Vote Buying SEARCanare Mendoza Lopez 2018

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/323082978
An Empirical Analysis of Vote Buying Among the Poor: Evidence from

Elections in the Philippines (forthcoming in Southeast Asia Research)
Article · February 2018
CITATION READS
1 33,179
3 authors:
Tristan Canare Ronald U. Mendoza

Ateneo de Manila University Ateneo de Manila University
47 PUBLICATIONS 213 CITATIONS 151 PUBLICATIONS 2,148 CITATIONS
SEE PROFILE SEE PROFILE
Mario Antonio Guatelara Lopez

Asian Institute of Management
7 PUBLICATIONS 30 CITATIONS
SEE PROFILE
All content following this page was uploaded by Ronald U. Mendoza on 04 August 2019.
The user has requested enhancement of the downloaded file.

Article
South East Asia Research

2018, Vol. 26(1) 58–84
An empirical analysis of ª SOAS University of London 2018
Reprints and permission:
vote buying among the sagepub.co.uk/journalsPermissions.nav

DOI: 10.1177/0967828X17753420
journals.sagepub.com/home/sear
poor: Evidence from elections
in the Philippines
Tristan A Canare
Ateneo de Manila University, Philippines
Ronald U Mendoza
Ateneo de Manila University, Philippines
Mario Antonio Lopez

Asian Institute of Management, Philippines
Abstract
Recent literature suggests that the poor are more likely to be targeted for vote buying and to sell
their votes. However, there is limited empirical analysis on the patterns of vote buying among low-
income voters. This paper attempts to fill this gap using a survey conducted in Metro Manila,
Philippines after the 2016 elections. Data analysis shows that vote buying among the poor is indeed
very common, but the incidence varies depending on the vote buying type. The most prevalent
form uses more benign goods such as food and clothing, but offers of money is still reported by
more than a quarter of respondents. Different vote-buying types also have different correlates,
including some socio-economic factors, suggesting that it is a finely targeted activity. In addition,
money vote buying is predominant in tight elections, but buying votes using non-monetary offers is
more common when there is a clear winner even before the election. Most of those who were
offered accepted the goods or money, but only about two-thirds voted for the candidate. In
addition, evidence suggests that the good or money is not the deciding factor in voting for the
candidate. This supports the premise that vote buying is just part of a bigger effort by politicians to
build clientelism and patronage among his/her constituencies. Dependency and loyalty is merely
punctuated by election-related transfers, rather than an effort to completely change votes.
Keywords
Vote buying, elections, election fraud, philippines
Corresponding author:
Ronald U Mendoza, School of Government, Ateneo de Manila University, Quezon City, Philippines.
Email: ronmendoza@post.harvard.edu
Canare et al. 59
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
ic
ia
ay
nd
ya
a
ca
a
l
a
ga
tin
al
bi
an
bl
er
fri
gu
en
la
ne
m
pu
ig
tsw
en
ai
K
ra
te
lo
Se
N
Re
Th
rg
ua
Pa
h
Co
Bo
A
ut
G
an
So
ic
in
om
D
Figure 1. Estimated prevalence of vote buying across selected democracies.

Sources: For Latin America, the source is the Latin American Public Opinion Project (LAPOP) – Americas
Barometer Survey 2012. The survey question used in generating the graph is: “Has a candidate or someone
from a political party offered you something?” Possible answers are “Frequently,” “Rarely,” and “Never.” The
figures in the graph are a combination of “Frequently” and “Rarely” responses. For Africa, the source is the
Afrobarometer Survey 2014/2015. The question used is: “In your opinion, how often do the following things
occur in this country’s elections: Voters are bribed?” Possible answers are: “Never,” “Sometimes,” “Often,”
“Always,” and “Don’t Know”. The figures in the graph are a combination of “Sometimes,” “Often,” and
“Always” responses. For Thailand, the source is Finan and Schechter (2012), citing Phongpaichit et al. (2000).
Introduction
By now there are numerous studies of how policymakers target low-income individuals and
households for more efficient social protection and public service provision (Barrientos and
Hulme, 2008; Devereux, 2002; Keefer and Khemani, 2004). Yet, another form of targeting may
also be taking place—this time involving vote buyers and the poor. Vote buying is rampant in
many developing democracies. As seen in Figure 1, some of the highest vote buying incidences can
be found among African countries, with Latin American nations posting lower vote buying rates,
and Thailand situated in the middle. Vote buying brings a host of costs to society: it can weaken
accountability in a democratic society and impede the growth of institutions needed for democracy
(Kitschelt et al., 2010; Lyne, 2007; Stokes, 2005). Vote buying is also correlated with greater
corruption and inefficient government institutions (Hicken and Simmons, 2008; Keefer, 2007;
Kitschelt and Wilkinson, 2007). However, while several studies have found evidence that poor
voters do get targeted more for vote buying (Carkoglu and Aytac, 2015; Jensen and Justesen, 2014;
Schaffer, 2005), there has so far been scant analysis of specific vote buying patterns and the
correlates of vote buying within low-income groups.
As a contribution to the nascent literature, this study examines patterns of vote buying among
poor voters in Metro Manila, Philippines, using a survey dataset produced closely following the
2016 national and local elections. Specifically, it will examine the correlates of—the factors linked
60 South East Asia Research 26(1)
Party preference of recipient

vis-a-vis Party Offering Reward
Indifferent or
Favors Party
Favors Opposition
“Rewarding Loyalists” “Vote Buying”
Inclined to Vote Requires no monitoring Requires monitoring of
Reward Recipient
vote choice
Inclined to Vote or
“Turnout Buying” “Double Persuasion”
Not Vote
Inclined Not to Vote Require monitoring of Requires monitoring of
turnout vote choice and turnout
Figure 2. Framework of vote vs turnout buying.

Source: Nichter (2008).
to—vote buying, hopefully uncovering evidence of how this is targeted among low-income voters.
This study also looks at the correlates of other aspects of vote buying, including accepting the offer
and voting for the candidate who made the offer.
The main focus is on the poor, as extensive anecdotal evidence in the Philippines suggests that
this could be the group most prone to sell their votes, in large part because of their needs. This is
supported by Schaffer (2005), who found evidence that vote buying is particularly effective among
low-income voters. There are studies with similar results using data from Argentina (Brusco et al.,
2004) and Turkey (Carkoglu and Aytac, 2015). In addition, similar empirical research found
evidence that selling of votes is more prevalent among poor voters in Argentina (Stokes, 2005),
Egypt (Blaydes, 2006), and Nigeria (Bratton, 2008), and that poor people are more likely to be
targeted by vote buying in Africa (Jensen and Justesen, 2014), the Philippines (Schaffer, 2005), and
Turkey (Carkoglu and Aytac, 2015). While there have been several studies concluding that low-
income voters are disproportionately targeted for vote buying, there has been little evidence
presented regarding specific targeting patterns within this group. Analyzing the correlates of vote
buying among the low-income group would provide insights on who among the vulnerable
members of society are more likely to be targeted by vote buyers, who are more likely to accept this
vote buying offer, and who are more probable to keep their end of the bargain by voting for a
candidate who gave them money. This can potentially help policy makers in designing and tar-
geting programs on voter education.
This article argues that vote buying among poor voters in Metro Manila appears to be a finely
targeted activity, as evidenced by some factors turning out to be significant correlates of vote
buying. Moreover, various things of value can be used to buy votes, while money is used primarily
in tight elections. Based on the survey that was the basis of this article, the evidence suggests that in
the May 2016 national and local elections, vote buyers offered an average of around PhP500 per
targeted voter in poor communities in Metro Manila.
Related literature
Early work on vote buying emphasized the distinction between paying to change voting decisions
from one candidate to another, versus paying to simply encourage voter turnout (or even refraining
from voting, if the vote buyer wishes to erode the opponent’s voter base). Nichter (2008), for
example, developed a theoretical framework distinguishing vote buying from turnout buying (see
Canare et al. 61
Figure 2). In particular, he considered how vote buying involves targeting weakly opposed voters,
while turnout buying focused on targeting unmobilized strong supporters. This framework helped
to clarify the possible different strategies behind vote buying efforts—likely influenced by the
political context, the ability to hold voters accountable to the agreement, and other factors. Using
data on Argentina, Nichter’s empirical analysis found greater evidence that campaign operators
targeted for stronger turnout, rather than for changing voter preferences.
Moreover, Brusco et al. (2004) implemented a vote buying survey in Argentina in 2002, to
examine the possible factors associated with vote buying in that country. They hypothesized that
vote buying is best seen as a form of probabilistic selective incentive—good(s) that a voter receives
in the future only with her/his continued support for the political party buying her/his vote. This
framework, they argued, proves effective in holding the voter to the vote buying agreement, due to
factors such as social networks linking the voter to the party, the ability to monitor the voter such as
attendance in rallies and meetings, and the conducive environment of smaller towns and cities
whereby voters are more easily known and monitored.
Alternatively, they also acknowledged that norms of reciprocity could be leveraged to ensure
the vote buying agreement is carried out by the voter. These authors found evidence that the
Argentine poor are most likely targeted with clientelistic appeals, including vote buying. Political
party affiliation was also critically important, given the types of policies some parties were
associated with—for instance, receiving a handout from a Peronist increased the likelihood that a
voter would vote for a Peronist presidential candidate in 1999 from just under to well over 50%
(Brusco et al., 2004).
The importance of reciprocity—or the likelihood that the vote seller will vote for the vote
buyer—in vote buying was also emphasized by Finan and Schechter (2012). According to these
authors, vote buyers target reciprocal voters; and this targeting can be performed by political
operators or “middlemen” who know and are familiar with the communities assigned under their
watch. Indeed, in a survey conducted among voters and middlemen in Paraguay, the authors found
evidence that reciprocal persons do receive more offers of vote buying.
Recent work, including by Vicente and Wantchekon (2009) and Vicente (2014), also turn to
data from field experiments on election campaigns. These authors utilize field experiments in
Benin and Sao Tome and Principe to examine the extent to which vote buying activities are
effective in garnering votes for politicians. They also examine the extent to which information
campaigns against vote buying could curb the effectiveness of this practice. Their experimental
design included agreements with politicians to randomize the content of their campaigns
(including an agreement not to diffuse clientelistic or purely redistributive messages and vote
buying in some of their stronghold villages), as well as a randomized information campaign against
vote buying1 which allowed them to generate rich empirical results. They found evidence that
clientelistic messages were more effective among men, and were approaches that were deployed
more effectively by incumbent politicians (i.e. those with control over public services provision).
Nevertheless, the effectiveness of clientelistic approaches by incumbents was counteracted by vote
buying strategies used by political challengers. These were successful in increasing voter turnout.
Furthermore, the anti-vote buying campaign eroded the effectiveness of more clientelistic
approaches, but only among women voters. Field experiments on vote buying have also been
1. The intervention was essentially a leaflet distributed to 10,000 households containing the slogan: “Do not
let your conscience be bent by vote buying—Your vote should be free and in good conscience.”
recently conducted in the Philippines. Cruz et al. (2016) found evidence that voters with more
information are more targeted for vote buying. Because voters know more, candidates respond by
buying more votes. In a related study, Hicken et al. (2014) tested which are the more effective anti-
vote selling strategies. They found evidence that encouraging voters to vote for their preferred
candidate in spite of accepting a vote buying offer from another candidate is less effective than
encouraging voters not to sell their votes.
In addition, research on Latin American voters also emphasized the importance of trust. For
instance, Carreras and Irepoglu (2013) utilized the 2010 wave of the Latin America Public Opinion
Project (LAPOP) surveys to analyze how low trust in elections and exposure to vote buying affects
electoral participation in that region. They found evidence that perceptions of unfairness in
elections led to lower willingness to participate. On the other hand, receiving vote buying
incentives to participate counteracted this effect.
Broad social and economic conditions, as well as the interaction between political and eco-
nomic elites, could also possibly influence election fraud and vote buying prevalence. Chaves,
Fergusson, and Robinson (2015), for instance, developed a theoretical model whereby political and
economic elites engage in or tolerate election fraud. They predict that “fraud will be lower and
public good provision higher when land inequality is higher, the overlap between elites lower,
and the strength of the state higher” (Chaves et al., 2015: 124). They find evidence to support their
hypothesis using data from Colombia’s 1992 presidential elections.
Recent vote buying studies have also turned to new estimation methods to try to assess the correct
vote buying estimate, considering that survey respondents could face a social desirability bias. For
example, based on a survey of over 1000 respondents after the 2008 municipal elections in Nicaragua,
Gonzalez-Ocantos et al. (2011) estimated vote buying patterns in that country using the list method
and found that 24% of registered voters in Nicaragua were offered a gift or service in exchange for
votes. Nevertheless, only 2% of survey respondents reported the behavior when using a direct
question on vote buying. They also found evidence of two key vote buying linkages in the Nicaraguan
context. First, there were strong perceptions that the government could monitor citizens’ votes. And
second, state actors easily engaged citizens through the government-sponsored (and Sandinista-
controlled) Consejos de Poder Ciudadano (Citizen Power Councils, CPCs) which distributed goods
for clientelistic purposes. Both these mechanisms made it easier for vote buying to take place.
Moreover, Carkoglu and Aytac (2015) turn to an expanded version of a list experiment to
overcome possible social desirability bias in responses to vote buying survey questions. Using the
list method, they found that up to one-third of the electorate was targeted for vote buying in the
2011 Turkish parliamentary elections—over twice the estimated vote buying admitted by voters
surveyed using direct survey questions. They also utilized a population-based list experiment with
an estimator to analyze what types of voter groups are targeted by vote buyers. They found evi-
dence that less-educated, low-income, and urban residents were the preferred targets, along with
strong partisans of the ruling political party (i.e. an approach that was meant not to curb votes but
increase turnout).
Vote buying in the Philippines

The vote buying process and why the poor matter
Before the turn of the current century, there had not been many studies on vote buying. As Lehoucq
(2002) points out, only a few papers on vote buying existed then and even fewer used actual vote
Canare et al. 63
buying data. Although the last decade and a half has seen an increase in papers examining vote
buying, with a number of studies considering vote buying in the Philippines, in South East Asia,
and other developing countries, there have been few, if any, empirical papers on vote buying that
have specifically focussed on low-income voters. As discussed earlier, there is a special interest in
studying vote buying among the poor because there is literature suggesting that vote buying is
more effective among low-income voters and that this is the oft-targeted group.
Vote buying entails an implicit contract between the vote buyer and the seller. What makes it
different compared to most legal contracts is that vote buying and selling, as an illegal activity,
cannot be enforced through the usual legal means. Contract enforcement in vote buying involves
some level of trust between the buyer and the seller that the latter will keep his/her end of the
bargain. Schaffer (2002a) outlines three ways by which candidates may expect compliance from
vote sellers. The first is instrumental compliance, wherein the vote seller changes or does not
change his/her vote as a direct effect of the money or other material things given to him/her.
The second is normative compliance, wherein the seller changes or does not change his/her vote
because the money from the candidate persuades him/her of the candidate’s goodness, or because
the voters feels obliged to vote for the candidate because of the money. Third is coercive com-
pliance, wherein the vote buyer coerces the seller to change or not change his/her vote.
How vote buyers hope to enforce the vote buying contract changes with the political context. In
an era of automated elections in the Philippines, which began in 2010, monitoring vote buying
compliance has become more difficult. The old methods of monitoring compliance during the time
of manual elections (see discussion below) have become obsolete, and, as discussed by Cruz
(2013), new methods of monitoring instrumental compliance are difficult to implement. Fur-
thermore, instrumental compliance is not the preferred vote buying method of most established
politicians and political operatives, who tend to favor a normative compliance approach. Cruz’s
(2013) interview of politicians and political operatives in the Philippines indicates that vote buying
is most effective when buyers exploit the social norms held to by some voters that they have to
reciprocate the money given to them by the candidates by voting for them. This is not a finding
exclusive to Cruz’s study. Finan and Schecter (2012) found evidence that voters who are more
likely to display these norms of reciprocity are more likely to be targeted for vote buying. For vote
buying candidates, one of the best ways to find reciprocity is to turn to voters with whom they have
established dependency and patronage relationships.
Hicken (2006) and Schaffer (2002a) also emphasized the role of patronage and clientelist
politics in enhancing the occurrence and effectiveness of vote buying. When there is a patronage
and clientelist relationship between voters and candidates, vote buying serves to enhance the
effects of these relationships. As compliance monitoring of individual vote buying is difficult,
candidates tend to focus on group monitoring. A survey in Isabela province, Philippines found
evidence that individuals with more social connections are more likely to be targeted for vote
buying (Cruz, 2013). Moreover, group monitoring is easier among patron and client voters.
Our own interview with political operatives in the Philippines confirms this kind of strategy.
According to our sources, there are two common types of vote buying in the country: “retail” and
“large-scale.” The former pertains to buying votes from individual voters while the latter is the
group vote buying described above. Retail vote buying accounts for a much smaller share, rein-
forcing the findings in literature cited earlier that vote buyers focus on group vote buying. This is
because it is more difficult to monitor compliance in retail vote buying and targeting of reciprocal
voters is easier with group vote buying. Although monitoring is easier in groups, a politician needs
a large network to implement this. For a targeted group, it is common to assign a person to act as a
monitor. This person, usually the barangay captain (village leader), should be familiar with and
knowledgeable about the group’s dynamics so that monitoring can be implemented effectively.
Ideally, this person should also have some form of authority over the group assigned to him/her.
Our interviews also cohere with findings in the literature on how political patronage, cliente-
lism, and vote buying reinforce each other. According to our sources, the actual act of trying to buy
votes occurs as the election approaches, but it involves a lengthy process that starts way before the
campaign period. Several years before the election, potential vote buyers would start by imple-
menting small mass-oriented programs such as school feeding, sponsoring inter-village sports
tournaments, film showings in the barangay (village) plaza, donating small service vehicles to the
barangay government, and the like. The main goal of these seemingly benign actions is to establish
dependence, patronage, and clientelism with the constituents. One source refers to this part of the
process as “soft” vote buying, while another interviewee refuses to call this phase vote buying.
Eventually, as the election approaches, the process shifts to the actual voting buying phase—the
vote buying as studied in the literature—which involves giving cash and other material things. The
degree of actual vote buying depends on how closely the race is being contested—the closer
the election, the more intensified the actual vote buying.
The workings of vote buying provide another interesting angle on why vote buying among the
poor should be studied. Because low-income households have more needs and less social pro-
tection, it is easier to develop dependency relationships and reciprocity with them. Indeed, Brusco
et al. (2004) and Hicken (2006) argue that poverty positively affects clientelism, patronage, and
vote buying. This could explain the findings of empirical papers that vote buying and selling is
more prevalent among low-income voters (Blaydes, 2006; Bratton, 2008; Jensen and Justesen,
2014; Stokes, 2005), including in the Philippines (Schaffer, 2005).
The 2016 Philippines election and vote buying

The Philippines holds national and local elections every three years. All local positions (provincial
level, municipal and city level, and district representatives) are held for a three-year term, while the
President, Vice President, and Senators each have six-year terms. The terms of the 24 senators
overlap such that 12 senators are elected every three years. The country is known to have one of the
most colorful elections in the world, attracting media attention because of election-related violence
and allegations of cheating such as vote buying and vote rigging.
Elections are a big source of spending and provide an economic boost in the Philippines. It is not
uncommon for the real Gross Domestic Product (GDP) growth rate to peak during an election year.
Election-related spending spurs demand and this may include vote buying money. The last five
election years produced real GDP growth rate peaks. The last time an election year did not produce
a peak year in GDP growth was in 2001, four years removed from the Asian Financial Crisis and
one year after a severe political crisis in which a former president was forced to resign.
The 2016 Philippine national and local elections were only the third time that automated
elections had been held, the other two being in 2010 and 2013. Previously, voters wrote the
names of their candidates on a ballot and these were counted manually. A common way to
implement the implicit contract between the buyer and the seller was to bring a carbon paper into
the polling booth to produce proof that the seller indeed voted for the buyer. Another was for the
buyer to give the seller a filled out fake ballot which the latter would then cast (Cruz, 2013;
Schaffer and Schedler, 2006).
Canare et al. 65
With election automation, it supposedly became more difficult to enforce the vote buying
contract. The new preferred method became issuance of pre-filled ballots; however, this is difficult
to implement (Cruz, 2013). Forging ballots has become more problematic as they now need to
meet security features that can be read by the vote counting machines. This has contributed to the
situation discussed earlier wherein many politicians and political operatives prefer to take
advantage of developing reciprocity and dependency relationships with voters.
In addition to this, there is another way in which election automation may have affected vote
buying behavior in the Philippines: automation significantly reduces the ability of candidates to
cheat and rig election results, hence they engage more in vote buying. Officials of the Com-
mission on Elections (COMELEC) have pointed out that because candidates cannot get around
the automated vote counting machines to cheat, they instead focus on vote buying (Miasco,
2016; Rodriguez, 2016).
Vote buying beyond the Philippines

The mechanisms of vote buying discussed here are not limited to the Philippines. Vote buying
studies in neighboring South East Asia and other developing countries offer analogous discussions
on how vote buying works. For instance, a scan of vote buying literature in South East Asia shows
that Thailand—where vote buying, as in the Philippines, is a perennial problem—is an oft-studied
case. Hicken (2002) argues that a weak party system—a condition that also exists in the Philippines
—is a major contributor to the prevalence of vote buying in Thailand. Because party labels are
weak and even party members compete with each other, the personal campaign strategies of
individual candidates are more important than those of the party. And among the strategies
available to the candidates, vote buying is usually the least expensive.
There are also similarities in terms of operations and targeting. As in the Philippines, the tar-
geting of vote buyers in aggregate villages and groups of people through an influential inter-
mediary has also been observed in Paraguay, Taiwan, Thailand, Senegal, and Indonesia (Finan and
Schechter, 2012; Fionna, 2014; Schaffer, 2002a).
Methodology
The survey
The primary data used in this study is a survey among low-income voters in Metro Manila, Phi-
lippines. Metro Manila, also known as the National Capital Region, is an agglomeration of 17 cities
surrounding the national capital. The region accounts for almost 13 percent of the country’s
population and 36.6 percent of the 2016 GDP. It is also the country’s most urbanized region.
Respondents were selected using multi-stage area probability sampling. It covers all 17 cities in
Metro Manila, and the number of respondents per city is proportional to the number of registered
voters. The first stage involved randomly selecting low-income communities for each city/
municipality from a list provided by the city governments. The second stage saw systematic
sampling of 10 respondents from each drawn community. Face to face interviews with the
respondents were conducted. The first choice respondent was the household head of each selected
household. In his/her absence, the second choice was the spouse, and the third was any adult
member of the family.
Using survey data in any empirical investigation has some potential limitations, and these are
acknowledged in this article. One potential limitation of survey data is representativeness of the
sample, wherein the drawn observations do not represent the population they are supposed to
represent. In our survey, samples were drawn randomly via multi-stage random sampling. It must
be emphasized that the sample population was low-income voters, rather than all voters, in Metro
Manila during the 2016 elections. Thus, there is a high degree of likelihood that representativeness
was met. The survey limitation that could potentially affect the results of this study is the possi-
bility that the respondent would not answer truthfully if the question pertained to socially unde-
sirable actions or characteristics. This social desirability bias is the potential to commit errors in
self-reported measures to project a positive image of oneself to others or avoid humiliation (Fisher,
1993). Vote buying and selling are illegal; therefore, asking respondents if they received a vote
buying offer could be subject to this limitation. To test for social desirability bias, we included
indirect questioning on vote buying.
One of the most common types of indirect questioning is projective questioning. This involves
asking the respondent to answer from the point of view of other people. Early papers that utilized
this method of questioning include Robertson and Joselyn (1974), Calder and Burnkrant (1977),
and Anderson (1978). Fisher (1993) reviewed several papers that use indirect projective ques-
tioning and concluded that it significantly reduced social desirability bias for questions on socially
undesirable variables and had no effect on socially neutral variables. Chaudhuri and Christofides
(2013) discussed how other forms of indirect questioning can elicit more truthful answers on
socially sensitive topics.
Two sets of instruments are used in the survey underpinning this study. One uses direct ques-
tioning, i.e. respondents are asked if they themselves have received an offer to sell their vote. The
other uses indirect vote buying questions, i.e. respondents are asked if they know someone who has
received an offer to sell their vote. Since the objective of this article is to test for correlates of vote
buying, the indirect questions were used only to validate the results of the direct questions – that is, to
test for large differences in the vote buying incidence as generated using the two types of questions.
The objective of including indirect vote buying questions is to compare the incidences of
different vote buying types that it generates with that of direct questioning. If the recorded vote
buying incidence between the direct and indirect questions hardly differ, then social desirability
bias is not a major concern. As will be discussed later, the recorded vote buying incidences using
the indirect questions are only slightly higher than those using direct questions. We argue that this
slight difference can be explained by network effects. When a respondent is asked if he/she was
offered something in exchange for his/her vote, the question only refers to the respondent him/
herself. In contrast, for indirect questions, the respondent has to think of all other people in his/her
network. Thus, social desirability bias presents only a small problem, if at all, in this research. In a
similar vote buying study conducted in the province of Isabela, Philippines, Cruz (2013) tested for
social desirability bias using a list method, and similarly found no significant under-reporting of
vote buying in using direct questioning. Khemani (2015) also used direct vote buying questions in
the same province. The prevalence of vote buying borne out by these results suggests that it has
almost become socially acceptable.
The instrument contains other questions related to vote buying, such as whether the respondent
voted for the candidate who made the offer, if he/she accepted the offer, and what was offered in
exchange for his/her vote. The instrument also collects information on socio-economic conditions,
demographic information, and questions pertaining to access to information, how the respondents
view the government and certain public services, and their well-being and outlook in life. A total of
400 respondents each for Set 1 (direct) and Set 2 (indirect) instruments were selected. The survey
was rolled out in July 2016, around two months after the national and local elections.
Canare et al. 67
In addition to the above-mentioned points, we also contracted a survey firm whose enumerators
have undergone training and possess extensive experience undertaking surveys in low-income
communities. This firm has well over two decades of experience fielding these surveys in urban
informal communities. This further contributed to our efforts to elicit more truthful answers from
our respondents.
Correlates of vote buying

The general vote buying question in the instrument asks whether the respondent has been offered
anything in exchange (explicitly or implicitly) for his/her vote. As vote buying involves not only
cash, but a range of other things with value, a question is also asked concerning what items have
been offered. These questions could provide evidence for whether vote buying in general is tar-
geted at certain groups within the low-income group, and whether specific forms of vote buying
(money, food, other material things) are also targeted. Testing for the correlates of receiving an
offer to sell one’s vote could shed light on this. Related to this, questions on whether the respondent
accepted the offer and voted for the candidate who attempted to buy his/her vote are also asked.
Testing for the correlates of accepting the offer and of voting for the candidate could provide
evidence of who among the poor is more likely to sell his/her vote.
For each of the four vote buying types specified in the instrument (money; food, rice, and
groceries; other material things; and any help or favor), we implement a probit regression using
indicators such as gender, age, income proxy variables, and proxies for need. To test whether vote
buying is targeted, only indicators that are observable or potentially observable are included among
the independent variables. The closeness of the race in the previous election is also controlled for
using the margin of victory of the winning mayoral candidate as a share of total votes cast for
mayor. To control for the effectiveness of the government, indicators representing the quality of
public services are also included.2
The independent variables included are (with variable names in parenthesis): sex (male), age
(age), employed (employed), average household monthly income (income), dummy for ownership
of house the family lives in (own_house), dummy variable for being a 4Ps (Conditional Cash
Transfer (CCT)) beneficiary (pppp), dummy variable ¼ 1 if household experienced a calamity in
the past year (calamity), dummy variable ¼ 1 if the respondent rates the water service to be good or
very good, (water_good), dummy variable ¼ 1 if the respondent rates the local government per-
formance as good or very good (localgovt_good), number of appliances the household owns
(assets), dummy ¼ 1 if the respondent experienced involuntary hunger (hunger), and margin of
victory of the winning mayoral candidate as a share of total votes cast for mayor (margin_victory).
We also implemented probit regressions examining the factors linked to accepting the vote buying
offer, and voting for the candidate who made the offer, as dependent variables. The independent
variables include income and asset proxies, socio-demographic indicators, and variables that rep-
resent outlook in life and satisfaction with government performance and public services.
The possible correlates tested are: sex (male), age (age), employed (employed), average house-
hold monthly income (income), dummy variable ¼1 if respondent said that his/her family’s life had
2. We are not claiming that the mayoral candidates are the principal actors engaged in vote buying. Our
interviews of key political operatives revealed that mayoral candidates are key players in the election process
at the local government level. They tend to have a strong and readily available network of campaigners and
allies at the local level, and they also offer connections to mass voters for national level politicians.
Table 1. Summary statistics of independent variables.
Variable Obs Mean Std. Dev. Min Max
male 400 0.1875 0.3908 0 1

age 400 44.9850 13.2480 20 79
employed 400 0.5450 0.4986 0 1
income 399 11,844.730 8803.373 0 72,000
children 400 3.3550 2.0517 0 12
owns_house 400 0.6825 0.4661 0 1
pppp 400 0.3350 0.4726 0 1
calamity 400 0.3175 0.4661 0 1
water_good 400 0.6050 0.4895 0 1
localgovt_good 400 0.4450 0.4976 0 1
assets 400 4.7250 1.9504 0 8
hunger 400 0.2925 0.4555 0 1
margin_victory 400 49.6991 38.8145 0.3891 100
elemgrad 400 0.3625 0.4813 0 1
highschool 400 0.3625 0.4813 0 1
posths 400 0.2125 0.4096 0 1
life_improve 400 0.3500 0.4776 0 1
cheating 400 0.1575 0.3647 0 1
new 400 9.4423 8.5711 0 74
internet 400 3.0819 9.7270 0 94
credit 400 0.7000 0.4588 0 1
believe 400 0.1525 0.3600 0 1
assets_acquired 400 1.4400 1.5629 0 8
natlgovt_good 400 0.5025 0.5006 0 1
crime 400 0.4375 0.4967 0 1
happy 400 0.6150 0.4872 0 1
expect_improve 400 0.6475 0.4783 0 1
Source: Authors’ calculations using survey data.
improved or somewhat improved over the past three years (life_improve), dummy ¼ 1 if respondent
witnessed cheating in any of the past elections (cheating), hours spent listening to or reading the news
per week (news), hours spent accessing the internet per week (internet), presence of an immediate
source of credit (credit), dummy variable ¼ 1 if the household experienced a calamity in the past year
(calamity), dummy ¼ 1 if respondent said he/she believes that candidates will know who he/she
voted for in the elections (believe), number of appliances the household acquired over the past three
years (assets_acquired), dummy variable ¼ 1 if the respondent rates national government perfor-
mance as good or very good (natlgovt_good), dummy variable ¼ 1 if the respondent rates local
government performance as good or very good (localgovt_good), dummy variable ¼ 1 if the
respondent rates the water service to be good or very good, (water_good), dummy ¼ 1 if the
respondent said that crime is prevalent or really prevalent in his/her city (crime), dummy ¼ 1 if
respondent said that he/she is very happy or rather happy (happy), and dummy ¼ 1 if respondent said
that he/she expects his/her quality of life to improve in the next three years (expect_improve). Also
included are categorical variables for educational attainment: elementary graduate and attained some
high school units (elemgrad), graduated from high school (highschool), and attained any post-high
Canare et al. 69
Table 2. Selected survey results.
Variable Direct Question Indirect Question
Vote Buying Offer (General) 100.00% 100.00%

Vote Buying Offer (Money) 28.50% 34.85%
Vote Buying Offer (Food, rice, or groceries) 35.25% 41.12%
Vote Buying Offer (Other material things) 54.75% 54.52%
Vote Buying Offer (Help or favor) 17.25% 21.61%
Vote buying not to vote 4.00% 10.75%
Accepted offer 90.75% 97.40%
Voted for the candidate 67.25% 81.52%
Reasons for voting for candidate (among those who voted)
Qualified 94.42% 79.52%
Because of the offer 16.73% 37.65%
Intimidated 0.00% 0.60%
Might lose government benefits 1.12% 4.14%
Reasons for not voting for candidate (among those who
did not vote)
Not qualified 62.28% 48.28%
Because candidate attempted to buy vote 30.70% 33.33%
Note: “Don’t know” answer not included in computing for % share among indirect question respondents.
school education (posths), with less than elementary degree as the base variable. The analysis also
includes a proxy for tightly contested races which is the margin of victory of the winning mayoral
candidate as a share of total votes cast for mayor (margin_victory). The vote buying type is also
controlled for by including dummies of each of the four types of vote buying.
We examine the correlates of receiving a vote buying offer, of accepting the offer, and of voting
for the candidate who made the offer using observations from respondents to the direct question
instrument. Table 1 shows the summary statistics of the independent variables.
Empirical results
This section is divided into four sub-sections. The first part reports the general results of the survey.
The second discusses the correlates of vote buying, while the third and fourth discuss the correlates
of accepting the offer and of voting for the candidate who made the offer, respectively.
Survey results
The first remarkable result is that 100 percent of respondents in the survey answered “Yes” to the
general vote buying question, both in the direct and indirect questions. Since the survey was focused
on a sample of low-income voters in Metro Manila, this result suggests that a vast majority of this
group may have been targeted for vote buying by campaign operatives. It means that all respondents
reported that they have been offered something in exchange for their vote (for the direct question), or
that they know another person who has been offered something in exchange for their vote (for the
indirect question). Further survey results are summarized in Table 2, indicating statistics on the type
of vote buying offer, the acceptance of the offer, and the reasons for voting for the candidate.
As mentioned earlier, another noteworthy pattern was that responses differed when vote buying
was disaggregated into four different types. The most direct form of vote buying—offering money
in exchange for votes—was reported by 28.5 percent of direct question respondents and by 34.9
percent for respondents of the indirect question. This is not a major difference across the two
methods, suggesting that there is not a great risk of a social desirability bias in our survey results.
This is supported by the results of Cruz (2013), who also found evidence that social desirability
bias is not a cause for concern in vote buying studies in the Philippines. The amounts offered for
vote buying ranged from PhP50 to PhP5000 (or roughly US$1.03 to US$103 using the exchange
rate at the time of writing this study), with an average of PhP493 (roughly US$10.2).
Offers of food, rice, or groceries were reported by 35.3 percent of respondents of the direct
question and by 41.1 percent for the indirect. The share of respondents who received an offer of
other material things3 was highest among the four vote buying types, at 54.8 percent for the direct
question and 54.5 percent for the indirect. Finally, 17.3 percent direct question respondents
reported that they had been offered any help or favor, lower than the 21.6 percent among those who
were asked the indirect question. Moreover, four percent of direct question respondents said that
they had received an offer in exchange for them not to vote in the 2016 elections. Among indirect
question respondents, 10.8 percent reported this.
A large majority of respondents (90.8 percent for direct, 97.4 percent for indirect) reported that
they had accepted the offer made to them. However, the reported acceptance rate widely differed
across vote buying types, at least for the direct question respondents. Among those who were offered
money, 84.2 percent said they had accepted the offer. The acceptance rate among those who were
offered food, rice, or groceries was higher at 95.0 percent; and 96.8 percent for those who were
offered other material things. “Other help or favor” was the vote buying type with the smallest
acceptance rate at 75.4 percent. The acceptance rate had less variation across vote buying types for
indirect question respondents. Offers of money, food, other material things, and help or favor had
acceptance rates of 96.0 percent, 98.1 percent, 98.6 percent, and 94.9 percent, respectively.
Compared to those who accepted the offer, a smaller share reported that they voted for the
candidate who made the offer (67.3 percent for the direct question; 81.5 percent for the indirect
question). Interestingly, among those who voted for the candidate, 94.4 percent of direct question
respondents said that they did so because the candidate was qualified, and only 16.7 percent said
that it was because of the offer.4 This pattern coheres with those uncovered by other studies
suggesting that vote buying does not really change people’s minds on their votes—rather it merely
caps a protracted patron-client relationship (and thus there is a pre-formed view of the politician by
the voter) (Schaffer, 2005). Furthermore, a mere 1.1 percent said this was because they feared
losing government benefits that they were receiving. This is an important finding given rumors
prior to the election that candidates were using the country’s social protection system as a bar-
gaining chip with voters—supposedly some candidates threatened to remove poor voters from the
beneficiaries list if they did not support them. This result suggests that this threat did not seem to be
as effective as some expected it to be.
For the indirect question respondents, the share of those who voted for the candidate because
they are qualified is 79.5 percent, because of the offer is 37.7 percent, and because they might lose
3. The other material things identified by respondents were mostly shirts, umbrellas, mugs, fans, and the like.
4. It must be noted that the possible answers are not mutually exclusive. The respondent can say that he/she
voted for the candidate both because of qualifications and the vote buying offer.
Canare et al. 71
government benefits is 4.1 percent. None among the direct question respondents said that they
voted for the candidate because he/she was intimidated (while for the indirect question respon-
dents, 0.6 percent said so).
Among those who did not vote for the candidate who made the offer, 62.3 percent of direct
question respondents said it was because of the candidate’s lack of qualifications, while 30.7
percent said it is because the candidate tried to buy the respondent’s vote. Among indirect question
respondents, these figures are 48.3 percent and 33.3 percent, respectively.
Related to this, among the respondents who accepted the offer, 69.7 percent of those asked the
direct question did vote for the candidate who made the offer. For the indirect question respon-
dents, this proportion is 82.3 percent. These figures suggest that a minority of respondents who
accepted the vote buying offer chose not to vote for the candidate. Possibly, this again reflects the
fact that most vote buying offers are made within the context of longstanding relationships between
candidates and voters who already tend to support them. In addition, there is anecdotal evidence
that campaign operatives often feign knowledge of voters’ choices, so that voters themselves feel
obliged to carry out the vote buying agreement.5
Correlates of vote buying

Turning to our analysis of vote buying patterns using probit methodology, we examine some of the
characteristics and factors linked to vote buying. This approach highlights whether and to what
extent targeting among poor voters could be taking place. Tables 3 to 6 show the average marginal
effects of the different independent variables on the likelihood of receiving a vote buying offer for
each of the four types of vote buying.
The first column shows marginal effects with only indicators of need and wealth as the independent
variables. In a similar survey after the 2013 Philippine elections, Mendoza et al. (2015) observed that
respondents with less income and fewer tangible assets reported more vote buying. Indeed, individuals
with lower income and more needs could be more susceptible to vote buying as politicians may try to
take advantage of their condition to buy votes. The second column shows marginal effects after
demographic factors are controlled for, and the third further includes perception on the performance of
the government and public services. The fourth column includes the margin of victory in the mayoral
race as a share of the total votes cast for mayor during the last election.
The results show that different vote buying types have different correlates. As expected, the
closeness of the previous election—measured by margin of victory of the winning mayoral candidate
as share of total votes cast for mayor—is positively correlated with vote buying, but only when
money is involved. This coheres with the findings from our discussions with campaign operators who
emphasize that monetary vote buying is most intense during tightly contested races, particularly for
the mayoral position which is key to mobilizing large numbers of voters at the local level.
Respondents are more likely to have been offered money when the ex-post margin of victory
was smaller (i.e. when elections were tighter). However, the story was entirely different with other
forms of vote buying. The variable on margin of victory was insignificant in vote buying using
food, rice, and groceries and in vote buying using “help or favor”. In sharp contrast to monetary
vote buying, respondents from areas where the elections are least contested are more likely to
receive offers of other material things such as shirts, mugs, umbrellas, and the like.
5. See, among others, Schaffer (2005), Mendoza et al. (2015), and our own discussion below on vote buying
techniques detailed from discussions with actual campaign agents.
Table 3. Average marginal effects after probit on receiving a vote buying offer (money).
Average Average Average Average

Variable Marginal Effect Marginal Effect Marginal Effect Marginal Effect
Employed –0.01955 –0.02774 –0.02196 –0.02523

(0.04532) (0.04639) (0.04642) (0.04550)
Income 0.00000 0.00000 0.00000 0.00001
(0.00000) (0.00000) (0.00000) (0.00000)**
Children –0.00121 –0.00601 –0.00417 –0.00387
(0.01162) (0.01260) (0.01282) (0.01255)
own_house –0.06556 –0.08142 –0.08582 –0.08149
(0.04816) (0.04927)* (0.04918)* (0.04852)*
Pppp 0.07822 0.09392 0.09942 0.07819
(0.04770) (0.04857)* (0.04841)** (0.04838)
Assets 0.00990 0.01092 0.01052 0.01098
(0.01276) (0.01264) (0.01275) (0.01253)
Hunger 0.08124 0.08973 0.09857 0.09807
(0.05071) (0.05104)* (0.05082)* (0.05024)*
Calamity 0.05358 0.05534 0.06185 0.03961
(0.04794) (0.04788) (0.04788) (0.04772)
Male 0.01655 0.01632 –0.00128
(0.06016) (0.06023) (0.05955)
Age 0.00268 0.00250 0.00245
(0.00185) (0.00186) (0.00184)
water_good 0.06703 0.06964
(0.04759) (0.04695)
localgovt_good –0.04168 –0.02534
(0.04707) (0.04635)
margin_victory –0.00196
(0.00059)***
N 399 399 399 399
Notes: *p < 0.1; **p < 0.05; ***p < 0.01; standard errors in parenthesis.
Several indicators of wealth and need were significantly correlated with some vote buying types.
Individuals with less wealth and more needs are more likely to receive offers of money and food in
exchange for votes. For vote buying offers using money, respondents who do not own their house and
those who experienced hunger are more likely to receive an offer. For vote buying using food, having
more children and fewer tangible assets increases the likelihood of receiving an offer for this type of
vote buying. In smaller communities wherein campaign operators are able to build strong ties,
knowledge of these voter characteristics may be built up over time. Indeed, similar patterns have
been observed in vote buying activities in other countries (e.g. Argentina; see Brusco et al., 2004).
Being a CCT6 beneficiary is also a positive correlate of receiving a monetary vote buying offer,
but its significance disappears as the closeness of the election race is controlled for. Once again,
with repeated interactions with communities at a local level, it is possible for campaign operators
6. See Fernandez and Olfindo (2011) for an overview of the Philippines’ CCT program.
Canare et al. 73
Table 4. Average marginal effects after probit on receiving a vote buying offer (food, rice, or groceries).

Employed 0.02251 0.03778 0.03610 0.03575

(0.04703) (0.04755) (0.04757) (0.04761)
Income –0.00000 –0.00000 –0.00000 –0.00000
(0.00000) (0.00000) (0.00000) (0.00000)
Children 0.03992 0.04037 0.04172 0.04172
(0.01085)*** (0.01165)*** (0.01170)*** (0.01170)***
own_house –0.00507 0.00378 0.00374 0.00393
(0.05008) (0.05161) (0.05113) (0.05118)
Pppp –0.05514 –0.06448 –0.06914 –0.07005
(0.05088) (0.05190) (0.05171) (0.05196)
Assets –0.02178 –0.02243 –0.02183 –0.02181
(0.01323)* (0.01325)* (0.01320)* (0.01320)*
Hunger 0.01697 0.01405 0.01135 0.01133
(0.05386) (0.05374) (0.05357) (0.05354)
Calamity 0.03623 0.03927 0.03351 0.03256
(0.05083) (0.05100) (0.05070) (0.05119)
Male –0.06762 –0.06680 –0.06772
(0.06405) (0.06433) (0.06454)
Age –0.00152 –0.00144 –0.00144
(0.00191) (0.00190) (0.00190)
water_good –0.05155 –0.05133
(0.04895) (0.04890)
(0.04883) (0.04927)
margin_victory –0.00010
(0.00063)
N 399 399 399 399
to be able to identify beneficiaries of the social protection program. Targeting these beneficiaries
should be straightforward. Nevertheless, it is possible that in more tightly contested races, the
targeting of CCT beneficiaries by campaign operatives is relaxed from being tightly focused on
CCT beneficiaries (who tend to be among the poorest in these low-income communities) to an
approach that covers more low-income voters including those who are not necessarily identified as
poor under the government’s social protection program.
A seemingly contrasting result is the positive and significant effect of income in monetary
vote buying—respondents with higher incomes are more likely to receive monetary vote buying
offers—but it is significant only in the equation where everything is controlled for. There are two
possible ways to interpret these findings. First, even among low-income voters, the threshold for
securing support may vary depending on socio-economic and other conditions. Unsurprisingly,
voters with less may be ingratiated with smaller offers, while those with more (proxied by higher
income) will likely expect more. In addition, campaign operatives may anticipate this and
provide different offers across target communities. Indeed, even where money is offered to buy
Table 5. Average marginal effects after probit on receiving a vote buying offer (other material things).

Employed –0.00984 –0.01360 –0.00716 –0.00409

(0.04909) (0.05043) (0.04981) (0.04963)
Income 0.00000 0.00000 0.00000 –0.00000
(0.00000) (0.00000) (0.00000) (0.00000)
Children –0.01030 –0.01359 –0.01565 –0.01589
(0.01212) (0.01313) (0.01313) (0.01291)
own_house 0.01415 0.00462 0.00310 0.00095
(0.05268) (0.05405) (0.05361) (0.05367)
Pppp –0.10647 –0.09696 –0.08493 –0.07481
(0.05177)** (0.05302)* (0.05278) (0.05302)
Assets 0.01093 0.01186 0.01097 0.01020
(0.01395) (0.01401) (0.01397) (0.01388)
Hunger –0.16992 –0.16414 –0.15639 –0.15845
(0.05422)*** (0.05493)*** (0.05444)*** (0.05394)***
Calamity –0.03531 –0.03308 –0.02078 –0.01045
(0.05260) (0.05282) (0.05246) (0.05295)
Male 0.00236 –0.00003 0.00905
(0.06680) (0.06603) (0.06603)
Age 0.00162 0.00151 0.00149
(0.00202) (0.00200) (0.00199)
water_good 0.13455 0.13299
(0.04955)*** (0.04941)***
localgovt_good 0.07980 0.07128
(0.04989) (0.04996)
margin_victory 0.00108
(0.00064)*
N 399 399 399 399
votes, there is a large variation across the 17 cities sampled. Among the 17 cities, the average
amount of money offered ranged from PhP50 to around PhP1800. The mean of this average is
around PhP540 and the median is P380. Three cities have an average of at least PhP1000, while
six have an average of PhP300 or less, indicating a wide variation in monetary vote buying
offers across cities.
In contrast to vote buying using money and food, there is less evidence that needs and wealth
have similar effects on vote buying using other material things and help or favor. If anything, some
results show the opposite. Respondents who experienced hunger are less likely to receive offers of
other material things. The same is true for CCT beneficiaries, although the significance of this
variable disappears as more variables are controlled for. For vote buying using help or favor,
respondents who had experienced a calamity are less likely to receive such offers.
Interestingly, some measures of quality of public services are also significant correlates of vote
buying attempts. Respondents who said that water service is good or very good are more likely to
receive an offer of other material things in exchange for their vote; while respondents who said that
Canare et al. 75
Table 6. Average marginal effects after probit on receiving a vote buying offer (any help or favor).

Employed 0.01957 0.01560 0.01534 0.01614

(0.03771) (0.03938) (0.03915) (0.03899)
Income 0.00000 0.00000 0.00000 0.00000
(0.00000) (0.00000) (0.00000) (0.00000)
Children 0.01468 0.01600 0.01293 0.01284
(0.00826)* (0.00884)* (0.00889) (0.00891)
own_house 0.04031 0.04079 0.04855 0.04826
(0.04162) (0.04147) (0.04119) (0.04123)
Pppp 0.03420 0.03334 0.03241 0.03371
(0.04005) (0.04061) (0.04042) (0.04069)
Assets –0.00632 –0.00634 –0.00614 –0.00615
(0.01087) (0.01089) (0.01087) (0.01089)
Hunger –0.04252 –0.04332 –0.05098 –0.05102
(0.04512) (0.04546) (0.04532) (0.04536)
Calamity –0.13198 –0.13308 –0.13662 –0.13510
(0.04461)*** (0.04530)*** (0.04562)*** (0.04587)***
Male 0.02365 0.02072 0.02185
(0.04994) (0.04978) (0.05000)
Age –0.00014 –0.00005 –0.00006
(0.00151) (0.00149) (0.00149)
water_good –0.02859 –0.02804
(0.04047) (0.04016)
localgovt_good 0.07981 0.07863
(0.03781)** (0.03751)**
margin_victory 0.00016
(0.00046)
N 399 399 399 399
the city government performance is good or very good are more likely to be offered help or favor in
exchange for their vote.
Correlates of accepting a vote buying offer

The empirical results show that indicators of need and wealth, perceptions about one’s sur-
roundings, and life expectations are among the key correlates of accepting a vote buying offer. The
average probit marginal effects of each independent variable on the likelihood of accepting the
offer are reported in Table 7. Two columns of marginal effects are shown; the second column
shows the marginal effects after the vote buying types have been controlled for.
In general, indicators of wealth have a negative effect and indicators of need have a positive
effect on the likelihood of accepting the offer. That is, among the poor, those with relatively less
wealth and relatively more needs are more likely to accept a vote buying offer. Respondents with
lower income and those who are unemployed are more likely to accept the offer. Respondents who
Table 7. Average marginal effects after probit on accepting the vote buying offer.
Variable Average Marginal Effect Average Marginal Effect
Male 0.01417 0.02087

(0.03845) (0.03398)
Age –0.00130 –0.00113
(0.00110) (0.00106)
Elemgrad –0.06765 –0.03961
(0.07146) (0.05451)
Highschool –0.04305 –0.00915
(0.07290) (0.05667)
Posths –0.06392 –0.04628
(0.07637) (0.05930)
Employed –0.04851 –0.04977
(0.02904)* (0.02519)**
Income –0.00000 –0.00000
(0.00000)* (0.00000)*
life_improve –0.00203 0.00630
(0.02831) (0.02486)
Cheating –0.00849 0.00574
(0.03729) (0.03127)
News 0.00164 0.00127
(0.00179) (0.00176)
Internet 0.00096 0.00068
(0.00157) (0.00134)
Credit 0.01653 –0.01622
(0.03032) (0.02559)
Calamity 0.04685 0.02396
(0.03257) (0.02841)
Believe 0.04334 0.05561
(0.04364) (0.04212)
assets_acquired –0.01585 –0.00921
(0.00842)* (0.00781)
natlgovt_good 0.01406 0.03084
(0.02790) (0.02481)
(0.02533)* (0.02275)
water_good –0.01704 –0.03745
(0.02913) (0.02519)
Crime –0.05508 –0.04856
(0.02775)** (0.02436)**
Happy 0.00707 –0.02008
(0.02717) (0.02545)
expect_improve 0.03173 0.04869
(0.02780) (0.02325)**
margin_victory –0.00073 –0.00107
(0.00034)** (0.00033)***
(continued)
Canare et al. 77
Table 7 (continued)
Variable Average Marginal Effect Average Marginal Effect
q1ci_ofer_mony –0.06586
(0.02917)**
q1ciii_ofer_food 0.07002
(0.02728)**
q1civ_ofer_mtrl 0.11391
(0.02836)***
q1cvi_ofer_help –0.09592
(0.02663)***
N 399 399
acquired fewer tangible assets over the last three years are also more likely to accept the offer, but
the significance of this particular variable disappears as vote buying types are controlled for.
Perceptions about one’s surroundings and life expectations are also correlates of accepting a
vote buying offer. Those who said that crime in their city is prevalent or really prevalent are less
likely to accept the offer, while those who expect their lives to improve are more likely to accept it.
Interestingly, respondents in cities where the mayoral elections were more closely contested are
more likely to accept the offer.
The four vote buying type dummies all have statistically significant marginal effects and
with different signs, indicating that what the offer was affected whether the respondent accepted
it or not. Offers of money and help or favor are relatively less likely to be accepted compared to
the other two. Other material things are the offers most likely to be accepted, followed by food,
rice, or groceries.
Correlates of voting
The average probit marginal effects of each independent variable on the likelihood of voting for the
candidate who made the offer are reported in Table 8. Three columns of marginal effects are shown
in the table. The second column shows the marginal effects after the vote buying types have been
controlled for, while the third reports marginal effects after controlling for whether the offer was
accepted or not.
In contrast to accepting the offer, the correlates of voting for the candidate who made the offer
are mostly perceptions about the voter’s surroundings and life experiences. It is also affected by
age, which has a positive and significant marginal effect on the likelihood of voting for the can-
didate who made the offer. That is, older respondents are more likely to vote for the candidate who
offered to buy their votes.
Similar to accepting the offer, those who said that crime in their city is prevalent or very prevalent
are less likely to vote for the candidate who made the offer. On the other hand, respondents who rated
the city government performance as good or very good are more likely to vote for the candidate; and
those who are happy or very happy are less likely to vote for the candidate.
Most vote buying type dummies are insignificant, except for other material things. Those who
were offered other material things are more likely to vote for the candidate who made the offer
Table 8. Average marginal effects after probit on voting for the candidate who made the offer.
Variable Average Marginal Effect Average Marginal Effect Average Marginal Effect
Male 0.03146 0.03365 0.02715

(0.06393) (0.06315) (0.06166)
Age 0.00629 0.00621 0.00657
(0.00194)*** (0.00195)*** (0.00193)***
Elemgrad 0.03183 0.01686 0.02651
(0.09918) (0.09957) (0.09814)
Highschool –0.00540 –0.01317 –0.01079
(0.10219) (0.10253) (0.10043)
Posths 0.03108 0.01334 0.02562
(0.10970) (0.11071) (0.10860)
Employed –0.06595 –0.06658 –0.05367
(0.04845) (0.04783) (0.04770)
Income –0.00000 –0.00000 –0.00000
(0.00000) (0.00000) (0.00000)
life_improve –0.01180 –0.01084 –0.01592
(0.04962) (0.04948) (0.04882)
Cheating 0.01436 0.01244 0.01239
(0.06195) (0.06207) (0.06080)
News –0.00197 –0.00257 –0.00275
(0.00261) (0.00261) (0.00252)
Internet 0.00202 0.00222 0.00199
(0.00249) (0.00247) (0.00245)
Credit –0.03261 –0.04462 –0.04919
(0.05215) (0.05193) (0.05141)
Calamity 0.06126 0.06530 0.05948
(0.05028) (0.05101) (0.04992)
Believe 0.00212 0.01388 –0.00170
(0.06453) (0.06471) (0.06348)
assets_acquired 0.00354 0.00659 0.01008
(0.01555) (0.01540) (0.01532)
natlgovt_good –0.03137 –0.02840 –0.03434
(0.04819) (0.04805) (0.04755)
localgovt_good 0.09793 0.09199 0.10191
(0.04953)** (0.04959)* (0.04910)**
water_good 0.02472 0.01000 0.02006
(0.04949) (0.04964) (0.04889)
Crime –0.09057 –0.08428 –0.07458
(0.04639)* (0.04625)* (0.04571)
Happy –0.07824 –0.08398 –0.08033
(0.04837) (0.04798)* (0.04767)*
expect_improve 0.00433 0.01121 0.00132
(0.04976) (0.04951) (0.04882)
margin_victory 0.00075 0.00055 0.00083
(0.00062) (0.00062) (0.00062)
(continued)
Canare et al. 79
Table 8 (continued)
Variable Average Marginal Effect Average Marginal Effect Average Marginal Effect
q1ci_ofer_mony –0.00616 0.01275

(0.05290) (0.05295)
q1ciii_ofer_food 0.04818 0.02238
(0.05109) (0.05133)
q1civ_ofer_mtrl 0.11591 0.07599
(0.05207)** (0.05382)
q1cvi_ofer_help 0.02018 0.06341
(0.06717) (0.06904)
q1d_accept 0.26091
(0.07776)***
N 399 399 399
compared to other vote buying types. However, its significance disappears once the dummy for
accepting the offer has been controlled for. As expected, respondents who accepted the offer are
more likely to vote for the candidate who made the offer, and it appears to be the most important
determinant of voting for the candidate.
Summary and conclusions

This study exposed several important patterns on vote buying, using data on low-income voters in
Metro Manila collected after the 2016 national and local elections. Although all respondents
reported that they received a vote buying offer, the incidences of the different vote buying types
were widely distributed. The most common form of vote buying was offering other material things,
identified mostly by respondents as shirts, umbrellas, mugs, and the like. The next most prevalent
type of vote buying was offering food, rice, or groceries. The type of vote buying most of the
literature identifies with—offering cash in exchange for votes—was only the third most prevalent
type. The least prevalent type was offering any help or favor to the potential voter.
A large majority of respondents said that they accepted the offer, but the acceptance rate also differs
across vote buying types. Any help or favor has the lowest acceptance rate, followed by money. Other
material things and food have the highest acceptance rates. The lower acceptance rate of money and the
higher acceptance rate of the relatively benign offers (other material things and food) could indicate
that there are some negative views attached to monetary vote buying, and that accepting material things
other than money in exchange for votes could be deemed more socially acceptable.
Compared to those who accepted the offer, there is a smaller share of respondents who voted for
the candidate who made the offer. Moreover, among those who accepted the offer, only about two
out of three voted for the candidate. This indicates weak enforcement of the implicit contract
between the vote buyer and the vote seller. Most (94 percent) of those who said that they voted for
the candidate did so because of the candidate’s qualifications and only 17 percent because of the
offer, indicating that they probably would have voted for the candidate anyway. Among those who
did not vote for the candidate who made the offer, 62 percent said this was because of the can-
didate’s lack of qualifications, and 31 percent because of the vote buying. Once again, this coheres
with earlier research that vote buying is part of a longstanding relationship between politicians and
their low-income constituents. This is also supported by earlier research on vote buying. Schaffer
(2002a) analyzed vote buying in several countries and argued that vote buying may or may not be
effective. A related vote buying survey among low-income voters in the Philippines by Schaffer
(2002b) also found evidence that material offers in exchange for votes affected the vote of only
about 30 percent of those who accepted them. Ibana (1996) similarly argues that vote buying is
seen by ordinary voters as a way to achieve equality and they view it as an obligation of the
candidates to support them, rather than as an economic transaction.
This could also be explained by Nichter’s (2008) concept of “turnout buying,” wherein the
candidate makes transfers to already known supporters to induce them to vote. Dependency and
loyalty is merely punctuated by election-related transfers, rather than an effort to completely
change votes. These figures could suggest that vote buying is already an accepted practice and
while it can influence a voter’s decision, there could be other more important factors in selecting
candidate to support. As discussed in earlier sections, vote buying is a long process that merely
culminates with the actual offers of material things as an election approaches. It starts way before
an election, with candidates slowly building dependency and loyalty among their constituents.
Nevertheless, it should be noted that in close elections, vote buying—and using money to buy
votes in particular—could become more intensively used by campaign operators. This was cor-
roborated by our interviews with political operatives. The results also provide evidence that
supports this, suggesting that that the poor are more likely to be targeted for monetary vote buying
if the election is closely contested.7 Moreover, the results also suggest that these vote buying offers
are more likely to be accepted in close elections.
As noted earlier, the co-existence of these different vote buying approaches suggests that
campaign operatives use a variety of tools to try to secure votes. Further analysis on their correlates
suggests that these tools are deployed strategically across different contexts and targeted towards
different voters depending on their characteristics. The significance of key variables such as
ownership of tangible assets and number of children also suggests that visible proxies could be
used by campaign operatives to target certain communities of voters. In particular, the significance
of these two variables provides evidence that among low-income voters, certain forms of vote
buying are targeted at those with relatively even greater needs. Similar to the government’s poverty
targeting for social programs, campaign operators appear to strive for efficiency by targeting low-
income voters in ways that leverage their vulnerability and dependence, and particularly in the
context of tight races.
Results show evidence that the closeness of the election race is one of the most important
correlates of money-based vote buying. The closer the election, the more likely that monetary vote
buying will occur. Closeness of the race, however, has different effects on vote buying depending
on vote buying type. The likelihood of vote buying using food, rice, or groceries and help or favor
is not affected by closeness of the race. In contrast to monetary vote buying, there is more vote
buying using other material things when the race is not tight. This could imply that candidates
consider money as the most potent type of vote buying in terms of attracting votes. This could also
7. It must be noted, though, that although the closeness of the mayoral race was used to represent the clo-
seness of elections, it cannot be concluded that it is the mayoral candidates who do the vote buying. The
mayoral race was chosen as proxy for closeness of election because it allows for some variability across
observations. In contrast, races for national positions do not allow for variability across observations; that
is, all respondents in the survey face the same candidate choices for national posts.
Canare et al. 81
show that even if a candidate is confident of winning, he/she can still attempt to buy votes, but by
using more “benign” forms. The likely purpose of this is to maintain and to continue building the
patron and dependency relationship between the candidate and voters, which could prove useful in
the future, when elections might be closer.
Aside from closeness of the race, there are also some asset and need indicators that turned out to
be significant correlates of vote buying, although these effects are again different across vote
buying types. Candidates would be expected to target voters with greater needs and fewer assets, as
they could be more likely to sell their votes. However, this effect is only observed in monetary and
food vote buying. A possible explanation for this is that money and food could be the greater
immediate necessity of those with relatively more needs and less wealth. Variations in the direction
of effects, significance, and number of significant correlates across the four vote buying types
indicate that there could be different schemes in targeting potential voters.
Correlates of accepting a vote buying offer are more or less in line with what is expected. Lower
income, unemployment, and fewer acquired tangible assets are all associated with a higher
probability of accepting the vote buying offer. In relation to this, some needs indicators are sig-
nificant correlates of accepting a vote buying offer, but they are not correlated with voting for the
candidate who made the offer. This means that needs could have induced the voters to accept the
offer, but needs do not affect whether or not the voter will vote for the candidate who made the
offer. Interestingly, accepting the offer appears to be the most important correlate of voting,
indicating that somehow, although imperfectly, vote sellers keep their end of the bargain. This
coheres with the literature cited earlier that vote buyers target those more likely to embody norms
of reciprocity (see Finan and Schechter, 2012). Another interesting result is that respondents are
more likely to accept the offer if the race is more tightly contested.
The type of offer also affects whether it will be accepted or not. Other material things are the most
likely to be accepted, followed by food, rice, or groceries. Offers of money and help or favor are less
likely to be accepted. This indicates that potential vote sellers could be selective in accepting offers,
and are more likely to accept the more benign ones. This selectivity against accepting money could
also be due to the attached social undesirability of accepting money in exchange for votes, such that
only in certain contexts is this strategy truly effective (and therefore also used by campaign
operators). That is, although there is evidence that vote buying in general has come to be socially
accepted, there could still be some undesirability attached to accepting money.
All this evidence should be useful for policymakers in introducing reforms to prevent vote
buying and enhance citizens’ awareness of their voting responsibilities. Clearly, information
dissemination on the adverse effects of vote buying, voters’ rights, and candidates’ responsi-
bilities needs to be deeply engrained years before elections, and it is also critical to fine-tune
voter education and outreach campaigns for different voter groups—notably poor and low-
income households—so that these campaigns are intensified where needed (e.g. in jurisdic-
tions with close races).
Acknowledgements
We thank the National Academy of Science and Technology (NAST) and the Young Public Servants (YPS)
for supporting this study, the key resource persons interviewed for their insights on how vote buying in the
Philippines works, and the anonymous referee for suggesting revisions that greatly improved this paper.
However, the views expressed herein are ours and do not necessarily reflect the views of these institutions
or persons. This paper is dedicated to the memory of our colleague and co-author, Mario Antonio Lopez.
Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or
publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publi-
cation of this article: This study was funded by the National Academy of Science and Technology (NAST)
and the Young Public Servants (YPS).
References
Anderson J (1978) The validity of Haire’s shopping list projective technique. Journal of Marketing Research
15(4): 644–649.
Barrientos A and Hulme D (2008) Social Protection for the Poor and Poorest: Concepts, Policies and
Politics. New York: Palgrave Macmillan.
Blaydes L (2006) Who votes in authoritarian elections and why? Determinants of voter turnout in contem-
porary Egypt. In: 2006 annual meeting of the American Political Science Association, Philadelphia, 31
August–3 September 2006.
Bratton M (2008) Vote buying and violence in Nigerian election campaigns. Electoral Studies 27(4):
621–632.
Brusco V, Nazareno M and Stokes SC (2004) Vote buying in Argentina. Latin American Research Review
39(2): 66–88.
Calder BJ and Burnkrant RE (1977) Inter-personal influence on consumer behavior: An attribution theory
approach. Journal of Consumer Research 4(1): 29–38.
Carkoglu A and Aytac SE (2015) Who gets targeted for vote buying? Evidence from a list experiment in
Turkey. European Political Science Review 7(4): 547–566.
Carreras M and Irepoglu Y (2013) Trust in elections, vote buying and turnout in Latin America. Electoral
Studies 32(4): 609–619.
Chaudhuri A and Christofides TC (2013) Indirect Questioning in Sample Surveys. Berlin: Springer-Verlag.
Chaves I, Fergusson L and Robinson JA (2015) He who counts elects: Economic elites, political elites and
electoral fraud. Economics and Politics 27(1): 124–159.
Cruz C (2013) Social networks and the targeting of illegal electoral strategies. In: 2013 American Political
Science Association annual meeting, Chicago, 29 August–1 September 2013.
Cruz C, Keefer P and Labonne J (2016) Incumbent advantage, voter information, and vote buying. Inter-
American Development Bank Working Paper Series No. IDB-WP-711.
Devereux S (2002) Social protection for the poor: Lessons from recent international experience. Institute of
Development Studies Working Paper 142.
Fernandez L and Olfindo R (2011) Overview of the Philippines’ conditional cash transfer program: The
Pantawid Pamilyang Pilipino Program (Pantawid Pamilya). The World Bank Philippine Social Protection
Note No. 2.
Finan F and Schechter L (2012) Vote buying and reciprocity. Econometrica 80(2): 863–881.
Fionna U (2014) Vote buying in Indonesia’s 2014 elections: The other side of the coin. ISEAS Perspective
2014: 35.
Fisher RJ (1993) Social desirability bias and the validity of indirect questioning. Journal of Consumer
Research 20(2): 303–315.
Gonzalez-Ocantos E, Kiewiet de Jonge C, Melendez C, et al. (2012) Vote buying and social desirability bias:
Experimental evidence from Nicaragua. American Journal of Political Science 56(1): 202–217.
Hicken A (2002) The market for votes in Thailand. In: Trading political rights: The comparative politics of
vote buying international conference, Center for International Studies, Massachusetts Institute of Tech-
nology, Cambridge, Massachusetts, 26–27 August 2002.
Canare et al. 83
Hicken A (2006) How do rules and institutions encourage vote buying? In: Schaffer FC (ed.) Elections for
Sale: The Causes and Consequences of Vote Buying. Boulder, CO: Lynne Rienner Publishers, pp. 47–60.
Hicken A and Simmons J (2008) The personal vote and the efficacy of education spending. American Journal
of Political Science 52(1): 109–124.
Hicken A, Leider S, Ravanilla N, et al. (2014) Temptation in vote selling: Evidence from a field experiment in
the Philippines. Center for Economic Studies and IFO Institute Working Paper No. 4828.
Ibana R (1996) Epilogue: Ordinary people in everyday life. In: Alejo MJ, Rivera MEP and Valencia NIP (eds)
[De]scribing Elections: A Study of Elections in the Lifeworld of San Isidro. Quezon City: Institute for
Popular Democracy, pp. 125–138.
Jensen PS and Justesen (2014) Poverty and vote buying: Survey-based evidence from Africa. Electoral
Studies 33(1): 220–232.
Keefer P (2007) Clientelism, credibility, and the policy choices of young democracies. American Journal of
Political Science 51(4): 804–821.
Keefer P and Khemani S (2004) Why do the poor receive poor services? Economic and Political Weekly
39(9): 935–943.
Khemani S (2015) Buying votes versus. supplying public services: Political incentives to under-invest in pro-
poor policies. Journal of Development Economics 117: 84–93.
Kitschelt H and Wilkinson S (2007) Citizen-politician linkages: An introduction. In: Kitschelt H and Wilk-
inson S (eds) Patrons, Clients and Policies: Patterns of Democratic Accountability and Political Com-
petition. Cambridge: Cambridge University Press, pp. 1–49.
Kitschelt H, Hawkins K, Luna JP, et al. (2010) Latin American Party Systems. Cambridge: Cambridge
University Press.
Lehoucq F (2002) When do parties buy votes? Theoretical and empirical perspectives on electoral corruption.
In: Trading political rights: The comparative politics of vote buying international conference, Massachu-
setts Institute of Technology, Cambridge, Massachusetts, 26–27 August 2002.
Lyne M (2007) Rethinking economics and institutions: The voter’s dilemma and democratic accountability.
In: Kitschelt H and Wilkinson S (eds) Patrons, Clients and Policies: Patterns of Democratic Account-
ability and Political Competition. Cambridge: Cambridge University Press, pp. 159–181.
Mendoza RU, Lopez MA, Yap D II, et al. (2015) The 2013 Philippine mid-term election: An empirical analysis
of dynasties, vote buying, and the correlates of senate votes. In: Mendoza RU, Beja EL Jr, and Teehankee JC
et al. (eds) Building Inclusive Democracies in ASEAN. Manila: Anvil Publishing, pp. 119–128.
Miasco M (2016) Comelec admits vote buying massive in automated polls. The Philippine Star, 9 May.
Available at: http://www.philstar.com:8080/cebu-news/2016/05/09/1581423/comelec-admits-vote-buy
ing-massive-automated-polls (accessed 21 August 2017).
Nichter S (2008) Vote buying or turnout buying? American Political Science Review 102(1): 19–31.
Phongpaichit P, Treerat N, Chaiyapong Y, et al. (2000) Corruption in the public sector in Thailand: Percep-
tions and experience of household. Working paper, Chulalongkorn University, Thailand.
Robertson D and Joselyn R (1974) Projective techniques in research. Journal of Advertising Research 14(5):
27–31.
Rodriguez F (2016) Comelec urges public to report vote-buying. Rappler, 3 April. Available at: http://www.
rappler.com/newsbreak/in-depth/122509-vote-buying-education-comelec (accessed 21 August 2017).
Schaffer FC (2002a) What is vote buying? In: Trading political rights: The comparative politics of vote
buying international conference, Center for International Studies, Massachusetts Institute of Technology,
Cambridge, Massachusetts, 26–27 August 2002.
Schaffer FC (2002b) Disciplinary reactions: Alienation and the reform of vote buying in the Philippines. In:
Trading Political Rights: The Comparative Politics of Vote Buying International Conference, Center for
International Studies, Massachusetts Institute of Technology, Cambridge, Massachusetts, 26–27 August
2002.
Schaffer FC (2005) Clean elections and the great unwashed: Vote buying and voter education in the Philip-
pines. Massachusetts Institute of Technology School of Social Science Occasional Paper No. 21.
Schaffer FC and Schedler A (2006) What is vote buying? In: Schaffer FC (ed.) Elections for Sale: The Causes
and Consequences of Vote Buying. Boulder, CO: Lynne Rienner Publishers, pp. 17–32.
Stokes S (2005) Perverse accountability: A formal model of machine politics with evidence from Argentina.
American Political Science Review 99(3): 315–325.
Vicente P (2014) Is vote buying effective? Evidence from a field experiment in West Africa. The Economic
Journal 124(574): 356–387.
Vicente P and Wantchekon L (2009) Clientelism and vote buying: Lessons from field experiments in African
elections. Oxford Review of Economic Policy 25(2): 292–305.
View publication stats

Vote Buying SEARCanare Mendoza Lopez 2018

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vote Buying SEARCanare Mendoza Lopez 2018

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

An Empirical Analysis of Vote Buying Among the Poor: Evidence from

Article · February 2018

Tristan Canare Ronald U. Mendoza

SEE PROFILE SEE PROFILE

Mario Antonio Guatelara Lopez

The user has requested enhancement of the downloaded file.

South East Asia Research

vote buying among the sagepub.co.uk/journalsPermissions.nav

Mario Antonio Lopez

Figure 1. Estimated prevalence of vote buying across selected democracies.

Party preference of recipient

Figure 2. Framework of vote vs turnout buying.

Vote buying in the Philippines

The 2016 Philippines election and vote buying

Vote buying beyond the Philippines

Correlates of vote buying

Table 1. Summary statistics of independent variables.

Variable Obs Mean Std. Dev. Min Max

male 400 0.1875 0.3908 0 1

Source: Authors’ calculations using survey data.

Table 2. Selected survey results.

Variable Direct Question Indirect Question

Vote Buying Offer (General) 100.00% 100.00%

Correlates of vote buying

Average Average Average Average

Employed –0.01955 –0.02774 –0.02196 –0.02523

Average Average Average Average

Employed 0.02251 0.03778 0.03610 0.03575

Average Average Average Average

Employed –0.00984 –0.01360 –0.00716 –0.00409

Average Average Average Average

Employed 0.01957 0.01560 0.01534 0.01614

Correlates of accepting a vote buying offer

Variable Average Marginal Effect Average Marginal Effect

Male 0.01417 0.02087

Variable Average Marginal Effect Average Marginal Effect

Male 0.03146 0.03365 0.02715

q1ci_ofer_mony –0.00616 0.01275

Summary and conclusions

Declaration of Conflicting Interests

View publication stats

You might also like