Access to Information and Other Correlates of Vote Buying and Selling Behaviour: Insights from Philippine Data

Abstract

Access to information is a key factor influencing political behaviour and decisions. Recent studies on vote buying and selling have tried to unpack the possible drivers behind this phenomenon; yet, few studies have empirically examined the role of different sources of information. This study contributes to the nascent literature in this area by turning to a unique dataset from a survey of low-income voters in Metro Manila, the Philippines. It empirically examined the relationship between access to information and vote selling behaviour by low-income voters. It also studied other correlates of vote selling and the possible factors linked to receiving an offer. The results suggest that the quantity of information has no significant relationship with the likelihood of accepting the offer and voting for the candidate for whom the offer was made. However, the quality of information does matter. In particular, access to sources of ‘good quality information’ is negatively associated with completing the vote selling transaction (i.e., accepting the offer and voting for the candidate). This study also found evidence that when money is used for vote buying, it appears to be targeted at those with greater needs, confirming the literature that vote buying activities tend to be well targeted at poor and low-income communities. Unsurprisingly, vote buying offers are more likely in areas where elections are closely contested, and they are also more likely in socially cohesive communities. Our findings also suggest that vote buying may not necessarily be effective in the sense that it encourages only few voters to change their candidate preference. This coheres with earlier studies suggesting that vote buying and selling merely caps a longstanding patron–client relationship between politicians and low-income voters.

JEL: D72, D91, K49

Keywords

Asian politics political behaviour political economy Southeast Asian studies

Background and Objectives

Various forms of political patronage and clientelism, including vote buying and selling, can adversely affect growth and development because they induce inefficiencies in the redistribution of resources (Vicente & Wantchekon, 2009). Vote buying and selling has also been shown to be associated with corruption and inefficient government (Keefer, 2007; Kitschelt & Wilkinson, 2007) and with weaker accountability and slower institutional growth (Lyne, 2007; Stokes, 2005). Although these practices can produce suboptimal social outcomes, these persist because they can be an effective strategy for candidates, particularly when voters lack (or possess imperfect) political information (Kramon, 2016; Wantchekon, 2003; Vicente, 2014), when voters’ incomes are low, when monitoring can be enforced (Brusco et al., 2004) and when norms of reciprocity exist in society (Finan & Schechter, 2012).

Vote buying is commonly defined as the offering of cash or other objects of value in exchange for votes (Brusco et al., 2004; Kramon, 2016). Vote selling is the acceptance of this offer and agreeing to vote for the candidate linked to the offer. In the strict sense, vote selling only takes place when the voter shifts from either undecided or changes their mind away from another candidate, and in favour of the candidate linked to the offer.

However, Schaffer and Schedler (2006) argue that vote buying and selling are more than just simple market transactions of exchanging a service for a fee. For one, because it is illegal, the vote buying and selling ‘contract’ between the candidate and the voter cannot be enforced under the usual mechanisms of law. Completing the vote buying and selling transaction, which includes the voter accepting the offer and voting for the candidate for whom the offer was made, is, therefore, subject to normative and cultural subtexts (Schaffer & Schedler, 2006). Moreover, with a secret ballot, the success of a vote buying and selling transaction rests on the ability of the candidate to properly target vote sellers (Cruz, 2019; Nichter, 2008).

Because of the negative effects of these practices, several studies attempted to determine what interventions may mitigate them, including campaigns to convince voters to promise not to sell votes (Hicken et al., 2017), voter education campaigns, including those focused on improving information on the costs of vote buying and selling (Vicente, 2014; Vicente & Wantchekon, 2009), and disseminating information on the performance of incumbent officials (Banerjee et al., 2011).

Often, the central theme of most of these interventions is providing the voters with more information. Access to good quality information has been shown to not only affect voting behaviour and political decisions (Ferraz & Finan, 2008; Ghirardato & Katz, 2002; Houser et al., 2011) but also produce better decisions (Lupia, 1994a; Bassi et al., 2011). However, the effect of information on voting behaviour depends largely on the quality of this information (Ghirardato & Katz, 2002), particularly in the contemporary context of widespread disinformation (Alcott & Gentzkow, 2017).

Building on earlier literature, the main objective of this study is to evaluate the relationship between access to information and vote selling behaviour. Specifically, we looked at the relationship between information access and the propensity to accept the vote buying offer and to vote for the candidate for whom the offer was made. The primary contribution of this article to the literature is its focus on the source, or quality, of information rather than the quantity or presence of information as mentioned in the studies of Kramon (2016), Banerjee et al. (2011), Fujiwara and Wantchekon (2013), Gottlieb (2016) and Keefer and Khemani (2014).

The secondary objective of this study is to empirically examine the correlates of receiving a vote buying offer. Although a rich body of literature suggests that the poor are disproportionately targeted by vote buyers (Carkoglu & Aytac, 2015; Jensen & Justesen, 2014; Schaffer, 2005), the other characteristics of voters being targeted for vote buying are less studied.

This study used data drawn from a survey of low-income households in Metro Manila, the Philippines conducted in July 2019, 2 months after the May national and local elections. The focus is on these families because anecdotal and empirical evidence suggests that it is the poor who receive the bulk of vote buying offers (Jensen & Justesen, 2014; Schaffer, 2005). Moreover, in studies of vote buying in the Philippines, evidence suggests that social desirability bias against this practice tends to be low in the country (Cruz, 2013; Cruz et al., 2016; 2017; Khemani, 2013).

In what follows, the second section contains a brief review of related literature on information and political decisions. The third section then elaborates on the survey methodology and the estimation method. The fourth section then discusses the main results and its implications. A brief concluding section summarises the main findings and outlines future directions for follow-up research.

Literature Review and Framework

Information and Political Decisions

There is a rich strand of literature on the possible effects of information on political decisions. It ranges from voter turnout (Bond et al., 2012; Houser et al., 2011; Larcinese, 2007; Lassen, 2005) to voting for underperforming officials (Banerjee et al., 2011; Ferraz & Finan, 2008; George et al., 2018) and political clientelism (Kramon, 2016; Vicente, 2014; Vicente & Watchekon, 2009). In a laboratory experiment of undergraduate students in California, Lupia (1994b) also found evidence that access to even a small amount of information can induce poorly informed voters to vote as if they are fully informed. The author came to this conclusion because the response to little information was the same as the response to complete information.

In most of these studies, information was provided as an intervention through experimental methods either on the field or in the laboratory. An example of the former is the study by Vicente (2014), which randomly distributed information campaign leaflets against vote buying prior to an election in São Tomé and Príncipe. This study found evidence that the intervention decreased voter turnout and reduced the influence of vote buying in the voting behaviour of programme recipients.

In a related paper, Ferraz and Finan (2008) released audit findings in random municipalities prior to a Brazilian election. Results suggest that corrupt politicians were less likely to be re-elected in areas where audit results were released, and the effect was stronger in areas with access to radio broadcasts. Furthermore, Banerjee et al. (2011) and George et al. (2018) conducted similar experiments in India. These studies found similar effects on votes for poorly performing candidates (i.e., candidates who performed poorly based on some indicators received less votes). Other studies used laboratory experiments, like Houser et al. (2011), which concluded that campaign advertisements increase voter participation. Furthermore, Bassi et al. (2011) found evidence that enhanced access to information weakens the tendency of voters to vote for candidates whom they can identify with. Put simply, access to information enables voters to widen their basis for selection, enabling them to choose from beyond initial biases.

On the other hand, non-experimental studies typically turned to survey data in order to examine the links between information and political decisions. Larcinese (2007) used British survey data in the 1997 general election to show that political knowledge can influence the likelihood of voting. Similarly, Hobolt (2007) studied voting patterns in referendums in three European countries, that is, Denmark, Ireland and Norway, and concluded that greater exposure to information increases the likelihood of voting based on issues rather than on personalities. A related study by De Vries et al. (2011) using European parliamentary elections in 2009 had the same conclusion.

Although studies have shown that the presence of and quantity of information do affect political decisions, including voting behaviour; there is far less research on the potential effect of the quality of information. Ghirardato and Katz (2002) constructed a theoretical model explaining how quality of information can affect voting behaviour, but empirical analysis on this topic is still scant.

Access to Information and Vote Buying Decision

In this section, we discuss why and how information can affect vote buying decisions—and why differing qualities of information can have varying effects. The foremost argument is that information on the illegality and corresponding penalties for vote selling can incentivise voters to refrain from selling their votes. Vote buying and selling are illegal in the Philippines, as these practices are in many countries. In the Philippines, they are punishable by imprisonment of 1–6 years.¹

Related to this, information on the negative effects of vote buying and selling can have the same effect. Studies have shown that vote buying and selling can adversely affect the economy, governance and public service provision (Keefer, 2007; Kitschelt & Wilkinson, 2007; Lyne, 2007). Knowledge of these effects can discourage voters from selling their votes. A crucial assumption for this mechanism to work is the correctness of information. This assumption is driven by what Ghirardato and Katz (2002) referred to as ‘ambiguities’, or the likelihood that the information is wrong. Good quality information will have minor ambiguities.

There are other less direct reasons why access to information can affect vote buying behaviour—and the direction of relationship can be ambiguous. Some voters justify selling their votes by the poor quality of public services and inferior performance of elected officials (Schaffer & Schedler, 2006; Vicente, 2014). If this is the case, information on the performance of incumbents—such as those in the experiments of Ferraz and Finan (2008), Banerjee et al. (2011) and George et al. (2018)—can either exacerbate or curb vote buying and selling, depending on the nature of the information released to voters. Following this argument, information on politician performance may encourage vote buying if the incumbent is a poor performer. On the other hand, it can limit vote buying and selling if the official performs well.

Related to this, information on illegality and penalties of vote buying and selling, which we argued should help disincentivise these practices, can induce an unintended positive effect on vote buying and selling if the general environment exhibits weak enforcement. In the Philippines, the prevalence of these practices suggests that it has become socially acceptable (Cruz et al., 2016; 2017; Khemani, 2013), which suggests weak enforcement.

These arguments suggest that access to information, even those of high quality, do not necessarily lead to lower propensity to sell votes. Good information only helps the voters weigh the expected benefits and costs of vote selling; yet, the decision on whether to sell their votes or not will still depend on factors such as the preferences of the voter as well as the expected rationality of the voting decision itself.

Data and Methodology

Source of Data

The data were developed from a survey of poor and low-income voters in Metro Manila, the Philippines. Metro Manila, or the National Capital Region (NCR), is a group of 17 contiguous cities including the capital. It accounts for about 13% of the national population (roughly 12 million residents) and 37% of the country’s gross domestic product (GDP).

Respondents were selected using multi-stage stratified random sampling covering all 17 cities of NCR, with the number of respondents per city proportional to the number of voters. The first stage comprised of randomly selecting low-income communities from a list provided by the city governments. For each city, congressional districts² served as strata, which means that each congressional district from each city was represented. The second stage comprised of randomly selecting respondents from each drawn community through systematic sampling.

The survey instrument consists of questions on vote buying and selling, socio-economic variables, demographics, views on the government and society and access to information, among others. The instrument starts with two screener questions: one on whether the respondent is a registered voter in the city of his/her residence and another on whether he/she or any member of his/her immediate family was ‘approached by a candidate, a representative of a candidate, or just anyone to campaign or to persuade you in any way to vote for a candidate’. The interview proceeds if the respondent answered in the affirmative to both screener questions.

The questions start by asking the respondent whether anything has been offered to him/her or his/her immediate family in exchange for his/her/their vote/s, that is, ‘You mentioned that someone from a political campaign approached you or your immediate family during the last election. Was there anything offered to you or to your immediate family in exchange for your vote or the vote of your immediate family?’. The respondents who answered ‘yes’ were then asked: (a) what was offered; (b) whether the respondent accepted the offer; and c) whether the respondent voted for the candidate for whom the offer was made. This allows us to distinguish whether the transaction was actually completed (i.e., not simply the acceptance of the offer, but also whether the candidate linked to the offer was actually voted).³

The access to information questions consist of inquiries on both quantity and quality of information. Quantity was represented by number of hours reading, listening to and watching news from any source. Quality was proxied by the respondent’s sources of information. We proxy for quality by categorising the news sources. Respondents were first asked for his/her sources of information, with 12 possible non-mutually exclusive answers: (a) social media, (b) Internet blogs, (c) websites of newspapers, (d) websites of TV and radio companies, (e) online-only news outlets, (f) other sources of news and information in the Internet, (g) printed broadsheets, (h) printed tabloids, (i) AM radio stations, (j) non-cable TV channels, (k) cable TV news channels and (l) others. For better comprehension, the respondents were given examples of each. After identifying which on the list the respondents use as sources of news and information, they were asked for their top three sources of information.

These choices were patterned after news and information consumption studies such as the 2014 Digital News Report of the University of Oxford Reuters Institute for the Study of Journalism (Newman & Levy, 2014) and the 2018 News Consumption Survey of the United Kingdom Office of Communications (U.K. Ofcom, 2018), with the choices modified to fit the Philippine context. Sources of information were used as proxies for quality of information because the literature suggests that the quality of news and information from these sources vary; for instance, some tabloids, as opposed to broadsheet newspapers, are generally associated with partially and entirely false stories (Alcott & Gentzkow, 2017) and with sensationalist and generally lower quality news (Conboy, 2006; Rowe, 2010). There have also been questions on the quality of news and information being posted in social media because many of them lack third-party checking and editorial processing (Alcott & Gentzkow, 2017). David et al. (2019) also found evidence that less exposure to social media is associated with greater political knowledge—further suggesting that social media does not tend to be associated with better informed voters. Furthermore, an analysis of fake news stories flagged by fact-checking organisations like PolitiFact and FactCheck.org by Kaplan (2018) suggests that these have been picked up and reported by some radio stations. This suggests that the quality of information relayed through radio may also be suspicious to some extent. To reconfirm these findings from the available literature, we consulted several experts to rank these various sources of information according to their perceived quality based on their expert opinions.⁴

Because of the sensitivity of the topic, the survey instrument was designed in two versions. The first asked direct vote buying questions, while the second used indirect questioning⁵. Although existing literature on vote buying and selling in the Philippines suggests that social desirability bias is low in the country (Cruz et al., 2016, 2017; Khemani, 2013), asking indirect vote buying and selling questions can help validate this. Six hundred respondents were interviewed for each of the two versions.⁶

Estimation Method and Specification

To assess whether information is correlated with the likelihood that the respondent actually accepted the offer in exchange for his/her vote and voted for the candidate for whom the offer was made, we estimated the following econometric equation:

v o t e s e l l_{i} = β_{0} + β_{1} n e w s_{i} + β_{2} s o u r c e_{i} + β_{3} X_{i} + μ_{i}

(1)

where votesell_i represents a dummy = 1 if respondent i accepted the money, food or other material things of significant value offered in exchange for his/her vote and voted for the candidate for whom the offer was made; news_i represents the number of hours per week the respondent spends watching, listening to or reading news from any source; source_i is a vector of 12 dummy variables representing the source of information identified earlier. Each dummy = 1 if the source is among the respondent’s top three sources of information and news. The variable news_i can be thought of as the quantity of information, while source_i refers to the quality of information.

The vector X_i represents a set of control variables that can affect the respondent’s decision to accept the offer and vote for the candidate, including measures of need and quality of life (household income, number of children and expected improvement in the quality of life over the next 3 years); demographics (age, sex and dummy for high school graduates), political values and beliefs (dummy for those who believe that vote buying is acceptable, dummy for those who believe that politicians would know whom you voted for); and perceptions about the community and the government (performance rating of the national government, performance rating of the city government, perception on the prevalence of crime in the city, dummy for respondents who witnessed any form of cheating in a previous election and perception on the quality of public services—water, electricity and public transportation).

For the secondary objective of studying the correlates of being targeted for or receiving a vote buying offer, the following equation was estimated:

\begin{matrix} v o t e b u y_{i} = β_{0} + β_{1} n e e d_{i} + β_{2} d e m o g_{i} + β_{3} p o l i t i c a l_{i} + \\ β_{4} s o c i a l_{i} + β_{5} c l o s e e l e c t_{i} + μ_{i} \end{matrix}

(2)

where votebuy_i represents a dummy = 1 if respondent i received an offer of money, food or other material things of significant value in exchange for his/her vote (but did not necessarily accept the offer nor voted for the candidate). The vector need_i represents a set of variables that measure need, quality of life and income of respondent i. A rich body of literature found evidence that poor voters are disproportionately targeted for vote buying (Carkoglu & Aytac, 2015; Jensen & Justesen, 2014; Schaffer, 2005), primarily because voters with more needs have greater propensity to sell their votes. These variables include income, being a beneficiary of the government’s conditional cash transfer programme, dummy for respondents who experienced a calamity or disaster in the past year, number of family members with jobs, number of children and dummy for those who experienced hunger in the past three months.

The vector demog_i represents the set of demographic indicators for respondent i, that is, age, sex, dummy for at least high school graduates and presence of an overseas worker in the family. These were included because social acceptance of vote buying has been found to be associated with voter characteristics (Gonzalez-Ocantos et al., 2014), including demographics.

The vector political_i represents a set of political beliefs and perception variables, that is, dummy for those who believe that politicians would know whom you voted for, performance rating of public services (water, electricity and public transportation), performance rating of the city government and performance rating of the barangay⁷ government. Some studies argue that vote buyers target voters with zero or low monitoring cost (Brusco et al., 2004; Nichter, 2008; Stokes, 2005), and voters who believe that candidates would know whom they voted for likely do not require much monitoring. A measure of government performance was included because some vote sellers justify this practice because of the poor service that they receive from politicians (Schaffer & Schedler, 2006; Vicente, 2014), and studies have uncovered evidence that more vote buying and selling is associated with lower pro-poor public services (Khemani 2013). In common, vernacular, low-income communities often note that they matter only every 3 years, when their vote is important for politicians.

The vector social_i represents a set of variables of social cohesion as perceived by respondent i. Studies suggest that targeting social networks is crucial to vote buyers because it makes monitoring easier and less costly—communicating and logistics are less challenging and the influence of network members on each other can make vote buying more effective (Cruz, 2019; Wang & Kurzman, 2007). The social cohesion variables are the respondent’s number of years of residency in the barangay, respondent’s perceived level of harmoniousness of his/her relationship with his/her neighbours and how often social events and parties occur in the neighbourhood. Finally, closeelect_i represents a vector of variables that measure the closeness of election in the respondent i’s city and congressional district. It includes two variables: the difference in number of votes between the winner and the second-placed candidate expressed as share of the number of votes of the winning candidate for the mayoral and district representative races. The lower the value of this indicator, the closer the election was for that position. Anecdotal and empirical evidence suggest that certain types of vote buying and selling are more prominently used when the race is closely contested (Canare et al., 2018).

Equations 1 and 2 were estimated using logistic regression and cluster robust standard errors, with clustering in the community level. Because political behaviour tends to exhibit geographical and social patterns (Shiller, 1995), we had to control for possible correlation of errors within the community. Equation 1 was estimated using only respondents who received a vote buying offer, while all observations were used for Equation 2.

The dependent variable of Equation 1 pertains to accepting the vote buying offer and voting for the candidate for whom the offer was made, while that of Equation 2 pertains to receiving an offer in exchange for one’s vote. Because the offers involve things of different value, we considered different vote buying types depending on what was offered. In general, we only considered things of significant value such as money, food and other material things with considerable value such as health cards, cellular phones, shoes and appliances⁸. We further classified offers of significant value into (a) money, (b) food and (c) food plus other materials with considerable value. Equations 1 and 2 were both estimated separately for each of the abovementioned vote buying types: (a) all offers with significant value, (b) money, (c) food and (d) food plus other materials with considerable value.

These equations were estimated using the observations that responded to the direct vote buying and selling questionnaire. Some studies suggest that social desirability bias in asking questions concerning vote buying and selling in the Philippines is minimal, primarily due to the prevalence of the practice—and even its transparency—in the country (Cruz, 2013; Cruz et al., 2016, 2017; Khemani 2013). Moreover, Ibana (1996) argued that poor voters in the Philippines view vote buying and selling as a means to achieve income equality, regarding it as an obligation from politicians. This implies that these practices are acceptable to them, further reducing social desirability bias.

Nonetheless, we compared the responses between the direct and indirect groups to the question on whether they accepted the offer and voted for the candidate for whom the offer was made. Among those who answered in the affirmative to receiving a vote buying offer, we compared the proportion of respondents who answered in the affirmative to the question on accepting the offer and voting for the candidate. A test of comparison of proportions suggests that there is no statistically significant difference between the direct and indirect groups for all vote buying types. This confirms the existing literature that there is little if any social desirability bias involved in vote buying/selling.

Analysis of Empirical Results

Some Survey Statistics on Vote Buying

Before proceeding with the regression results, we first present some descriptive vote buying and selling statistics from the survey data. Slightly less than half (43.2%) of the direct questionnaire respondents said that they received an offer to receive anything with significant value in exchange for voting for a certain candidate. The proportion varies across vote buying types. Sixteen per cent received an offer of money, while 30.5% were offered food. An additional 10.5% were offered other materials of considerable value (e.g., health cards, cellular phones, shoes and appliances). Almost 72% received an offer of things with relatively smaller value (e.g., ballers, caps, shirts, keychains, pens, mugs, fans, towels and calendars); although as discussed earlier, these offers were not the priority for the analysis in this study.

While the proportion of respondents who received an offer varies widely across vote buying types, the proportion of those who accepted the offer and voted for the candidate for whom the offer was made is almost consistent. Seventy-nine per cent of respondents who received an offer of anything with significant value accepted the offer and voted for the candidate. When disaggregated, the proportion is almost the same for money (80%), food (81.4%) and other material things with considerable value (82.5%). This is summarised in Figure 1.

Figure 1.

Source: The authors.

Respondents who voted for the candidate for whom the offer was made were also asked for the reason behind this vote. A large majority (94.6%) said that the qualifications of the candidate motivated this vote. In contrast, only 7.8% said that the offer was a reason for the vote. (Note that multiple responses were allowed.) Moreover, only 4% said that they were supposed to vote for another candidate but changed their mind because of vote buying.

Estimation Results

Accepting the Offer and Voting for the Candidate

The logit average marginal effects of Equation (1) are reported in Table 1. It shows the average marginal effects on the likelihood of accepting a vote buying offer and voting for the candidate for whom the offer was made. It has four columns of results, with the first column representing vote buying using anything of significant value. This was disaggregated into vote buying types in the second to fourth columns—the second column represents vote buying using money; the third column represents vote buying using food; and the fourth column represents vote buying using food and other materials of considerable value. In Table 1, some dummies on the top three sources of information have blank results. The reason is that very few respondents included these dropped variables among their top three sources of information.

Table 1.

Logit Average Marginal Effects on Likelihood of Accepting the Offer and Voting for the Candidate

Variables	(1)	(2)	(3)	(4)
Variables	All Offers with Significant Value	Money	Food	Food and Other Materials with Considerable Value
newshours	−0.00171	0.00135	−0.00115	−0.00334
	(0.00295)	(0.00371)	(0.00325)	(0.00276)
socmed_top3	0.0653	−0.0113	0.123	0.0841
	(0.0565)	(0.0698)	(0.0756)	(0.0638)
tvradioweb_top3	0.104	0.121	0.105	0.0778
	(0.0694)	(0.0916)	(0.0701)	(0.0699)
broadhseetprint_top3	−0.305*		−0.297**	−0.279*
broadhseetprint_top3	(0.183)		(0.138)	(0.157)
tabloidprint_top3	0.158***		0.174***	0.141***
	(0.0439)		(0.0305)	(0.0436)
amradio_top3	0.105**	0.140**	0.120***	0.0951*
	(0.0506)	(0.0616)	(0.0452)	(0.0514)
noncabletv_top3	0.116	0.255	0.0377	0.0468
	(0.136)	(0.270)	(0.0930)	(0.0792)
others_top3	0.0389	−0.00662	0.0999	0.104
	(0.128)	(0.259)	(0.0654)	(0.0732)
income	−2.44e-06	−2.87e-06	−4.44e-06*	−3.24e-06
	(2.75e-06)	(6.27e-06)	(2.32e-06)	(2.24e-06)
children	0.0249**	0.0281	0.0361***	0.0274***
	(0.0115)	(0.0347)	(0.0135)	(0.00986)
male	−0.0779	−0.00410	−0.0506	−0.0465
	(0.0976)	(0.0724)	(0.101)	(0.0935)
age	0.00213	0.00227	0.00128	0.00155
	(0.00224)	(0.00381)	(0.00298)	(0.00237)
highschool	0.100**	−0.0333	0.141***	0.132***
	(0.0455)	(0.132)	(0.0487)	(0.0466)
votebuyacceptable	−0.00476	0.0969	−0.0991	−0.0383
	(0.0675)	(0.0884)	(0.0661)	(0.0705)
knowwhovoted	−0.0579	−0.0545	0.000825	0.0205
	(0.0606)	(0.0788)	(0.0557)	(0.0532)
rating_national	0.00800	0.0190	−0.0212	−0.00213
	(0.0251)	(0.0395)	(0.0339)	(0.0258)
rating_city	0.0202	0.0127	0.0612**	0.0442*
	(0.0300)	(0.0546)	(0.0278)	(0.0261)
crime_prev	0.0513*	−0.0543	0.111***	0.0881***
	(0.0264)	(0.0478)	(0.0417)	(0.0337)
public_service	−0.0212	0.124**	−0.0347	−0.0618*
	(0.0454)	(0.0541)	(0.0419)	(0.0336)
cheatingwitness	0.00274	−0.0549	0.0855	0.0652
	(0.0761)	(0.109)	(0.0533)	(0.0660)
qualitylife3yrs	0.0195	0.0807**	0.0193	0.0130
	(0.0335)	(0.0393)	(0.0338)	(0.0349)
Observations	224	80	159	192
Pseudo R-Squared	0.1260	0.2359	0.2876	0.1978

Source: The authors.

Note: Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05 and * p < 0.1.

The variable newshours, which is the proxy for quantity of information intake, turned insignificant in all regressions. Its coefficient was negative in three of the four runs but were statistically insignificant. This suggests that the likelihood of accepting a vote buying offer and voting for the candidate for whom the offer was made does not vary with the quantity of information consumed by the voter. This is true for all vote buying types.

Although the variable that measures quantity of information turned insignificant, some of the dummy variables representing the voter’s top sources of information turned significant. Survey respondents who included broadsheet newspapers among their top three sources of news and information are 30.5 percentage points less likely to accept a vote buying offer and vote for the candidate for whom the offer was made. The marginal effects are likewise negative and significant for the different vote buying types with almost similar magnitude (29.7 for vote buying using food and 27.9 for vote buying using food and other materials of considerable value).

Some news source variables also turned positive and significant. In contrast to broadsheets, tabloid newspapers have positive and significant marginal effects. Survey respondents who included tabloids among their top three sources of information are 15.8 percentage points more likely to accept a vote buying offer and vote for the candidate. Disaggregating into vote buying types, the marginal effects were 17.4 when the offer is food and 14.2 for food and other materials of considerable value.

Similar to tabloids, AM radio stations also posted positive and significant marginal effects in the overall and disaggregated vote buying types. Including AM radio stations among the top three sources of news and information is associated with 10.5 percentage points higher likelihood of accepting a general vote buying offer and voting for the candidate. For the vote buying types, the marginal effects are 14.0 percentage points for money, 12.0 percentage points for food and 9.5 percentage points for food and other materials of considerable value.

Other top sources of information—social media, websites of TV and radio stations, non-cable TV stations and others—all turned out insignificant. The other sources of information included in the survey were dropped from the regressions because too few respondents included them in their list of top sources of news and information.

Turning the discussion to the control variables, some turned out statistically significant, but their signs and significance depend on the vote buying type. Income, number of children and expectation of improvement in the quality of life in the next 3 years—our indicators of need—are significantly associated with accepting a vote buying offer and voting for the candidate. Higher family income is negatively associated with the likelihood of accepting a food vote buying offer and voting for the candidate, and greater number of children is positive and significant for overall vote buying, food and food and other material things of considerable value. Interestingly, neither income nor number of children were significant in money vote buying. Moreover, expectation in the improvement in quality of life was positive and significant in money vote buying, suggesting that those who are optimistic in quality-of-life improvement are more likely to accept a money vote buying offer and vote for the candidate.

The difference between money and other vote buying types does not stop there. More educated respondents—those who earned at least a high school degree—were more likely to accept the offer and vote for the candidate in all but money vote buying. The education variable, on the other hand, was negative but insignificant in money vote buying. In addition, perception of crime prevalence is positive and significant in vote buying types except money, in which it was negative but insignificant; that is, a respondent who perceives that crime is more prevalent is more likely to accept a vote buying offer (except for money) and vote for the candidate. Another is perception on the quality of public services. Higher quality of public service is associated with higher likelihood of accepting a money vote buying offer and voting for the candidate, but the sign was negative in other vote buying types.

Receiving a Vote Buying Offer

The logit average marginal effects of Equation (2) are reported in Table 2. It shows the average marginal effects on the likelihood of receiving a vote buying offer. The four columns correspond to the different types of vote buying—overall vote buying in column 1 followed by money (column 2), food (column 3) and food and other materials of considerable value (column 4). Indicators of need, particularly income and being a beneficiary of the conditional cash transfer (4Ps) programme, are significant correlates of receiving a vote buying offer. A 4Ps beneficiary is 13.9 percentage points more likely to receive a vote buying offer, while higher income is associated with lower likelihood of receiving an offer. This relationship, however, is mostly driven by money vote buying. Income and being a 4Ps beneficiary are associated with receiving a money vote buying offer, but it has no significant relationship with the two other vote buying types.

Table 2.

Logit Average Marginal Effects on Likelihood of Receiving a Vote Buying Offer

Variables	(1)	(2)	(3)	(4)
Variables	All Offers with Significant Value	Money	Food	Food and Other Materials with Considerable Value
income	−3.46e-06*	−3.63e-06**	−1.36e-06	−1.57e-06
	(1.78e-06)	(1.83e-06)	(1.95e-06)	(2.01e-06)
4ps	0.139***	0.0932**	0.0495	0.0407
	(0.0534)	(0.0447)	(0.0499)	(0.0507)
disaster	0.0472	−0.0322	0.0396	0.0666
	(0.0490)	(0.0353)	(0.0433)	(0.0481)
jobfamily	0.0164	0.0213	−0.00529	−0.00529
	(0.0220)	(0.0231)	(0.0259)	(0.0265)
children	0.00476	0.00688	0.00975	0.0123
	(0.0103)	(0.00776)	(0.0109)	(0.0104)
hunger	0.0177	0.0482	−0.00910	−0.0206
	(0.0568)	(0.0431)	(0.0472)	(0.0519)
age	0.00260	−0.000229	0.00211	0.00268
	(0.00202)	(0.00183)	(0.00191)	(0.00194)
male	0.118**	−0.0117	0.0981*	0.110**
	(0.0563)	(0.0469)	(0.0548)	(0.0552)
highschool	−0.0276	0.0625**	−0.0811*	−0.0927*
	(0.0517)	(0.0313)	(0.0487)	(0.0517)
ofw	0.0554	0.0144	−0.00355	0.00724
	(0.0491)	(0.0396)	(0.0430)	(0.0462)
knowwhovoted	−0.000144	0.0310	−0.0396	−0.0282
	(0.0448)	(0.0400)	(0.0427)	(0.0453)
rating_publicsvc	0.0721**	0.00609	0.0585*	0.0597*
	(0.0360)	(0.0263)	(0.0354)	(0.0350)
rating_city	−0.0449*	−0.0204	−0.0332	−0.0348
	(0.0265)	(0.0173)	(0.0232)	(0.0242)
rating_brgy	0.00982	0.0117	0.0171	0.0145
	(0.0201)	(0.0166)	(0.0206)	(0.0200)
neighbor_social	0.0415*	0.0170	0.0462***	0.0427**
	(0.0219)	(0.0120)	(0.0159)	(0.0217)
neighbor_relation	0.00774	0.0345	−0.0120	0.00398
	(0.0389)	(0.0263)	(0.0354)	(0.0384)
years_brgy	−0.00214	−0.00110	−0.000976	−0.00186
	(0.00194)	(0.00153)	(0.00192)	(0.00183)
mayor_diffvote	−0.00376***	0.000303	−0.00321***	−0.00392***
	(0.00121)	(0.000771)	(0.00117)	(0.00128)
rep_diffvote	−0.00203*	−0.00116*	−0.00201**	−0.00168*
	(0.00105)	(0.000613)	(0.000854)	(0.000984)
Observations	526	526	526	526
Pseudo R-Squared	0.0903	0.0615	0.0875	0.0878

Source: The authors.

Note: Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05 and * p < 0.1.

Among demographic variables, gender and education are significant correlates of receiving an offer, depending on the vote buying type. Males appear to be more targeted for vote buying–they are 11.8 percentage points more likely to receive an offer compared to females. However, this relationship can only be observed in vote buying using food and other material things of considerable value, as it turned insignificant in money vote buying. Money vote buyers also behave differently in targeting more educated voters. High school graduates are more likely to receive money vote buying offers, but they are less likely to receive offers of food and other material things of considerable value.

Some political variables also turned significant. Those who rate public services—water, electricity and public transportation—better are more likely to receive vote buying offers (except money). In contrast, those who rate the city government better are less likely to receive such offers. Social cohesion also showed significant variables. A voter from communities that hold more social events and parties is more likely to receive a vote buying offer (except money). It is possible that social cohesion is highly correlated with the pre-election activities of certain politicians. As noted in earlier literature, vote buying and selling in the Philippines tends to cap already existing politician–voter relationships. Those very same relationships could have been built on activities (e.g., politicians-sponsored events, celebrations, etc.) that also strengthen social cohesion.

A strong correlate of receiving a vote buying offer (all vote buying types) is the closeness of elections in the voter’s city or congressional district. Once again, this confirms the findings of earlier studies that close elections tend to intensify vote buying and selling activities.

Discussions and Implications

The regression results suggest some implications on the relationship between vote buying and access to information and its other correlates. The results suggest that more reliable sources of information—broadsheet newspapers in this case—are associated with lower probability of accepting the vote buying offer and voting for the candidate. On the other hand, less reliable sources (based also on expert opinions on quality and reliability of various sources of information)—tabloids and AM radio stations in this case—are associated with higher probability of completing the vote buying transaction.

It was discussed in earlier sections that there are different mechanisms by which good quality information can provide either an incentive or a disincentive to accept the offer and vote for the candidate. This study found evidence that the disincentive outweighs the incentive. Access to higher quality information is associated with lower likelihood of accepting the vote buying offer and voting for the candidate for whom the offer was made.

As expected, variables that proxy for needs—income and number of children—are associated with accepting the offer and voting for the candidate. Respondents with lower income and greater number of children—even among a sample of already poor households—were more likely to accept the vote buying offer and vote for the candidate. Interestingly, this relationship was not found in money vote buying, but only in vote buying using food and other material things with considerable value. In fact, those who said that they expect an improvement in their quality of life are more likely to accept an offer of money and vote for the candidate.

Moreover, vote buying behaviour is different when money is involved compared to the case when food and other material things of considerable value were offered. One possible explanation for this is that voters, particularly those in the lowest income classes, attach some form of social undesirability on money—those with lower income and greater needs would be more willing to accept food and other material things with considerable value rather than accept money. Another possible explanation is that for voters who are already relatively more well-off (i.e., those who expect their quality of life to improve and those with access to good public services), inducing them to complete the vote buying transaction requires an incentive more valuable than just food and other material things with considerable value. This more valuable incentive could be money.

Furthermore, money vote buying appears to be targeted at poor voters with greater needs. In contrast, needs indicators are not associated with receiving a vote buying offer using food and other material things with considerable value. The concentration of money offers on voters with greater needs is reasonable, as vote buyers try to maximise the returns on their investment.

Voters in socially cohesive communities are also more likely to receive vote buying offers. Vote buyers could be targeting social networks because it makes monitoring easier and because of the influence of the network members to each other (Cruz, 2019; Wang & Kurzman, 2007). Socially cohesive networks amplify these benefits, and as noted earlier, it is also possible that the very same elements of social cohesion are linked to the longstanding relationships with political leaders (e.g., via sponsorship of events and community-building activities).

The last significant set of correlates of receiving a vote buying offer was the closeness of the electoral race. As expected, voters are more likely to receive a vote buying offer in areas where elections are more closely contested. This applies to all vote buying types.

The survey results also have implications on the workings of vote buying in the Philippines. Almost 95% of those who voted for the candidate for whom the offer was made said that the candidate’s qualification is one of the reasons for doing so. In contrast, only around 8% identified the vote buying offer as a reason for voting for the candidate. Moreover, only 4% said that they were supposed to vote for another candidate but changed their mind because of the offer. These suggest that these voters would probably have voted for the candidates anyway. These results support the assertion that vote buying is merely a part of the long process of political clientelism and patronage that starts long before the election. The vote buying transfer enhances the patron–client relationship and induces supporters to vote. The voter accepts them as part of maintaining good relations (Canare et al., 2018; Nichter, 2008; Schaffer & Schedler, 2006).

Conclusions

To conclude, the contribution of this article to the literature is that it focused not only on the possible effect of quantity of information but also of the quality or source of information on vote selling behaviour. In addition, this study examined the other correlates of accepting a vote buying offer and voting for the candidate. It also analysed the correlates of receiving a vote buying offer.

The findings suggest that quantity of information has no significant relationship with the likelihood of accepting the vote buying offer and voting for the candidate for whom the offer was made. The variable that was associated with it was the quality, or source, of information. Voters whose top sources of information are those that generally contain good quality information are less likely to complete the vote buying transaction (accept the offer and vote for the candidate). In contrast, voters whose top sources of information are those that generally contain lower quality information are more likely to do so. The research findings herein also suggest that outside of access to information, the correlates of vote buying behaviour depend on the vote buying type.

There are also interesting correlates of receiving a vote buying offer. Money vote buying appears to be targeted at those with greater needs, while demographic and education variables are also significant correlates depending on the vote buying type. Vote buying offers are also more likely in socially cohesive communities and in areas where elections are closely contested. Over all, these findings suggest a sophisticated strategy to target voters and successfully influence their votes perhaps far in advance of the actual elections.

Finally, the vote buying survey results support the argument that vote buying is merely a part of the long process of political clientelism and patronage that starts long before the election. The vote buying transfer enhances the patron– client relationship and induces supporters to vote.

Footnotes

Acknowledgements

The authors thank Katherine Peralta for the excellent research assistance, the Konrad Adenauer Stiftung for financial support and the anonymous reviewer for the comments.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received financial support from the Konrad Adenauer Stiftung for this research.

Appendix

Table A1.

Summary Statistics and Variables Description

Variable	Definition	Obs.	Mean	Std. Dev.
accept&vote_allvalue	Dummy = 1 if accepted the vote buying offer and voted for the candidate for whom the offer was made (all offers with significant value, i.e., money, food and other materials with considerable value such as cell phones, shoes, health cards and appliances)	259	0.792	0.407
accept&vote_money	Dummy = 1 if accepted the vote buying offer and voted for the candidate for whom the offer was made (money)	95	0.800	0.402
accept&vote_food	Dummy = 1 if accepted the vote buying offer and voted for the candidate for whom the offer was made (food)	183	0.814	0.390
accept&vote_food& other_materials_value	Dummy = 1 if accepted the vote buying offer and voted for the candidate for whom the offer was made (food and other material things with considerable value)	220	0.818	0.387
offered_allvalue	Dummy = 1 if received a vote buying offer (all offers with significant value, i.e., money, food and other materials with considerable value such as cell phones, shoes, health cards and appliances)	600	0.432	0.496
offered_money	Dummy = 1 if received a vote buying offer (money)	600	0.158	0.365
offered_food	Dummy = 1 if received a vote buying offer (food)	600	0.305	0.461
offered_food& other_materials_value	Dummy = 1 if received a vote buying offer (food and other material things with considerable value)	600	0.367	0.482
income	Monthly household income (in PhP)	552	16,474.950	12,349.720
4ps	Dummy = 1 if beneficiary of the conditional cash transfer programme	600	0.240	0.427
disaster	Dummy = 1 if experienced disaster or calamity in the previous year	600	0.372	0.484
jobfamily	Number of family members with job	600	1.413	1.048
children	Number of children	600	3.475	2.198
hunger	Dummy = 1 if experienced involuntary hunger in the past three months	600	0.230	0.421
age	Age	600	45.375	13.891
male	Dummy = 1 if male	600	0.152	0.359
highschool	Dummy = 1 if earned at least a high school degree	600	0.582	0.494
ofw	Dummy = 1 if family has an overseas Filipino worker member	600	0.423	0.494
knowwhovoted	Dummy = 1 if respondent believes politicians will know whom he/she voted for	571	0.196	0.397
rating_publicsvc	Average rating of the following services: water, electricity and public transportation (1—very poor; 2—poor; 3—fair; 4—good; 5—very good)	600	3.497	0.640
rating_city	Satisfaction with the performance of the city government (1—very dissatisfied; 2—somewhat dissatisfied; 3—undecided; 4—somewhat satisfied; 5—very satisfied)	600	4.198	0.870
rating_brgy	Satisfaction with the performance of the barangay government (1—very dissatisfied; 2—somewhat dissatisfied; 3—undecided; 4—somewhat satisfied; 5—very satisfied)	600	3.947	1.061
neighbor_social	How often would there be any parties or social events within your neighbourhood? (1—never; 2—once a year; 3—once a month; 4—2 to 3 times a month; 5—more than 3 times a month)	600	3.058	1.228
neighbor_relation	How would you rate your relationship with your neighbours? (1—very poor; 2—poor; 3—cannot say; 4—harmonious; 5—very harmonious)	600	4.140	0.633
years_brgy	Number of years living in the barangay	600	28.005	15.013
mayor_diffvote	Difference between the number of votes received by the winning and the second-placed mayoral candidate, expressed as percent of the number of votes received by the winner.	600	48.977	25.752
rep_diffvote	Difference between the number of votes received by the winning and the second-placed candidate for representative, expressed as percent of the number of votes received by the winner.	600	60.446	31.368
votebuyacceptable	Dummy = 1 if respondent believes vote buying and vote selling are acceptable	600	0.125	0.331
rating_national	Satisfaction with the performance of the national government (1—very dissatisfied; 2—somewhat dissatisfied; 3—undecided; 4—somewhat satisfied; 5—very satisfied)	600	3.898	0.941
crime_prev	Respondent’s perceived prevalence of crime in the city (1—really not prevalent; 2—not prevalent; 3—can’t say; 4—prevalent; 5—really prevalent)	600	3.038	1.084
cheatingwitness	Dummy = 1 if respondent witnessed any form of cheating in a previous election	600	0.180	0.385
qualitylife3yrs	Has your quality of life improved over the past 3 years? (1—worsened; 2—somewhat worsened; 3—did not change; 4—somewhat improved; 5—improved)	600	4.543	0.690
newshours	Number of hours per week spent reading, listening to and watching news	600	11.068	9.675
socmed_top3	Dummy = 1 if ‘Social media’ are among the top 3 sources of news and information	600	0.358	0.480
internetblog_top3	Dummy = 1 if ‘Internet blogs’ are among the top 3 sources of news and information	600	0.022	0.146
newspaperweb_top3	Dummy = 1 if ‘Websites of newspapers’ are among the top 3 sources of news and information	600	0.007	0.081
tvradioweb_top3	Dummy = 1 if ‘Websites of TV and radio companies’ are among the top 3 sources of news and information	600	0.130	0.337
onlineonlynews_top3	Dummy = 1 if ‘Online-only news outlets’ are among the top 3 sources of news and information	600	0.002	0.041
otheronline_top3	Dummy = 1 if ‘Other sources of news, opinion, and information in the internet’ are among the top 3 sources of news and information	600	0.012	0.107
broadhseetprint_top3	Dummy = 1 if ‘Printed broadsheets’ are among the top 3 sources of news and information	600	0.023	0.151
tabloidprint_top3	Dummy = 1 if ‘Printed tabloids’ are among the top 3 sources of news and information	600	0.088	0.284
amradio_top3	Dummy = 1 if ‘AM radio stations’ are among the top 3 sources of news and information	600	0.283	0.451
noncabletv_top3	Dummy = 1 if ‘Non-cable TV channels’ are among the top 3 sources of news and information	600	0.930	0.255
cablenews_top3	Dummy = 1 if ‘Cable TV news channels’ are among the top 3 sources of news and information	600	0.002	0.041
others_top3	Dummy = 1 if ‘Others’ are among the top 3 sources of news and information	600	0.027	0.161

Source: The authors.

Notes

References

Alcott

, & Gentzkow

(2017). Social media and fake news in the 2016 election. Journal of Economic Perspectives, 31(2), 211–236.

Banerjee

, Kumar

, Pande

, & Su

(2011). Do informed voters make better choices? Experimental evidence from Urban India (Unpublished Manuscript). https://www.povertyactionlab.org/sites/default/files/publications/142%20-%20informed%20voters%20Nov2011.pdf

Bassi

, Morton

, & Williams

(2011). The effects of identities, incentives, and information on voting. Journal of Politics, 73(2), 558–571.

Bond

, Fariss

, Jones

, Kramer

, Marlow

, Settle

, & Fowler

(2012). A 61-million-Person experiment in social influence and political mobilization. Nature, 489, 295–298.

Brusco

, Nazareno

, & Stokes

(2004). Vote buying in Argentina. Latin American Research Review, 39(2), 66–88.

Carkoglu

, & Aytac

(2015). Who gets targeted for vote buying? Evidence from a list experiment in Turkey. European Political Science Review, 7(4), 547–566.

Canare

, Mendoza

, & Lopez

(2018). An empirical analysis of vote buying among the poor: Evidence from elections in the Philippines. South East Asia Research, 26(1), 58–84.

Conboy

(2006). Tabloid Britain: Constructing a community through language. Routledge.

Cruz

(2013). Social networks and the targeting of illegal electoral strategies. 2013 American Political Science Association Annual Meeting, Chicago.

10.

Cruz

(2019). Social networks and the targeting of vote buying. Comparative Political Studies, 52(3), 382–411.

11.

Cruz

, Keefer

, & Labonne

(2016). Incumbent advantage, voter information and vote buying (Working Paper Series, No. 711). Inter-American Development Bank.

12.

Cruz

, Labonne

, & Querubin

(2017). Politician family networks and electoral outcomes: Evidence from the Philippines. American Economic Review, 107(10), 3006–3037.

13.

David

, San Pascual

M. R. S.

, & Torres

M. E. S.

(2019). Reliance on Facebook for news and its influence on political engagement. PLoS ONE, 14(3), 1–15.

14.

De Vries

, Van Der Brug

, Van Egmond

, & Van Der Eijk

(2011). Individual and contextual variation in EU issue voting: The role of political information. Electoral Studies, 30(1), 16–28.

15.

Ferraz

, & Finan

(2008). Exposing corrupt politicians: The effects of Brazil’s publicly released audits on electoral outcomes. Quarterly Journal of Economics, 123(2), 703–745.

16.

Finan

, & Schechter

(2012). Vote buying and reciprocity. Econometrica, 80(2), 863–881.

17.

Fujiwara

, & Wantchekon

(2013). Can informed public deliberation overcome clientelism? Experimental evidence from Benin. American Economic Journal: Applied Economics, 5(4), 241–255.

18.

George

, Gupta

, & Neggers

(2018). Coordinating voters against criminal politicians: Evidence from a mobile experiment in India (Unpublished Manuscript). https://www.povertyactionlab.org/sites/default/files/publications/Coordinating-Voters-Against-Criminal-Investigators-mobile_experiment_November2018.pdf

19.

Ghirardato

, & Katz

(2002). Indecision theory: Quality of information and voting behavior (Working Paper, No. 1106). California Institute of Technology Social Science.

20.

Gonzalez-Ocantos

, De Jonge

C. K.

, & Nickerson

(2014). The conditionality of vote-buying norms: Experimental evidence from Latin America. American Journal of Political Science, 58(1), 197–211.

21.

Gottlieb

(2016). Greater expectations: A field experiment to improve accountability in Mali. American Journal of Political Science, 60(1), 143–157.

22.

Hicken

, Leider

, Ravanilla

, & Yang

(2017). Temptation in vote-selling: Evidence from a field experiment in the Philippines. Journal of Development Economics, 131, 1–14.

23.

Hobolt

(2007). Campaign information and voting behaviour in EU referendums. In De Vreese

(Ed.), The dynamics of referendum campaigns: An international perspective (pp. 84–114). Palgrave-Macmillan.

24.

Houser

, Morton

, & Stratmann

(2011). Turned on or turned out? Campaign advertising, information and voting. European Journal of Political Economy, 27(4), 708–727.

25.

Ibana

(1996). Epilogue: Ordinary people in everyday life. In Alejo

, Rivera

, & Valencia

(Eds.), [De]scribing elections: A study of elections in the Lifeworld of San Isidro (pp. 125–138). Institute for Popular Democracy.

26.

Jensen

, & Justesen

(2014). Poverty and vote buying: Survey-based evidence from Africa. Electoral Studies, 33(1), 220–232.

27.

Kaplan

(2018). Radio stations across North America have repeatedly spread fake news. Media Matters for America. https://www.mediamatters.org/fake-news/radio-stations-across-north-america-have-repeatedly-spread-fake-news

28.

Keefer

(2007). Clientelism, credibility, and the policy choices of young democracies. American Journal of Political Science, 51(4), 804–821.

29.

Keefer

, & Khemani

(2014). Radio’s impact on preferences for patronage benefits (Working Paper, No. 6932). World Bank Policy Research.

30.

Khemani

(2013). Buying votes Vs. supplying public services political incentives to under-invest in pro-poor policies (Working Paper, No. 6339). World Bank Policy Research.

31.

Kitschelt

, & Wilkinson

(2007). Citizen-politician linkages: An introduction. In Kitschelt

& Wilkinson

(Eds.), Patrons, clients and policies: Patterns of democratic accountability and political competition (pp. 1–49). Cambridge University Press.

32.

Kramon

(2016). Electoral handouts as information: Explaining unmonitored vote buying. World Politics, 68(3), 454–498.

33.

Larcinese

(2007). Does political knowledge increase turnout? Evidence from the 1997 British general election. Public Choice, 131, 387–411.

34.

Lassen

(2005). The effect of information on voter turnout: Evidence from a natural experiment. American Journal of Political Science, 49(1), 103–118.

35.

Lupia

(1994a). Shortcuts versus Encyclopedias: Information and voting behavior in California Insurance reform elections. American Political Science Review, 88(1), 63–76.

36.

Lupia

(1994b). The effect of information on voting behavior and electoral outcomes: An experimental study of direct legislation. Public Choice, 78(1), 65–86.

37.

Lyne

(2007). Rethinking economics and institutions: The voter’s dilemma and democratic accountability. In Kitschelt

& Wilkinson

(Eds.), Patrons, clients, and policies: Patterns of democratic accountability and political competition (pp. 206–226). Cambridge.

38.

Newman

, & Levy

D. A. L.

(2014). Digital news report 2014. University of Oxford Reuters Institute for the Study of Journalism.

39.

Nichter

(2008). Vote buying or turnout buying? American Political Science Review, 102(1), 19–31.

40.

Rowe

(2010). Tabloidization of news. In Allan

(Ed.), The Routledge companion to news and journalism (pp. 350–361). Routledge.

41.

Schaffer

(2005). Clean elections and the great unwashed: Vote buying and voter education in the Philippines (Occasional Paper, No. 21). Massachusetts Institute of Technology School of Social Science.

42.

Schaffer

, & Schedler

(2006). What is vote buying? In Schaffer

(Ed.), Elections for sale: The causes and consequences of vote buying (pp. 17–32). Lynne Rienner Publishers.

43.

Shiller

(1995). Conversation, information, and herd behavior. American Economic Review, 85(2), 181–185.

44.

Stokes

(2005). Perverse accountability: A formal model of machine politics with evidence from Argentina. American Political Science Review, 99(3), 315–325.

45.

U.K. Ofcom (Office of Communications). (2018). News consumption in the UK: 2018. U.K. Office of Communications.

46.

Vicente

(2014). Is vote buying effective? Evidence from a field experiment in West Africa. The Economic Journal, 124(574), 356–387.

47.

Vicente

, & Wantchekon

(2009). Clientelism and vote buying: Lessons from field experiments in African elections. Oxford Review of Economic Policy, 25(2), 292–305.

48.

Wang

, & Kurzman

(2007). The logistics: How to buy votes? In Schaffer

(Ed.), Elections for sale: The causes and consequences of vote buying (pp. 61–62). Lynne Rienner Publishers.

49.

Wantchekon

(2003). Clientelism and voting behavior: Evidence from a field experiment in Benin. World Politics, 55(3), 399–422.