and to improve the education and health of low-income children. The government
entity in charge, the Ministrio do Desenvolvimento Social (Ministry of Social Development, or MDS), casts federal assistance as a right of all citizens whose incomes fall
below an amount necessary to fulll basic needs and develop human capital. If this
is the case, poor families should receive regular nancial support free from political
discretion. Moreover, recipients should understand the program in this way.
The size, scope, and geographic concentration of the Bolsa Famlia in Brazils
impoverished Northeast, a historic stronghold of clientelism, call for an investigation
of its political effects. Specically, despite its bureaucratic appearance, has the Bolsa
reinforced patron-client relations? Or does it break with the historic tendency of
federal social assistance programs to fuel local patronage networks? Moreover, is
there any evidence that the program has fostered an understanding among citizens
of social assistance as a right or government entitlement rather than as a favor
of a personal benefactor?
Most research on the politics of CCTs examines their impact on the electoral
prospects of presidential candidates.5 Many studies have noted the Bolsas positive
effects on the electoral prospects of the former presidential candidate most closely
identied with it, Luiz Incio Lula da Silva.6 Tracy Fenwick and Alfred Montero
also examine the relationship between the program and gubernatorial dynamics.7
By cutting out state governments from administering the program, federal ofcials
effectively bypassed one important historic source of resource diversion. State ofcials cannot gain access to Bolsa Famlia resources to build their clientele. Yet the
Bolsas impact on municipal politics also warrants investigation, given that local
ofcials help people apply for the program and monitor recipients compliance with
its conditions. The paucity of attention paid to whether local manipulation exists is
surprising given how public resources have been distributed historically in Brazil.
The other major strand in the CCT literature is more policy based, focusing on
outcomes like improved service delivery and efciency. Measuring such outcomes in
technocratic terms, this literature does not address how they translate into the quality
of citizens interactions with the state.8 For example, can citizens gain access to
the Bolsa Famlia through exclusively bureaucratic channels? Do they view it as transparent and clean? For insight into how ordinary people experience the program,
researchers should hear directly from citizens.
This study aims to ll important gaps in both bodies of literature. We examine
the politics of the Bolsa Famlia at the municipal level, where the majority of direct
interactions between beneciaries and the state takes place. We also learn about the
program through the voices of participants themselves. Our survey and focus group
44

Natasha Borges Sugiyama and Wendy Hunter

research was conducted among people whose socioeconomic status (low income and
education) and location of residence (Northeast Brazil) render them likely candidates
for political manipulation.
Much is at stake in how the distribution of Bolsa grants is determined. Not reaching the neediest families would be a serious outcome of clientelistic manipulation.
Moreover, meaningful democratic development would be limited if recipients perceive
the program as subject to vote buying.9 This perception might also contribute to undermining collective efforts to claim a greater share of public entitlements in the long
term. As Scott Desposato writes, vote buying at its worst eliminates both participation and inuence. Voters do not participate in public decision making, instead trading
their votes for private gains.10 If the Bolsa Famlia is implemented in a strict bureaucratic fashion using clear eligibility criteria, it may help free poor Brazilians from
thinking that securing social assistance depends on supporting local political machines.

Theoretical and Empirical Debate

In countries where a sizable poor population votes in democratic elections, pro-poor
policies often become grist for political clientelism.11 Evidence abounds of Latin
American politicians using seemingly neutral antipoverty programs for particularistic
political gain. We must differentiate between clientelism and vote buying. Following
Luis Roniger, clientelism refers to a
nonuniversalistic quid pro quo between individuals or groups of unequal standing.
It implies mediated and selective access to resources and markets from which others
are normally excluded.Those in controlpatrons, subpatrons, and brokers
provide selective access to goods and opportunities and place themselves or their
supporters in positions from which they can divert resources and services in their favor.
Their partnersclientsare expected to return their benefactors help, politically
and otherwise.12

Examples of clientelism include offers of cash, material goods (bricks, food, small
appliances), and services such as health care, or public sector employment in exchange
for political support. Vote buying, which occupies the far end of the continuum
between programmatic and clientelist strategies of electoral mobilization, entails
the offering of particularistic material rewards to prospective voters near election time.
Compared to patronage, which often has an ambiguous legal status, vote buying is
almost always illegal.13
Many studies stress several background conditions that render people vulnerable
to clientelism, namely, poverty, inequality, and rural settings.14 Patronage and vote
buying are often viewed as especially prevalent where there is severe poverty, given
the high value placed on benets that patrons can deliver.15 With the low levels of
education that typically accompany extensive deprivation, indigent individuals are less
likely to organize for broad collective solutions. Recent studies also suggest inequality
45

Comparative Politics October 2013

as a key contributor to vote buying.16 Where the gap between the haves and have
nots is extreme, vote buying is an inexpensive strategy; politicians can pay out
relatively little to win over large numbers of poor people.17 Rural settings are also
highly conducive to clientelism. One explanation posited for the link concerns the
clients normative sense of obligation to the patron.18 Another explanation involves
vote monitoring. Sparsely populated rural communities tend to be small political
constituencies, where politicians are better able to gure out how people vote.19 In
fact, cross-national evidence suggests that rural Latin Americans are more likely to
report having been offered material goodies in exchange for their votes. A recent
AmericasBarometer conrms what anthropological and social scientic studies alike
have long concluded: that basic survival needs and a lack of options fuel the predatory
practices of vote-seeking politicians.20
A second issue raised in the literature concerns the opportunities for manipulation
that seemingly modern national programs entail for local politicians. Scholars have
long recognized the role of intermediaries in linking local to national units.21 Even
when states develop to the point of having institutionalized programs of social protection, poor people often encounter so many challenges in navigating the system
that they seek the assistance of brokers. The result is a modern version of the
traditional patron-client relationship.22 The more necessary local mediation is in
gaining effective access to national benets, the more political inuence is to be
had. Beyond local politicians, street-level bureaucrats can exercise tremendous discretion over who obtains and maintains benets.23 Although public employees should
follow bureaucratic rather than political criteria, they are themselves often enmeshed
in clientelistic networks, owing their jobs to patronage-bearing politicians.
Reasons to Expect Clientelistic Provisioning Why might the Bolsa Famlia be
expected to exhibit traditional patron-client dynamics? First, the sociodemographic
features describedpoverty, inequality, and ruralnesscharacterize the region where
the majority of beneciaries live.24 Second, the local authorities assigned role opens
up considerable opportunities for brokerage. Thus, while the CCT appears to operate
according to means-tested bureaucratic principles, there are compelling reasons to
think that municipal-level politicians might use the federal transfers in quid pro quo
exchanges. The logic of our approach conforms to a most likely case study for
conrming the theoretical predictions of clientelism.25 If in these highly probable
contexts people did not report local political manipulation, they would be even less
likely to do so in other settings. If expectations are not conrmed, the subsequent
search for an explanation may generate new and interesting causal insights.
The Northeast is not only Brazils least-developed area, but it also has the largest
concentration of rural poverty in Latin America.26 Compared to other regions of
Brazil, it has the highest percentage of households making less than one minimum
wage.27 Over half the population (57 percent) lives in poverty and over a third (37 percent) is indigent.28 Furthermore, a striking number of residents (roughly 26 percent) still
lives in rural areas, where the poverty rate reaches 76 percent.29 Income inequality is the
46

Natasha Borges Sugiyama and Wendy Hunter

highest in Brazil due to extraordinarily skewed patterns of land ownership.30 Amidst
this backdrop, one can expect local politicians wanting to take credit for Bolsa
Famlia transfers, which equal as much as 11 percent of some Northeastern locations municipal budgets.31
In fact, stark deprivation in the poor, arid zone has rendered Northeasterners
vulnerable to patronage appeals.32 Clientelism has assumed various forms, including
the manipulation of federal programs by subnational political elites.33 A well-known
example is SUDENE (Superintendency for the Development of the Northeast), an
economic development agency inaugurated in 1959 that soon became a huge
patronage operation.34 Similarly, resources from various social ministries in Braslia
have been frequently misused. For example, food from a Ministry of Education
school lunch program has fueled many local patronage networks. Survey evidence
also reects the notorious wheeling and dealing in the region. A recent crossregional poll found that 46 percent of Northeasterners claimed personal knowledge
of people trading votes for material benets, and 21 percent reported that they had
done so themselves.35
A second reason to expect local shenanigans with Bolsa Famlia grants concerns
the municipal authorities assigned role. Although the federal Ministry of Social
Development retains ultimate decision-making authority, local governments serve as
citizens main contact point, creating a ripe opportunity for political brokerage. People
apply for the Bolsa Famlia through municipal ofces, where local ofcials do all the
intake, reviewing documents, interviewing families, and entering personal data into
a national database. In some small cities, the application process takes place in the
prefeitura (city hall), in which case the perceived distance between the municipal
and federal roles may be nonexistent. There are several ways in which enterprising
politicians could facilitate the poors entry and maintenance in the program, such as
transporting destitute families to the nearest Bolsa Famlia ofce or helping applicants
obtain the numerous identity documents required. Also, in many places mayors
appoint the municipal supervisors of the Bolsa. Local ofcials (teachers, nurses, and
social workers) responsible for monitoring beneciaries compliance with conditionality requirements may themselves owe their jobs to patronage dynamics. In short,
the political insulation that comes with a professional civil service is far from
guaranteed.36 Given these factors, local political interference in the programs operation
would not be surprising.
Finally, media reports cast doubt on the Bolsa Famlias independence from
local politics. For instance, the Folha de So Paulo reported allegations that
mayoral and city council candidates had used the program for vote buying before
the October 2008 municipal elections. The Folhas investigation found isolated
cases of electoral interference in Cear and Piau, and identied complaints in
three other Northeastern states where ofcials allegedly promised some voters a
Bolsa Famlia card in exchange for their support and threatened backers of
opposition candidates with a loss of benets.37 It is crucial to discern whether
these select reportswhich themselves could have resulted from overzealous
47

Comparative Politics October 2013

investigative journalists, sentiments against the program, or personal animosities
against certain ofcialsrepresent systematic and pervasive malfeasance.
In sum, this investigation is born out of theoretical debates in comparative
politics and empirical features of Brazilian politics. Given the demographic prole
of recipients and the region, the intake and monitoring role of municipal governments, and occasional media investigations, there are multiple reasons to expect
efforts to appropriate Bolsa Famlia grants for political gain. To assess whether this
is happening, we turn to the poor for insight.
Research Design and Empirical Test
To investigate whether local politicians use the Bolsa Famlia in ways that reinforce
clientelism, we conducted eld research in three Northeastern municipalities in 2009 and
2011. A signicant degree of clientelism could be expected in all three locations. A
common feature is poverty, arguably the single most important condition that renders
people vulnerable to vote buying. Each municipality displays at least one related or additional characteristic highlighted by studies of clientelism: extreme poverty (Camaragibe, in
the state of Pernambuco), a sparsely populated rural setting (Pau Brasil, in Bahia state),
or high socioeconomic inequality (Jaboato dos Guararapes, also in Pernambuco).
Camaragibe is extremely poor even by the standards of Northeast Brazil. The
most impoverished of the three locations, nearly three-quarters (74.5 percent) of its
population fall below the poverty line.38 This medium-sized city (population 136,381)
has a Gini coefcient of 0.41, and 35 percent of all households receive the Bolsa.
In Pau Brasil 52 percent of all residents live below the poverty line and income
inequality is less marked than in Camaragibe (Gini coefcient of .38).39 Its small
size (12,111 people), location in the states interior (32 percent of the population
are in a rural area), and low employment prospects for agricultural workers render
its electorate vulnerable to clientelism. Of our three cities, Pau Brasil has the highest
proportion of households receiving the Bolsa Famlia (63 percent). Finally, Jaboato
is a large industrial city (population 630,008) that neighbors the state capital, Recife.
Jaboato exhibits considerable poverty (54 percent) and has the highest income
inequality (Gini coefcient of .45) of the three cities.40 Approximately 32 percent
of households receive the family grant. Empirical accounts of egregious political
behavior are also rife. A recent mayor, following a sequence of others political
malfeasance, was known to offer cofns in exchange for votes. In short, these
municipalities reect the most signicant socioeconomic and demographic preconditions for clientelism. Therefore, it could be expected that respondents would report
a high degree of political interference with the Bolsa Famlia.
Survey Data and Methodology To understand how poor people view the politics
of the Bolsa Famlia, we administered a face-to-face survey in each of the case study
cities in 2009.41 In addition, we met with Bolsa Famlia recipients from each city
48

Natasha Borges Sugiyama and Wendy Hunter

through eleven focus groups.42 In 2011 we interviewed local and federal ofcials
involved with the program. Insights from the focus groups and interviews inform
the discussion of the statistical results.43
The survey consisted of an approximately fteen-minute, face-to-face questionnaire, administered by a team of trained university students who went door-to-door
in poor neighborhoods located throughout each city. We employed a quota sampling
method to capture the experience of recipients and nonrecipients. Quota samples
are non-probabilistic samples designed to capture a predetermined distribution of
respondents based on important parameters, such as sociodemographics.44 One of
the main drawbacks is that interviewers may introduce selection bias. We attempted
to correct for this through training and supervising the interviewers. In addition, we
required even geographical distribution of subjects in our cities.
Practical, ethical, and theoretical considerations informed our sampling methodology. First, given that telephone coverage in Brazil remains low among the poor, a
randomized phone survey would not have captured our population of interest.45 The
use of face-to-face surveys was the best method to include the poor. Second, given
the sensitive nature of some questions and the economic vulnerability of the relevant
population, it was essential that respondents trust the independence of the research
team. For this reason, we interviewed participants in their homes, rather than recruit
them in public venues such as municipal buildings or town squares. The survey team
made clear that it was not afliated with any government agency (local, state, or
national) and had no authority over peoples social benets. Finally, recruitment
based on place of residence was intended to avoid any bias that might come from
selecting participants from ofcial program rolls. Given the theoretical possibility
that local politicians might manipulate the program for political purposes, we did
not utilize ofcial lists of Bolsa Famlia beneciaries or applicants produced by local
or federal governmental authorities.
As our sampling method is non-random, we cannot reliably estimate the standard
errors and coefcients for the broader population in the Northeast or in Brazil more
generally. In other words, we are not willing to assert that errors in our models will
be independent and identically distributed. This is not to say that quota sampling can
never produce estimates that have good inferential qualities. According to T.M.F.
Smith, under certain circumstances inferences about the parameters of a conditional
distribution (in our case, the conditional distribution of the log-odds of responses to
various vote-buying questions) can be valid.46 The conditions are: (1) the mechanism
that selects which houses to visit is not a function of the variables measured in the
survey; (2) that the choice of whether to include the observation in the sample is not a
function of variables measured in the survey (save the Bolsa Famlia participation
question); and (3) that the parameters governing selection are distinct from those that
describe the conditional distribution of responses (which, since the other two conditions are true, seems reasonable here). This suggests that the statistical inferences
about the coefcients in the model can be made, if only to the population selected
into the sample.47
49

Comparative Politics October 2013

The quota sample, by design, is 80 percent Bolsa Famlia participants and 20 percent nonparticipants. Actual participation rates among the poor (that is, those potentially included in our sample) for the three cities in our study are: Camaragibe
(62.76 percent), Jaboato (74.05 percent), and Pau Brasil (69.9 percent).48 Thus, Bolsa
participants are oversampled in our data, which could bias our coefcient estimates.
In an effort to reduce this type of bias, we reweighted the data to impose the appropriate population distribution. With this weighting, we expect the relationship between
our sample and the population of poor people in these cities to be similar. Thus, we
use these data to estimate the effects of our variables of interest on vote buying and
to draw inferences about the experiences of the poor in our case study cities.
Dependent Variables We develop three models to analyze the degree of clientelism
and political interference associated with the Bolsa Famlia. To capture respondents
experiences and perceptions of clientelism, the survey asked three questions: (1) Do
politicians buy votes in your city? (2) In the last election, did a politician offer you
something in exchange for your vote? (3) Is the Bolsa Famlia used in exchange for
votes? We are interested not only in the responses to each question but also in how
respondents differed across all three scenarios.
The rst question seeks to gauge general perceptions about politics and
clientelism in respondents cities. Overall, 65 percent reported that politicians buy
votes in their city. The second question inquires more directly about peoples personal
experiences in the most recent election. Arguably more sensitive in nature, the question was worded so as not to compromise the respondent. (Reporting that a politician
offered you something does not necessarily mean that you accepted.) More than a
quarter (28.3 percent) reported that they had received a recent vote buying offer.
The third question asks people about vote buying in specic relation to the Bolsa
Famlia. Far fewer (14.9 percent) reported that they thought the Bolsa was used for
vote buying. All the dependent variables are coded dichotomously.
Independent Variables
Municipality: Each city presents a somewhat distinctive political and socioeconomic
scenario associated with clientelism. Camaragibe has the highest rates of poverty,
Pau Brasil is predominantly small and rural, and Jaboato has the highest income
inequality. Given the diversity of settings, we anticipate city effects on vote buying
and test their independent effects.
Our models include four demographic control variables that may also be relevant
to respondents vote-buying experiences.
Bolsa Famlia Participation: While all respondents share socioeconomic vulnerability to vote buying, participation in the Bolsa Famlia situates recipients to assess
whether local politicians use it to buy votes. Arguably, Bolsa recipients are better positioned to know whether they had to sell their votes to receive or maintain the benet.
50

Natasha Borges Sugiyama and Wendy Hunter

Education: It is reasonable to expect educational attainment to play an important
role in the lives and worldviews of respondents. As education increases, the life prospects for individuals also increase through better employment and higher incomes.
We anticipate that respondents with more education will be less vulnerable to patronage
and vote buying. The average respondent had less than six years of formal schooling.
Sex: Studies of development show that men and women experience poverty
and social exclusion differently.49 We anticipate women to be more informed about
the Bolsa Famlia as grants are targeted to mothers. Further, social networks and information sharing are likely to form along gendered lines. Given the gendered effects
of poverty and the Bolsas program design, we control for sex.
Age: Finally, we control for the potential effects of respondents age as it may
inform their experiences and perceptions of clientelism. We reason that younger
respondents will have voted in fewer elections and thus had fewer opportunities to
observe vote buying. Conversely, older citizens will have had more elections from
which to recall quid pro quo offers. Age, however, may not matter for experiences
with recent vote buying (in the last election) and perceptions of vote buying with the
Bolsa Famlia.
For further details on all the above variables, including measurement, coding,
and descriptive statistics, see the Appendix.
Model Estimation and Results
Table 1 displays the ndings from the three logistic models constructed around survey
participants perceptions and experiences of general vote buying.
High levels of vote buying are evident everywhere; 66 percent of the sample
reported that politicians typically offer quid quo pro exchanges for votes. Even
so, model 1 shows that perceptions of vote buying vary according to location. Compared to Pau Brasil, respondents from Camaragibe are signicantly less likely to
report that politicians bought votes in their city, as indicated by the negative coefcient. Jaboato and Pau Brasil were not signicantly different from each other in
this model.
The predicted probability of reporting vote buying can be analyzed using the log
odds ratio (see Table 2 for predicted probabilities for all models). The probability varies
depending on where the respondent resides. For the mean respondentwho receives
the Bolsa Famlia, is thirty-nine years old, studied for 5.5 years, and is femalethe
probability of reporting vote buying is highest if she lives in Jaboato or Pau Brasil
(69 percent), but drops if she resides in Camaragibe (54 percent). Clearly, this model
reveals that our respondents identify a tremendous amount of general vote buying in
their cities, and its extent varies according to local conditions.
The second model, which examines offers for votes in the most recent election,
also yields insights into the effects of local conditions. Respondents were more likely
to report receiving an offer in Jaboato followed by Camaragibe and then Pau Brasil.
51

Comparative Politics October 2013

Table 1

Logistic Models for Clientelism

Model 1
Do politicians buy
votes in your city?
Coef.

Citya
Camaragibe
Jaboato dos Guararapes
Controls
Bolsa Famlia
Participation
Age
Education
Male
Constant
N
Log Likelihood
Prob > Chi2
Pseudo R2

Std. Err.

Model 2
In the last election, did
anyone try to buy
your vote?
Coef.

Std. Err.

Model 3
Is Bolsa Famlia used
for vote buying?
Coef.

Std. Err.

.6541
.0141

.1777
.1870

.5223
.9661

.2016
.2013

.1618
.3572

.2398
.2416

.4441

.1430

.0166

.1430

.9250

.1678

.0254
.0581
.0104

.0057
.0208
.2605

.0212
.0081
.1757

.0062
.0209
.2630

.0154
.0544
.9382

.0074
.0251
.2715

1.919

.3411

.7524

.3493

1.1142

.4147

1100
680.6047
0.000
0.0537

1100
647.9518
0.0000
0.0356

1100
474.9464
0.0000
0.0657

Pau Brasil is the comparison case.

p .05. p .01. p .001.

How do the city-level variables affect the predicted probability that a poor resident reports having received a vote-buying offer? For the average respondent, residence in Jaboato and Camaragibe results in an offer about one-quarter of the time;
in Pau Brasil, such reports are less likely. Indeed, city context matters for the predicted
probability of reported quid pro quo offers.
Finally, the results in model 3 offer a sharp contrast to the previous logistic
models. Unlike those models, which identied varied local experiences with vote
buying, in model 3 respondents across all cities report that the Bolsa Famlia is not used
for vote buying. In other words, whereas city of residence explains variation within
models 1 and 2, in model 3 city variables are not signicant. Respondents views on vote
buying and the Bolsa Famlia are consistent across all three municipalities. Instead, the
control variables yield the leverage in explaining reported differences in the use of
the Bolsa Famlia for vote buying. The most analytically relevant nding in this regard
is that recipients of the program were the least likely to view it as grist for vote buying.
This question reveals a dramatically different scenario regarding the likelihood
that poor families view the Bolsa Famlia in exchange for votes. Drawing once again
on the log odds ratio to generate predicted probabilities, city of residence has no
52

Natasha Borges Sugiyama and Wendy Hunter

Table 2

Predicted Probabilities for Models 1, 2, and 3

Model 1
Jaboato dos Guararapes
Camaragibe
Pau Brasil

Model 2
Jaboato dos Guararapes
Camaragibe
Pau Brasil

Model 3
Jaboato dos Guararapes
Camaragibe
Pau Brasil

Predicted Probability
Pr(y51|x): 0.6949
Pr(y50|x): 0.3051
Pr(y51|x): 0.5386
Pr(y50|x): 0.4614
Pr(y51|x): 0.6919
Pr(y50|x): 0.3081

95% Condence Interval

[ 0.6437,
0.7460]
[ 0.2540,
0.3563]
[ 0.4832,
0.5940]
[ 0.4060,
0.5168]
[ 0.6294,
0.7543]
[ 0.2457,
0.3706]

Predicted Probability
Pr(y51|x): 0.3584
Pr(y50|x): 0.6416
Pr(y51|x): 0.2638
Pr(y50|x): 0.7362
Pr(y51|x): 0.1753
Pr(y50|x): 0.8247

95% Condence Interval

[ 0.3061,
0.4107]
[ 0.5893,
0.6939]
[ 0.2169,
0.3108]
[ 0.6892,
0.7831]
[ 0.1257,
0.2249]
[ 0.7751,
0.8743]

Predicted Probability
Pr(y51|x): 0.1216
Pr(y50|x): 0.8784
Pr(y51|x): 0.1023
Pr(y50|x): 0.8977
Pr(y51|x): 0.0883
Pr(y50|x): 0.9117

95% Condence Interval

[ 0.0892,
0.1541]
[ 0.8459,
0.9108]
[ 0.0728,
0.1317]
[ 0.8683,
0.9272]
[ 0.0542,
0.1224]
[ 0.8776,
0.9458]

Calculations for average respondent: Bolsa Famlia participant, 38.9 years old, 5.58 years of education,
and female.

meaningful effect on the probability of vote buying with the Bolsa. For the average
respondent, the probability is only 12 percent for Jaboato, 10 percent for Camaragibe,
and 9 percent for Pau Brasil. As reported in Table 1, demographic factors drive the
variation. Holding all else constant, simply participating in the Bolsa Famlia alters the
probability that respondents view the program for vote buying (from a high of 23 percent to a low of 12 percent). Given this result, familiarity with the programs operation
matters. Whereas those with direct knowledge of how the Bolsa functions are more
likely to view it in bureaucratic terms, those without personal experience are more
likely to presume that it is prone to political manipulation.
The side-by-side comparisons of these models allow us to contextualize views on
vote buying with the Bolsa Famlia. Most important, whereas models 1 and 2 suggest
that perceptions and experiences of vote buying vary according to local conditions, the
model about the Bolsa Famlia program generates a standard response. Had there been
political interference with this social policy, we would have found some degree of
local variation. Instead, we found the opposite. That the experiences of poor families
53

Comparative Politics October 2013

from three very different cities were uniform on the question of whether the Bolsa
Famlia was insulated from vote buying suggests that the program stands apart from
conventional political wheeling and dealing. In the next section, we explore why that is.

Discussion of Findings
Our focus groups and eld research corroborated the statistical results of the three
models. Focus group testimonies revealed not only the general prevalence of clientelism
but also the exception represented by the Bolsa Famlia in these communities. Given
clientelisms widespread operation, it is noteworthy that people saw the Bolsa Famlia
as different. Focus group participants told us that vote buying was frequent and
typically took the form of exchanges for goods or favors. For example, some related
that council members sought to buy votes for amounts between 5 to 50 Reais (approximately $2.50 to $25 US dollars).50 Other respondents indicated that they had received
basic goods, such as a cesta bsica (food basket), bricks, and cement, in exchange for
votes.51 One participants brother was even promised a job in exchange for his entire
familys vote.52 The readiness with which participants acknowledged vote buying suggests that clientelism remains a reality of local elections.
At the same time, participants insisted that Bolsa Famlia grants were not being
traded for votes. When asked whether politicians were either promising Bolsa
resources or threatening to withdraw them our respondents resoundingly answered
no. One person stressed,
I think that politicians would be wrong to try to use Bolsa Famlia for [vote buying].
It would be wrong because the benet is federal. It comes from a federal agency.53

As she went on to explain, the poor would not believe local ofcials promises to
provide or threats to withdraw Bolsa Famlia grants because they understand that the
program falls outside their inuence. This sentiment was echoed across the three communities.54 Braslia decides was commonly repeated to emphasize that local political
efforts to inuence the program would violate its very terms and not even be possible
in practice. Importantly, the focus groups also revealed that local politicians do not play
a regular brokerage role. No one testied to this kind of political assistance, despite
expressions of difculty understanding some aspects of the programs operation, like
the amount of time it took to enroll and changes in grant amounts. Clearly, people
seemed to know the limits of what local authorities could readily deliver or withdraw.

Exploring Causes: Perspectives from Below Suggest How Program Design Matters
What accounts for the nding that the Bolsa Famlia does not appear to grease
clientelistic networks in a most likely case? Testimony from beneciaries suggests
54

Natasha Borges Sugiyama and Wendy Hunter

that policy design choices instituted from above can inuence, in meaningful ways,
perspectives from below. More specically, how the Bolsa operates causes ordinary
citizens to perceive limits on the scope of local political interference. We suggest
preliminarily some of the characteristics that could make social programs less available
for clientelism. The national agency in charge should control entry into the program,
institute a direct mechanism to transfer resources to beneciaries, provide user-friendly
bureaucratic channels for solving problems, and institute transparency and oversight
mechanisms. Moreover, it must inform the public about these measures. Conveying
clearly the terms of a program is a crucial rst step in reducing clientelistic interference,
which breeds on the misunderstandings and vulnerabilities of low-income populations.
Tools to control entry: Strong national oversight by technocrats committed to
running a clean program and equipped with the administrative and technical capacity
to do so appears to be a key precondition for insulating CCTs from local political
dynamics. Above all, strict targeting criteria need to be adhered to. Authorities must
convey that the programs terms do not allow for negotiation.
The Ministry of Social Development in Braslia has exercised such leadership.
In 2004 President Luiz Incio Lula da Silva established the MDS as the Bolsa
Famlias home, and staffed it with well-trained technocrats committed to targeting
resources toward the poorest Brazilians while keeping them away from the sticky
ngers of municipal-level politicians.55 Technocratic criteria structure decisions about
who receives the program. First, based on countrywide poverty maps, the MDS
determines the number of grants to be disbursed per municipality.56 Eligibility formulas, based on income level and household composition,57 increase the probability
that grants go to the most destitute families. The MDS claries its rules in a booklet
provided to applicants and beneciaries. We encountered its widespread availability
while in the eld. Entitled, Bolsa Famlia: Agenda de Compromissos da Famlia,58
it is illustrated by simple graphics and written in clear, accessible language. After
outlining the requirements of eligibility and maintenance in the program, the booklet
underscores in bold letters: Keep in mind: If your family meets the eligibility requirements of the program, receipt of the benet is your right, not a favor from anyone.59
Technocrats from the MDS also manage a national database (Cadastro nico)
that incorporates the household information and assessments described above. They
have means to verify some of the given information through several other government
databases. Crosschecking ability is key to heightening Braslias inuence. For example, if an applicant claims wages below the poverty line but other databases (such
as one for pensions) reveal a household income in excess of Bolsa Famlia eligibility,
she will be screened out. With such tools, personnel in the MDS make the nal
decision about whether a family receives the benet. We heard consistently from
focus group participants that Braslia decides entry into the program. While they did
not know the ins and outs of the process, they did understand the apolitical nature of
the screening.
Direct payment mechanism: National authorities should transfer resources directly
to beneciaries and thereby eliminate local governments as nancial intermediaries.
55

Comparative Politics October 2013

The advantage of a direct transfer is two-fold. First, it reduces the potential for funds
to be diverted, a problem experienced with many federal transfers. Second, the more
impersonal the nature of the pay out, the less the beneciaries have anyone to thank
for performing a favor. Focus group participants often stressed how much they
valued the personal autonomy afforded by the direct transfer.
A critical innovation featured by the Bolsa Famlia is the direct payment
mechanism to beneciaries through ATM-like cards. In most cities, recipients take
their Bolsa Famlia card to a bank (Caixa Econmica Federal) to receive the cash
grant.60 The direct payment mechanism helps account for peoples perception of the
central role of the federal government and the low utility of local patrons. As one
person stated, The program must be federal because the card says so.61 Participants in all three cities recognized that Braslia paid for the program and
ultimately determined who received the grants while municipal personnel dealt
with the data entry aspect of applications. Notably, the small booklet given to
beneciaries explains how payments work and contains a pocket where cards
can be inserted for safekeeping.
Bureaucratic problem-solving mechanisms: User-friendly bureaucratic features
should be instituted to induce beneciaries to seek help through ofcial channels
instead of local political intermediaries. For the impoverished segment of the population enrolled in many social programs, the temptation to resort to local sources of
assistance can result in brokerage.
MDS announcements encourage participants to solve problems through bureaucratic procedures rather than municipal brokers. The ATM card opens the way for
Caixa Econmica bank tellers, who are federal civil servants, to help people solve
problems with their payments. A toll-free number encourages applicants and recipients to call the ministry if they have questions. The small beneciary booklet not
only provides the number but also indicates who will be answering it: personnel
from the MDS in Braslia.62
Focus group participants reported that instead of turning to a local ofcial for
help they used bureaucratic (mainly federal) channels to overcome confusion and
roadblocks. Several indicated that bank tellers at the Caixa Econmica had provided
guidance. Others remarked on having called the [0-800] number provided by the
MDS. One participant even reported that her daughter called Braslia to nd out
why she hadnt received her Bolsa card and had her problem resolved.63 Participants
apparently feel they can turn to bureaucratic channels for assistance. Besides reducing
local political intervention, such institutional means of problem solving seem to have
created more uniform administration of the program across municipal governments.
Transparency and Oversight Mechanisms: The provision of transparency and
oversight mechanisms are also important in the detection of local wrongdoing.
As Mathew McCubbins and Thomas Schwartz note, policymakers use of realarms a system of rules, procedures, and informal practices that enable citizens
and organized interest groups to examine administrative decisions and seek
remediescan be an effective tool for policy oversight.64 In Good Government
56

Natasha Borges Sugiyama and Wendy Hunter

in the Tropics, Judith Tendler noted the signicance of citizen-oversight mechanisms
to enhance accountability in the Brazilian state of Cear.65 The states campaigns to
urge citizens to monitor the public servicein the cases of health, drought relief, and
procurementbrought about civic action. In Tendlers words, such messages put in
peoples minds a vision of better and more just government, implored citizens to stand
up against the abuses of government, and explained how to recognize infractions
and what to do about them.66 Notably, the dynamic she observed regarding the
state-municipal relationship in Cear holds for the national-municipal relationship
vis--vis the Bolsa Famlia.
The MDS has built into the Bolsa Famlia multiple mechanisms for citizen oversight, starting with easy public access to a complete list of beneciaries in a given
city.67 Heightening the incentive to report malfeasance is the zero-sum situation that
the municipal quota produces: one familys receipt of a grant reduces anothers ability
to receive one. Residents have several channels to report suspected fraud, including
a toll-free number in Braslia or a local municipal council.
Our eldwork uncovered an example of how the Bolsa Famlias re alarm
features can work. In Pau Brasil, the leader of a citizens group, Friends of Justice
Association, learned that nearly 200 utterly destitute families were not receiving the
Bolsa. He acknowledged being especially motivated to act after learning that the
federal governments municipal quota meant that every undeserving recipient hurt
a deserving familys chances of enrolling in the program. Using the list of existing
Bolsa beneciaries, he identied and reported suspicious cases to the municipal
department of social assistance and to Braslia (ofcials in the MDS). A public
prosecutor undertook an investigation, which resulted in the removal of a municipal
administrator and a reassessment of need among families in the municipality. Notably,
local media covered the story.68
Finally, federal audits constitute another important mechanism that has enabled
Braslia to transfer resources while preventing subnational ofcials from capturing
and diverting them to their clienteles. As discussed by Marcus Melo, the Federal
Internal Audit Ofce has made signicant efforts in recent years to detect shenanigans
in the transfer of federal funds to municipalities, especially in social programs. With respect
to the Bolsa Famlia, federal auditors have focused on clientelism and corruption in the
local management of the registration process, the last bastion of local control; auditors
have discovered cases of fraud but no major schemes of manipulation.69
While our research suggested several measures that limit clientelistic usage of
the Bolsa Famlia, the portrayal would not be complete without acknowledging that
these effects are not uniform. Technical improvements cannot assure a clean sweep
of all problems. Given the diverse degrees of administrative capacity found among
the countrys 5,000 municipalities, there is sometimes slippage and clientelist intervention. For example, in Pau Brazil we observed political criteria at work in the
appointment of the municipal head of the Bolsa Famlia ofce (the mayor is typically in charge of this appointment). An audit revealed that under one administrator,
whose appointment was blatantly political, at least 100 ineligible families had enrolled,
57

Comparative Politics October 2013

including friends of the mayor and members of the Bolsa Famlia staff. The ineligible
families were later removed, however.70
There is also the question of whether the nonclientelistic character of the Bolsa
Famlia depends on a programmatic party being in power in Braslia. Would the
bureaucratic checks described above be subverted if a party inclined toward
clientelism occupied federal power? In theory such a party could recongure the
MDS and throw out the technocrats who have run the ministry until now. We deem
this scenario unlikely. Any turn toward a politically based distribution of grants would
surely be picked up by the Brazilian media, which have scrutinized the program and
cast it in a more positive light as its terms have tightened.71 Such a shift would also
carry electoral risks due to the possible exclusion of people who have been receiving
the Bolsa as a social right and to the objection of higher income nonrecipients,
whose support for the program appears to hinge on its tight controls.
Conclusion
It is rare to nd well-operated social programs in the developing world. One of
the foremost challenges for democratic governance in such contexts is to provide
the poor with basic services while avoiding the predatory practices of politicians.
Meeting this goal becomes even more difcult in large federal countries where local
cooperation is necessary. Drawing on poor peoples experiences and perceptions, we
explore whether the Bolsa Famlia represents yet another example of social assistance that reinforces local patron-client exchanges or whether it promises a new
rights-based form of social protection that is insulated from political interference.
Given the programs size and scope, together with the federal-municipal cooperation
it requires, there are solid reasons to think that local politicians would get their
hands on the benets.
The study was designed as a crucial case of the Northeastern region, where
politicians would be most likely to use the program for their own political benet.
In the survey, many respondents attested to a considerable amount of quid pro quo
political activity in their cities, (model 1); some admitted to politicians trying to buy
their votes in the most recent election (model 2); yet very few regarded the Bolsa
Famlia as grist for clientelist networks (model 3). In other words, the overwhelming
majority was quick to perceive the Bolsa Famlia as a non-clientelistic form of
poverty relief, very different from the usual political wheeling and dealing. Discussions in the focus groups echoed these ndings. Participants attested to the continued
vitality of clientelism in general in their cities. Yet, across the boardeven in cities
known for high levels of malfeasancebeneciaries talked about the program in
ways that suggested its insulation from vote buying and other clientelistic dynamics.
The study thus concludes that good government in the tropics is possible.
What explains this remarkable nding? Broader comparative research is surely
needed to determine the characteristics of social programs that make them more or
58

Natasha Borges Sugiyama and Wendy Hunter

less available for clientelism. Based on our study of the Bolsa Famlia, we suggest
preliminarily what some of those characteristics might be, recognizing that this treatment falls short of a full-scale research project that examines variation in the features
of different social programs. Federal improvements in policy design appear to be
important in keeping social programs above the local political fray and in empowering
ordinary citizens. Our research to date suggests the promise of a central role for
national authorities in maintaining tight control over the terms of enrollment, providing a direct transfer of resources to recipients, instituting bureaucratic channels
for problem solving, and making available citizen oversight mechanisms to detect
and report wrongdoing. Notably, our ndings on the Bolsa Famlia, a national
program, reinforce a conclusion reached by Tendler with respect to the state government in Cear and its role in municipal health care. Rather than success being a
product of decentralization, it has more to do with something done by central, rather
than local, government.72
Our ndings have implications that go well beyond Brazil. Other developing
nations need to develop social protection programs that do not become enmeshed
in patron-client networks. Increasingly, they have sought to address poverty alleviation through the adoption of CCTs, much like Bolsa Famlia.73 We hope that our
ndings provide a road map for how social assistance can be insulated from patronclient networks.
A question for the long term is whether programs of this kind, especially
those that require considerable local administration, will eventually erode if
not dismantle altogether the norms and practices of subnational clientelism.
Whether this aspiration will be achieved is a matter for future research. For now,
however, we know that the principles embodied in the Bolsa Famlia have brought
poor Brazilians one step closer to achieving the goals of social inclusion and
meaningful citizenship.

Appendix
Table 1A

Survey Variables, Measurement, and Descriptive Statistics (Weighted)

Variable
Independent Variables
Jaboato dos Guararapes
Pau Brasil
Camaragibe
Bolsa Famlia Participant
Age
Education
Male

Measurement Frequency Min. Max

Dichotomous
Dichotomous
Dichotomous
Dichotomous
Continuous
Continuous
Dichotomous

404
258
461
726
76

0
0
0
0
16
0
0

1
1
1
1
82
16
1

Mean

Standard
Deviation

.3596
.2293
.4108
.6471
38.8396
5.7494
.0681

.4801
.4207
.4922
.4781
12.5251
3.6796
.2520

59

Comparative Politics October 2013

NOTES
Thanks go to David A. Armstrong, Javier Auyero, Daniel Brinks, Zach Elkins, Cristiano Ferraz, Ken
Greene, Evelyne Huber, Herbert Kitschelt, Ral Madrid, Cssio Muniz, Timothy J. Power, Maria Cristina
Raposo, Cibele Rodrigues, and Kurt Weyland. The authors would also like to thank the anonymous
reviewers for their valuable comments.
1. Robert R. Kaufman and Joan M. Nelson, eds., Crucial Needs, Weak Incentives (Washington, D.C.:
Woodrow Wilson Center Press, 2004); Joan M. Nelson, Reforming Health and Education (Washington/
Baltimore: Overseas Development Council/Johns Hopkins University Press, 1999).
2. Govindan Parayil, ed., Kerala: The Development Experience (London: ZED Books, 2000); V.K.
Ramachandran, Human Development Achievements in an Indian State, in Dharam Ghai, ed., Social
Development and Public Policy (New York: St. Martins Press, 2000); Judith Tendler, Good Government in
the Tropics (Baltimore: Johns Hopkins University Press, 1997).
3. Alberto Daz-Cayeros and Beatriz Magaloni, Aiding Latin Americas Poor, Journal of Democracy,
20 (October 2009), 3649, 40; James W. McGuire, Social Policies in Latin America, in Peter R.
Kingstone and Deborah J. Yashar, eds., Routledge Handbook of Latin American Politics (New York:
Routledge, 2012).
4. For an overview, see Anthony Hall, Brazils Bolsa Famlia, Development and Change, 39 (September 2008), 799882; Kathy Lindert et al., The Nuts and Bolts of Brazils Bolsa Famlia Program,
Social Protection Discussion Paper No. 0709 (Washington, D.C.: World Bank, 2007).
5. For example, Kathleen Bruhn, Social Spending and Political Support, Comparative Politics,
28 (January 1996), 15577; Matthew L. Layton and Amy Erica Smith, Social Assistance Policies and the
Presidential Vote in Latin America, in AmericasBarometer Insights (Nashville: Vanderbilt University, 2011).
6. Brian J. Fried, Distributive Politics and Conditional Cash Transfers, World Development,
40 (May 2012), 104253; Wendy Hunter and Timothy J. Power, Rewarding Lula, Latin American
Politics and Society, 49 (Spring 2007), 130; Elaine Cristina Licio, Lucio R. Renn, and Henrique
Carlos de O. de Castro, Bolsa Famlia e voto na eleio presidencial de 2006, Opinio Pblica, 15,
1 (2009); Cesar Zucco, The Presidents New Constituency: Lula and the Pragmatic Vote in Brazils 2006
Presidential Election, Journal of Latin American Studies, 40 (2008), 2949.
7. Alfred P. Montero, No Country for Leftists? Clientelist Continuity and the 2006 Vote in the Brazilian
Northeast, Journal of Politics in Latin America, 2 (January 2010), 11353; Tracy Beck Fenwick, Avoiding
Governors: The Success of the Bolsa Familia, Latin American Research Review, 44, 1 (2009), 10231.
8. Ariel Fiszbein and Norbert Schady, Conditional Cash Transfers: A World Bank Policy Research
Report (Washington, D.C.: World Bank, 2009); Lindert et al.
9. Javier Auyero, From the Clients Point(s) of View, Theory and Society, 28 (April 1999), 297334,
297; Susan C. Stokes, Is Vote Buying Undemocratic? in Frederic Charles Schaer, ed., Elections for Sale
(Boulder, CO: Lynne Rienner, 2007).
10. Scott Desposato, How Does Vote Buying Shape the Legislative Arena? in Schaer, 101.
11. Javier Auyero, The Logic of Clientelism in Argentina, Latin American Research Review, 35,
3 (2000), 5581; Ernesto Calvo and Victoria Murillo, Who Delivers? American Journal of Political
Science, 48 (October 2004), 74257; Philip Keefer and Razvan Vlaicu, Democracy, Credibility, and
Clientelism, in Policy Research Working Paper (Washington, D.C.: World Bank, 2005); Herbert Kitschelt
and Steven I. Wilkinson, eds., Patrons, Clients, and Policies (New York: Cambridge University Press, 2007);
Simeon Nichter, Vote Buying or Turnout Buying? American Political Science Review, 102 (February 2008),
1931.
12. Luis Roniger, Political Clientelism, Democracy, and Market Economy, Comparative Politics,
36 (April 2004), 35375, 35354. Vote buying should be distinguished from constituency services
provided by members of Congress. We do not consider as clientelism antipoverty policies that gain the
programmatic vote of the poor.
13. Frederic Charles Schaer, Why Study Vote Buying, in Schaer, 56.
14. Steen W. Schmidt et al., Friends, Followers, and Factions (Berkeley: University of California
Press, 1977); Kitschelt and Wilkinson; Luis Fernando Medina and Susan C. Stokes, Monopoly and
Monitoring, in Kitschelt and Wilkinson; Stokes, Is Vote Buying Undemocratic?; Susan C. Stokes, Political
Clientelism, in Carles Boix and Susan C. Stokes, eds., The Oxford Handbook of Comparative Politics (New
York: Oxford University Press, 2007).

60

Natasha Borges Sugiyama and Wendy Hunter

15. Kitschelt and Wilkinson, 25; Stokes, Political Clientelism; Kurt Weyland, Democracy without Equity
(Pittsburgh: University of Pittsburgh Press, 1996).
16. Stokes, Is Vote Buying Undemocratic? 84.
17. On the eects of income inequality, see James A. Robinson and Thierry Verdier, The Political Economy
of Clientelism, CEPR Discussion Paper No. 3205 (London: Centre for Economic Policy Research, 2002).
18. James C. Scott, Patron-Client Politics and Political Change in Southeast Asia, in Steen W. Schmidt
et al., eds., Friends, Followers, and Factions (Berkeley: University of California Press, 1977).
19. Medina and Stokes, Monopoly and Monitoring, 77.
20. Brian M. Faughnan and Elizabeth J. Zechmeister, Vote Buying in the Americas, in AmericasBarometer
Insights (Nashville: Vanderbilt University, 2011).
21. Sydel F. Silverman, Patronage and Community-Nation Relationships in Central Italy, in Schmidt et al.
22. Ronald P. Archer, The Transition from Traditional to Broker Clientelism in Colombia, Working Paper
No. 140 (Notre Dame: Kellogg Institute of International Studies, 1990); Auyero, From the Clients Point(s)
of View; Auyero, Logic of Clientelism in Argentina.
23. Michael Lipsky, Street-level Bureaucracy (New York: Russell Sage Foundation, 1980).
24. For a state-level breakdown of where Bolsa Famlia recipients live, see Ministrio do Desenvolvimento
Social e Combate Fome (MDS), Bolsa Famlia, http://www.mds.gov.br/bolsafamlia.
25. On most likely case study designs, see Henry E. Brady and David Collier, Rethinking Social Inquiry,
2nd ed. (Lanham: Rowman Littleeld Publishers, 2010), 339.
26. World Bank, World Development Report 2008, (Washington, D.C.: World Bank, 2007), 59.
27. Instituto Brasileiro de Geograa e Estatstica (IBGE), Pesquisa Nacional por Amostra de Domiclios,
Srie: PD226-Situao urbana ou rural do domiclio (Braslia: IBGE, 2009).
28. In 2000, poverty was dened as a per capita household income of less than one half of a minimum
wage (R$75.50) and indigence as less than one quarter of a minimum wage (R$37.75).
29. The national average for Brazil is 14.94 percent. IBGE, Pesquisa Nacional por Amostra de
Domiclios. On rural poverty, see World Bank, World Development Report 2008.
30. The Gini coecient for the Northeast is .555. For other regions, see IBGE, Pesquisa Nacional por
Amostra de Domiclios (Braslia: IBGE, 2005).
31. Lucia Maria Modesto Pereira, Interview, July 1, 2011; Amlia Cohn and Eduardo Fagnani, Proteo
Social, Garantia de Direitos, e Gerao de Oportunidades, in Jos Celso Cardoso Jr. and Carlos Henrique
Romo de Siqueira, eds., Complexidade e Desenvolvimento Vol. 3: Dilogos para o Desenvolvimento (Rio de
Janeiro: Instituto de Pesquisa Econmica Aplicada, 2011), 160.
32. Barry Ames, Political Survival (Berkeley: University of California Press, 1987); Richard Graham,
Patronage and Politics in Nineteenth-Century Brazil (Palo Alto: Stanford University Press, 1990); Victor
Nunes Leal, Coronelismo (New York: Cambridge University Press, 1997).
33. Frances Hagopian, Traditional Politics and Regime Change in Brazil (New York: Cambridge University
Press, 1996).
34. Barry Ames, The Deadlock of Democracy in Brazil (Ann Arbor: University of Michigan Press, 2001), 256.
35. Pesquisa do Ibope mostra o que o brasileiro pensa do processo eleitoral, Globo, Sept. 21, 2010;
IBOPE, Pesquisa de Opinio Pblica e Poltica, in JOB101629 (2010); IBGE, Cidades Database
(Braslia: IBGE, 2003).
36. Civil servants who obtain their jobs through a competitive public exam are insulated from administrative
turnover when new mayors begin their terms of oce.
37. Bolsa Famlia sustenta novo voto de cabresto no Nordeste, Folha de So Paulo, Sept. 24, 2008.
38. IBGE, Cidades Database. Ipeadata, Pobreza Pessoas Pobres database (Braslia: IBGE, 2000)
reports dierent poverty gures.
39. Ibid. Note that the percentage of the population and the percentage of households are not identical.
Poorer households tend to contain more people.
40. Ibid.
41. Surveys were standard across all three cities, with the exception of the names of local politicians
(mayor and governor). Out of 1104 responses, only 4 individuals declined to answer one of the vote
buying questions. For comparable analysis, we drop those respondents from our statistical analysis.
42. We held four groups (two male/two female) in Camaragibe, four groups in Jaboato (one male/three
female), and three groups in Pau Brasil (one male/two female).
43. Focus group testimonies are anonymous to protect participants privacy.
44. This sampling method is employed in pre-election polls in Great Britain due to the short election period.
See Herbert F. Weisberg, The Total Survey Error Approach (Chicago: University of Chicago Press, 2005).

61

Comparative Politics October 2013

45. Only 60 percent of respondents said they had a phone number for contact. Of those, most oered
numbers of public phone booths or relatives cell phones.
46. T.M.F. Smith, On the Validity of Inferences from Non-Random Samples, Journal of the Royal
Statistical Society, 146, 4 (1983): 394403.
47. Making inferences to the set of observations not selected into the sample with these data would
require more information about the observations, both in and out of the sample. See Smith.
48. We draw these percentages from the Ministrys Secretaria Nacional de Renda e Cidadania, which
provides data on the number of families that applied for the program and met the program eligibility
requirements and the number of families that actually receive the Bolsa.
49. Amartya Sen, Development as Freedom (New York: Alfred A. Knopf, 1999); United Nations,
Millennium Development Goals Report (New York: United Nations, 2010).
50. One participant in Camaragibe reported the range was between R$15-R$50; Focus group testimony,
Camaragibe, Mens Group, June 12, 2009. Elsewhere, a participant reported her vote cost R$5; Focus
group testimony in Jaboato, Womens Group, June 13, 2009.
51. A cesta bsica typically contains basic household necessities for one month.
52. Focus group testimony, Pau Brasil, Womens Group, June 9, 2009.
53. Focus group participant, Pau Brasil, Womens Group, June 9, 2009.
54. Focus group participant, Jaboato, Womens Group, June 25, 2009.
55. Many technocrats were recruited from the Escola Nacional de Administrao Pblica.
56. Strictly speaking, targeting the poorest municipalities does not necessarily mean the poorest families
receive the grant. Nor does it mean that beneciarieseven if they are well chosenreceive the benet
free of clientelistic interference.
57. For further details, see Lindert et al. The average value of the grants received in the rst six months
of 2006 was US$30.
58. MDS, Bolsa Famlia (Braslia: MDS, 2006).
59. Ibid., 9.
60. In small cities without such a bank, beneciaries can go instead to a Casa Lotrica (Lottery Store),
which serves as an informal banking outpost.
61. Focus group testimony, Pau Brasil, Womens Group, June 9, 2009.
62. MDS, Bolsa Famlia: Agenda de Compromissos da Famlia, 7.
63. Focus group testimony, Pau Brasil, June 9, 2009.
64. Mathew D. McCubbins and Thomas Schwartz, Congressional Oversight Overlooked, American
Journal of Political Science, 28 (February 1984), 165179, 166.
65. Tendler.
66. Ibid., 152.
67. Only the names of beneciaries and their respective cities of residence are publicly available. Other
personal data are restricted.
68. Digenes Vierra dos Santos, Interview, June 25, 2009.
69. Marcus Andr Melo, Unexpected Successes, Unanticipated Failures: Social Policy from Cardoso to
Lula, in Peter R. Kingstone and Timothy J. Power, eds., Democratic Brazil Revisited (Pittsburgh: Pittsburgh
University Press, 2008), 183.
70. Luis Alberto Azevedo Perreira, Interview, June 24, 2009.
71. Kathy Lindert and Vanina Vincensini, Social Policy, Perceptions and the Press, Social Protection
Discussion Paper 1008 (Washington, DC: World Bank, 2010).
72. Tendler, 43.
73. Fiszbein and Schady, Conditional Cash Transfers.

62