ilable at ScienceDirect

Electoral Studies 27 (2008) 621–632

Contents lists ava

Electoral Studies

journal homepage: www.elsevier .com/locate/e lectstud

Vote buying and violence in Nigerian election campaignsq

Michael Bratton*

Department of Political Science, Michigan State University, 323 South Kedzie Hall, East Lansing, MI 48824, USA

a r t i c l e i n f o

Article history:Received 11 October 2007Revised 21 April 2008Accepted 23 April 2008

Keywords:ElectionsVote buyingViolencePublic opinionDemocracyNigeriaAfrica

q Thanks are due to Fred Schaffer and Macartanments on an earlier draft of this paper. Remaining e

* Tel.: þ1 517 353 3377; fax: þ1 517 432 1091.E-mail address: mbratton@msu.eduURL: http://www.msu.edu/~mbratton

0261-3794/$ – see front matter � 2008 Elsevier Ltddoi:10.1016/j.electstud.2008.04.013

a b s t r a c t

Vote buying and political intimidation are characteristic dimensions of African electioncampaigns. According to survey-based estimates, almost one out of five Nigerians is per-sonally exposed to vote buying and almost one in ten experiences threats of electoral vi-olence. But when, as commonly happens, campaign irregularities are targeted at the ruralpoor, effects are concentrated. These effects are as follows: violence reduces turnout; andvote buying enhances partisan loyalty. But, perhaps because most citizens condemn cam-paign manipulation as wrong, compliance with the wishes of politicians is not assured. De-fection from threats and agreements is more common than compliance, especially wherevoters are cross-pressured from both sides of the partisan divide.

� 2008 Elsevier Ltd. All rights reserved.

In a democracy, an election campaign is supposed to bea peaceful and open discourse of persuasion. Ideally, candi-dates compete for popular support by presenting reasonedarguments about why they are most qualified for electionto office. They stake out rival positions about programs ofpublic goods, all the while being tugged towards themedian voter at the center of the political spectrum(Downs, 1957). Voters then choose the contender whosepolicy positions most closely mirrors their own set ofpreferences.

In Africa, however, elections more closely resemble bit-ter struggles over access to the resources controlled by thestate, which are the biggest prize in society. Given thesehigh stakes, politicians resort to a variety of means –whether fair or foul – to attain public office. To be sure,candidates go through the motions of presenting program-matic promises. But the pledges of politicians commonly

Humphries for com-rrors are my own.

. All rights reserved.

lack credibility (Keefer, 2004), are virtually identical acrosscontending parties (Mohammed Salih and Nordlund,2007), or quickly dissolve into personal attacks (van deWalle, 2003). In this context, voters choose among candi-dates less on the basis of distinctive policy positions thanon the candidates’ assumed trustworthiness and reliabilityas prospective patrons (Kitschelt and Wilkinson, 2007; vande Walle, 2007; Bratton and Lewis, 2007; Bratton, 2007).

Instead of providing opportunity for public deliberation,African election campaigns are mainly moments for politi-cians to engage in mass mobilization and the manipulationof electoral rules. All too often, campaign strategies featurematerial inducement and political intimidation. In extremeforms, unconventional modes of electoral practice aremanifest in explicit acts of vote buying (Schaffer, 2007;Lindberg, 2003) and electoral violence (LeBas, 2006;Wilkinson, 2004). Both kinds of activity, which aim todeny citizens the freedom to express their electoral prefer-ences, are incontrovertibly illegal. But, because persuasionalone seldom generates enough support, candidates none-theless regularly attempt to purchase or compel votes.

The purpose of this article is to explore the nature, ex-tent and effectiveness of irregular modes of electioneering.What forms – in cash or kind – does vote buying take?

M. Bratton / Electoral Studies 27 (2008) 621–632622

What sorts of practices – of threats or action – characterizeelection violence? How frequently do these departuresfrom democratic procedure occur? Who are the victims?And, most importantly, do inducements and compulsionswork? If they do, which are more effective: the carrots ofvote buying or the sticks of political violence?

In order to estimate the effectiveness of different sortsof electoral malpractice, it is necessary to measure theirconsequences for electoral behavior at the individual level.To this end, I distinguish three alternative courses of actionfor citizens: to refuse, to defect, or to comply. First, the votercan ‘‘refuse.’’ With reference to vote buying, the individualcan decline to enter into an agreement to trade his vote. Orshe can seek to avoid violence, for example by publiclyshaming the perpetrator or, more likely, exiting the elec-toral arena. Second, the individual can enter a vote buyingagreement or shoulder a threat of intimidation with no in-tention of complying. When the time comes to cast a ballot,the individual instead ‘‘defects’’ by failing to vote at all or byvoting as he or she pleases. Finally, the citizen can ‘‘comply’’with inducement or intimidation by turning out to vote andvoting the ‘‘right’’ way, meaning in accordance with the in-structions of the vote buyer or political persecutor. Thesethree possible responses – which echo Hirschmann’s(1970) alternatives of exit, voice and loyalty – representan ascending scale of popular submission to elite effortsto manipulate elections.

In this article, I show that citizens in one African countryclearly regard vote buying and electoral violence as infrac-tions of public morality. Most ordinary people resist effortsof political elites to illegally influence voter behavior. Butsome individuals – especially society’s poorest and mostvulnerable members – have little choice except to comply.Faced with irregular carrots or sticks during the course ofan election campaign, their only other viable option is tofeign compliance while refusing in practice. I show thatthis strategy – which I have called defection – is a common-place ‘‘weapon of the weak’’ (Scott, 1985).

Importantly, the available evidence suggests that votebuying and political intimidation are ineffective campaignpractices. In reality, people who are paid or threatened dur-ing the election campaign are actually less likely to turn outto vote on polling day. Threats of violence lead to an espe-cially sharp reduction in voter turnout. Moreover, whilevoters may be willing to cast their ballots for parties whosecandidates have broken electoral laws, many would haveexpressed such support anyway, that is, without extra-legalincentives or punishments. Most importantly, many whoenter vote-buying agreements say that they will ultimatelydefect, that is, by taking the money but voting as theyplease. Defection is especially likely when voters arecross-pressured from both sides of a partisan divide orwhen exposed to both vote buying and violence.

The article ends by suggesting implications of thesecampaign malpractices for the health of democracy.One particularly interesting result, that deserves furtherexploration, is that vote buying behavior is determined col-lectively. People are most likely to defect if they think thatothers will do so too, thus availing themselves of the pro-tections provided by collective action. But if collectiveaction also shapes norms – that is, people justify wrongful

behavior for themselves because everyone else is doing it –then campaign irregularities can corrode the quality ofdemocratic citizenship.

1. The campaign context, Nigeria 2007

Nigeria’s general elections of April 2007 – which fea-tured contests for the federal presidency, state governor-ships and legislative assemblies at state and nationallevels – promised a political watershed (Mustapha, 2006;Ibrahim, 2007; Rotberg, 2007). For the first time since inde-pendence in 1960, a third sequence of regular electionswould be held under a civilian regime and one elected Pres-ident would succeed another. If peacefully and honestlyimplemented, the elections held out the prospect of legiti-mizing and strengthening Nigeria’s fragile new democracy.But the serious misconduct observed in Nigeria’s previous1999 and 2003 polls cast doubt on whether Nigeria wouldeasily attain a free and fair election.

In the event, Nigeria’s 2007 elections were deeplyflawed (Human Rights Watch, 2007). Before the campaigneven began, seasoned observers correctly predicted that,as ‘‘various powerful figures calculate their best interestsand shift their factional alignments.tremendous amountsof largesse will change hands and some of the players willlikely resort to force’’ (Sklar et al., 2006, p. 108). A bitterfeud between outgoing President Obasanjo and Vice-President Atiku Abubakar, an aspirant successor, domi-nated the election season. Dueling lawsuits, boycottthreats, and shifting alliances between opportunistic polit-ical parties and factions created a chaotic atmosphere ofuncertainty. Election preparations – such as a delayed voterregistration exercise – were woefully inadequate and ques-tions soon arose about the impartiality and competence ofthe Independent National Election Commission (INEC)(Transition Monitoring Group, 2007).

These concerns were borne out on polling days, April 14and 21, 2007. Voting for President and National Assemblyfailed to take place in certain polling stations in a half dozenstates in the southeast and northeast due to the non-deliv-ery of electoral materials. In numerous other locationsacross the country, ballot papers were misprinted or ar-rived late. In the southern Niger Delta zone, armed militiasbrazenly stole ballot boxes or substituted pre-stuffed con-tainers of their own. Despite guarantees from the InspectorGeneral of Police that public security would be assured, op-position candidates were harassed or arrested, voters wereturned away from polling places by gangs of young thugs,ballot secrecy was violated by party workers and police,and some 300 persons were killed in election relatedviolence. INEC announced an overwhelming victory forthe ruling Peoples’ Democratic Party (PDP), but polling sta-tion results often bore little resemblance to actual turnoutor voter intentions. All told, the fraudulent election leftNigeria’s voters feeling ‘‘frustrated’’ and ‘‘disenfranchised’’(Economist, 2007).

This article seeks to document some neglected aspectsof an election campaign awash with oil money and marredby escalating violence. Unlike most political analysts ofNigeria, I focus on ordinary citizens rather than politicalelites. And unlike most election observers, I concentrate

M. Bratton / Electoral Studies 27 (2008) 621–632 623

on the election campaign rather than polling and countingprocedures.

The object of interest is not wholesale electoral manip-ulation as represented by legal maneuvers to keep candi-dates off the ballot or administrative fixes to falsify votetallies, which also occurred. Instead, I look at campaignabuses at the retail level, one citizen at a time. Who wasaffected, and how did they respond?

Data are drawn from a pre-election Afrobarometer sur-vey conducted in Nigeria in January and February 2007with questions about the previous 2003 and upcoming2007 general elections. With a representative nationalsample of 2410 adult respondents, it is possible to make in-ferences to the Nigerian population as a whole with a mar-gin of sampling error of �2% at a 95% confidence level.1

2 The remaining 5% ‘‘don’t know’’ whether vote buying is right orwrong.

2. The perceived morality of campaign manipulation

As a first step, I ask whether African citizens regard votebuying and violence as right or wrong. One might expectdisagreement on this issue, with at least some surveyrespondents regarding these tactics as legitimate – or atleast acceptable – methods of electioneering. One mightexpect that voters would ascribe a different moral statusto each transgression, with violence being seen as more un-ambiguously wrong than vote buying. The logic here is thatviolence visits heavy costs upon unwilling victims, whereasthe purchase of votes at least holds out the promise of a ma-terial benefit to those who voluntarily participate.

A clear majority of the Nigerians we interviewed saidthat political violence was always wrong. Almost four outof five adult Nigerians (79%) see political violence as ‘‘neverjustified,’’ even ‘‘in support of a just cause.’’ The same pro-portions think it better to find lawful solutions to socialproblems rather than to ‘‘use other means’’ and that ‘‘poli-ticians and political parties should not be allowed to formtheir own private security forces.’’ Only 5% of Nigeriansstrongly support the ‘‘necessity’’ of using violence in pur-suit of political goals. As might be expected, younger andless educated people are somewhat more tolerant of anaggressive approach to politics, but the differences arenot statistically significant.

Most Nigerians also condemn efforts by politicians topurchase support at the polls. Almost six in ten (58%) saythat it is ‘‘wrong and punishable’’ for ‘‘a candidate or partyofficial to offer money in return for a vote.’’ An additional30% consider vote buying ‘‘wrong but understandable,’’adding the qualifying phrase perhaps because they thinkthat political patrons are obliged to steer kickbacks theirclients. van de Walle has suggested that, in Nigeria, voterstake vote buying offers as signals of a patron’s wealth andcapability of winning elections, features of a leader withwhich they wish to be associated (Kitschelt and Wilkinson,2007, p. 64; see also Chabal and Daloz, 1999, pp. 39–44).Less charitably, Banegas (1998, p. 78) reports that votersin neighboring Benin see the payment of money for votes

1 For full technical details on Afrobarometer objectives, organization,questionnaires, samples, response rates, and publications, as well as forthe data on which this paper is based, see www.afrobarometer.org.

as reparation for public funds that politicians are assumedto have stolen.

But, importantly, only 7% of Nigerians would go so far asto characterize vote buying as ‘‘not wrong at all.’’ In otherwords, even those who regard vote buying as understand-able still regard it as wrong. And education is a powerfulsolvent to moral acquiescence: even people with primaryeducation are only half as likely as those without formalschooling to see candidates making handouts as ‘‘notwrong at all.’’

But, when they look in the mirror, Nigerians are less crit-ical of their own behavior. Barely half (49%) think that it is‘‘wrong and punishable’’ for a voter to ‘‘accept money inreturn for a vote.’’ The other half of the adult populationis willing to excuse participation in a vote buying transac-tion as ‘‘wrong but understandable’’ (35%) or ‘‘not wrongat all’’ (10%).2 The main extenuating circumstance is pov-erty. People on the lowest rung of a five point poverty scaleare only half as likely as those on the top rung to say thatthe sale of votes is ‘‘wrong and punishable.’’3

Between the two electoral violations, which is regardedas the greater evil? Nigerians clearly see electoral violenceas a larger political problem than vote buying. When askedabout the most important issues in the upcoming 2007national elections, many more Nigerians demanded that‘‘Nigerians should be secure from violence’’ than that‘‘vote buying should be controlled’’ (14 versus 2% of all elec-tion issues mentioned). Taken together, the above findingssuggest that Nigerians are resistant to both vote buying andelectoral violence on moral grounds. But they worry moredeeply about political intimidation and many are inclinedto forgive voters for succumbing to campaign inducements.

3. The frequency of campaign irregularities

I now attempt to estimate the frequency and distribu-tion of vote buying and violence in Nigerian election cam-paigns. Data are generated from straightforward surveyquestions about whether, during national election cam-paigns in 2003 and 2007, individuals encountered offersof ‘‘something.(money, food or a gift). in return foryour vote’’ or threats of ‘‘negative consequences in orderto get you to vote a certain way.’’ We also asked respon-dents to report whether they thought ‘‘other people inyour neighborhood or village’’ had had such encounters.

By February 2007, mid-way through the general elec-tion campaign, some 12% of Nigerians interviewedacknowledged that a candidate or a party agent had offered‘‘something in return for your vote.’’ This level of directexperience with vote buying was slightly lower than thatrecalled for the 2003 general elections, at 16%. But the2007 figure captured only half the campaign period,whereas the 2003 figure covered the entire campaign. We

3 The poverty scale is the Afrobarometer’s Index of Lived Poverty, anaverage index of shortages of basic needs: food, medical care, clean water,cooking fuel and cash income. See ‘‘Poverty, Poverty Measurement, andDemocracy in Southern Africa’’, Afrobarometer Working Paper No. 4(2003), www.afrobarometer.org.

M. Bratton / Electoral Studies 27 (2008) 621–632624

know from research in Taiwan that the pace of vote buyingaccelerates as the day of the election approaches (Wangand Kurzman, 2007), so it is reasonable to project thatvote buying in Nigerian elections was at least as frequentin 2007 as in 2003, and quite possibly more so.

Moreover, vote buying may be more extensive in bothcampaigns than implied by the figures cited. Experiencefrom Argentina suggests that some people are understand-ably reluctant to admit that they had been approached witha forbidden offer, especially if they had subsequently en-tered an agreement and complied with its terms (Bruscoet al., 2004). The existence of undercounting in self-assess-ments is reflected in the respondents’ 2007 estimate thatfully 28% of other voters were offered gifts during the2003 campaign. I therefore consider that the ‘‘true’’ levelof vote buying exists within a zone bracketed by a wideconfidence interval. I contend that the real frequency ofthis activity probably lies somewhere between reportedlevels of personal experience (12% in Nigeria in 2007,which may be an underestimate), and assumed levels of in-volvement by fellow citizens (28%, which may be anoverestimate).

In vote-buying transactions in Nigeria, voters are usuallyoffered money (68% of all reported attempts in 2007), com-modities (such as food or clothing, 26%) or jobs (6%). In thelatest and previous Nigerian elections, the modal (i.e., mostcommon) inducement was 500 naira, or about US$4.4 Butthe median price of a vote payment rose between 2003and 2007, from 1750 naira to 2250 naira, largely becausethe proportion of large payments (10,000 naira or moreper vote) increased over time.

Relatively few Nigerians report being directly affected byelectoral violence. By February 2007, just 4% of survey respon-dents nationwide said that they had received ‘‘threats of neg-ative consequences.’’ To be sure, political intimidation wasgeographically concentrated in certain electoral ‘‘hot spots,’’especially the Niger Delta region of the South-South zone,where the Afrobarometer recorded personal and observed ex-periences with electoral violence at rates more than threetimes as high as the national average.5

The negative consequences of political intimidationinclude – in approximately equal proportions – threats topersonal safety, threats to the safety of family members,and the loss of property. For some reason, respondentsthought that intimidation was more likely to take theform of the loss of a job for other voters than for them-selves. Reports of violent encounters in 2007, mid-waythrough the campaign, were almost as frequent as for thewhole 2003 campaign (5%). If one assumes that the paceof violence also quickens as the vote approaches, one caninfer that the actual level of electoral violence was ulti-mately higher in 2007 than 2003 (see also Human RightsWatch, 2007). And because the gap between reported per-sonal and estimated third-party abuses (6 percentage

4 In 2007, n¼ 262; in 2003, n¼ 390. When the same question wasasked in Kenya three weeks before the December 29, 2007 election, themodal amount reported was a little over US$3 (n¼ 347).

5 Some 13% of South-South residents claimed personal experience withintimidation by February 2007 and some 18% saw the same among theirneighbors.

points in 2007) was smaller for violence than vote buying,we can invest greater confidence in the violence data.

If people have relatively few direct experiences with po-litical intimidation, then why do they see violence as sucha pressing campaign concern? My working assumption isthat an atmosphere of threat, regardless of whether hostil-ities are ever directly experienced, has a generally chillingeffect on the public mood. In a related survey from January2007, Nigerian NGOs reported that more than half (56%)reported that they were at least ‘‘a little fearful’’ of ‘‘becom-ing a victim of intimidation or violence in the forthcomingelections.’’6 In the Afrobarometer survey, a similar propor-tion (54%) thought that the previous 2003 elections hadbeen unfree and unfair due to ‘‘the use of violence byparties and candidates.’’7 And an even larger proportion(71%) said that competition between political parties‘‘often’’ or ‘‘always’’ leads to violent clashes. Intriguingly,this widespread expectation of impending conflict wasnegatively related to actual experiences. But it was none-theless positively related to the likelihood that a citizenwould see violence as a prominent campaign issue.

4. Who are the victims?

It seems reasonable to assume that, in seeking to controlvoter behavior, politicians would focus their efforts on themost vulnerable elements in society. One would thereforeexpect a disproportionate concentration of bribery andviolence on poor and uneducated people. The poor arelikely to be victimized by vote buying because their limitedmeans make them susceptible to material inducements,including offers of basic commodities or modest amountsof money. For their part, people with limited educationmay be unaware of individual political rights and thereforepossess weak defenses against intimidation.

In Nigeria, however, the survey revealed few demo-graphic correlates of exposure to electoral violence. Politicalintimidation was spread rather evenly across all socialgroups, whether rich or poor, urban or rural, even male orfemale. This smooth distribution constitutes furtherevidence that violence is a general ‘‘atmospheric’’ conditionthat tends to affect everyone touched by an election cam-paign. In the only observable statistical effect, education per-formed as expected: it tended to inoculate Nigerians againstexplicit threats of ‘‘negative consequences’’ for making the‘‘wrong’’ vote choice. Whereas 5% of people with no formalschooling experienced such threats, just 2% of those withpost-secondary education did so.8 So people who are illequipped to defend their rights are prone to be victimized,even by their own political patrons. But this conclusionshould be treated with caution due to small sub-sample sizes.

By contrast, there were clear demographic correlates ofvote buying. In both 2003 and 2007, poor Nigerians weremost likely to report an encounter with a politician (or

6 Alliance for Credible Elections and CLEEN (2007).7 Moreover, past experience influences future expectations. Those who

saw the last elections (2003) as less than free and fair are likely to projectthe same for upcoming elections (2007) (r¼ 0.364, p< 0.001).

8 Gamma¼�0.196, p¼ 0.014.

M. Bratton / Electoral Studies 27 (2008) 621–632 625

a politician’s agent) who offered to buy their vote.9 Let uscompare people who suffered ‘‘many’’ shortages of basicneeds in the previous year with those who covered theseneeds: the former were over four times more likely thanto be ‘‘often’’ approached with a vote-buying offer. But,even if poor Nigerians were more frequently exposed tomaterial inducements, they still drove a hard bargain.They did not quote a lower price than the going rate forselling a vote. Nor were they more likely than anyone elseto accept payment in the form of goods-in-kind as opposedto cold, hard cash.

Moreover, in 2007, educated Nigerians were less likelyto report that anyone had approached them with a vote-buying offer.10 Those with post-secondary education werefive times less likely to report such an encounter than thosewithout any formal schooling. I have already shown thateducated people are especially inclined to find vote buyingmorally wrong. A more cynical view would contend thateducated people, especially those with paid employment,can afford to be morally upright. To be sure, educated peo-ple are not entirely immune from temptation. It is worthnoting in this regard that those with education are almosttwice as likely as those without formal schooling to thinkthat a vote is worth 10,000 naira or more.

At face value, it is unclear whether vote buying would bemore prevalent in urban or rural areas. On one hand, poorand uneducated people are concentrated in rural villages,making these areas prime targets for the distribution ofpatronage. On the other hand, outlying areas are hard forpoliticians to reach and monitor, which suggests that ruraldwellers can easily refuse or defect from reward-drivenagreements. The Afrobarometer data show that, in Nigeria,vote buying is much more common in rural than urbanareas by a margin of 7 percentage points in 2003 and 6 per-centage points in 2007. These results are consistent withresearch from East Asia and Latin America (Ramsayer andRosenbluth, 1993; Shugart and Nielsen, 1999).11 Indeed, in2003, residential location was the most important demo-graphic consideration of all, trumping both poverty and ed-ucation. By 2007, however, education and urban residencewere equally important considerations in reducing thelikelihood of vote buying.

5. Voter behavior

This article now turns to voter behavior. Two types areexamined: voter turnout and partisan choice. Because theAfrobarometer survey was conducted before the electionin February 2007, the indicators of these behaviors areproxy measures based on voting intentions in the electionsscheduled for April 2007.

Voter turnout is measured by a question that asked re-spondents to situate themselves in relation to a range of

9 As Stokes (2007) reasons, with reference to the diminishing marginalutility of income, ‘‘the same outlay of resources.will by more votesamong poor than among wealthy voters.’’

10 This relationship did not hold in 2003.11 Hicken (2007) suggests that urbanization ‘‘destroys the traditional

patron-client networks through which candidate funds can be distrib-uted.’’

behaviors running from ‘‘I am not a registered voter and Iam not interested in voting’’ to ‘‘I am a registered voter andI will vote.’’ Intended voters are those who chose the last op-tion in 2007. As Table 1 shows, voter interest in elections washigher in 2007 than in 2003, as evidenced by higher reportedrates of voter registration and lower levels of planned ab-stention. Even discounting the block of eligible voters (12%)who in February 2007 had not yet decided whether tovote, it appeared that voter turnout, which then stood atan intended 66%, was likely to rise above 2003 levels.12

Partisan choice is measured by a question that asked, ‘‘Ifelections were held tomorrow, which party’s candidatewould you vote for as member of the national assembly?’’In summary form, this variable is coded on a three-pointscale running from incumbent party partisan, throughnon-partisan, to opposition party partisan. As Table 2shows, voters were split in February 2007 between sup-porters of the incumbent PDP (one-third) and supportersof all other opposition parties (a combined two-fifths).13

But, alone, the largest opposition party, the All NigeriaPeoples’ Party (ANPP), did not attract even one-quarterof the intended votes, which suggests that neither incum-bents nor opposition held a decisive edge in most nationalraces.

6. How effective are campaign manipulations?

How do citizens respond to the mix of less-than-legalcarrots and sticks employed by politicians in African elec-tion campaigns? Are vote buying and violence effectivestrategies of electoral manipulation?

6.1. Voter turnout

The first object to be explained is voter turnout, mea-sured as those among all eligible voters who expressedan intention to vote in Nigeria in 2007. The key explanatoryvariables are experiences with vote buying and politicalintimidation during the 2007 election campaign. Do thesemalpractices affect turnout? As preliminary controls, Iinclude partisan preference and a standard set of demo-graphic factors.

The results of a logistic regression analysis are shown inTable 3, of which several are noteworthy. Consider thecontrol variables. First, while poverty and education per-form as expected – the former is negative for turnout andthe latter is positive – neither is statistically significant socan be discounted for the purposes of this analysis. Second,rural residence remains important for turnout, but in theunexpected direction that, in Nigeria as in other parts ofAfrica and the wider agrarian world, country dwellers are

12 Because INEC has not published (and may never be able to publish)comprehensive and credible data on Nigeria’s 2007 election, it is not pos-sible to verify these estimates against official statistics. But the reliabilityof Afrobarometer’s 2007 indicator of intended voter turnout is partly con-firmed its strong correlation with reported actual voter turnout by thesame individuals in 2003 (r¼ 0.328).

13 The reliability of the partisan preference indicator is partly confirmedby its correlation with an individual’s choices for National President, Na-tional Assembly, State Governor and State Assembly (r¼ 0.779 or above).

Table 3Determinants of voter turnout, Nigeria 2007

B S.E. Sig.

Constant 0.773 0.327 0.016

ControlsPoverty �0.055 0.053 0.297Education 0.017 0.023 0.449Rural 0.427 0.098 0.000Age 0.009 0.004 0.019Female �0.511 0.095 0.000Incumbent partisan �0.035 0.055 0.518

Campaign malpracticesExperience of votebuying offer

�0.121 0.086 0.163

Experience of threat of violence �0.796 0.137 0.000

Nagelkerke r2¼ 0.069 (n¼ 2229).

Table 1Voter turnout (2003) and intention to vote (2007), Nigeria, February 2007

2003 2007

Don’t know 3 3I am not registered to vote, and I am not

interested in voting11 5

I am not registered to vote, but would liketo have voted/to vote

13 8

I am registered, but I chose not to vote(2003)/choose not to vote (2007)

11 5

I am registered to vote, but I have not yetdecided whether to vote

– 12

I am registered to vote and I voted (2003)/I will vote (2007)

62 66

Cell entries are percentages of survey respondents, i.e., adult Nigerians(n¼ 2410).

M. Bratton / Electoral Studies 27 (2008) 621–632626

more likely to vote than their urban counterparts (Yadav,2000; Bratton et al., 2007; Krishna, 2005). Third, otherdemographic considerations now enter the analysis: olderpeople are significantly more likely to vote than youngsters;and women are very much less likely to vote than men.

Political partisanship, however, is unrelated to voterturnout. If there is any tendency, it is for supporters ofthe incumbent party to be more complacent about bother-ing to vote than opposition partisans, but the relationship isnot statistically significant.

Our main interest is in campaign irregularities. Table 3shows that individuals who experience a vote buying offerare less likely to vote than those who do not. One possibleinterpretation is that recipients of such offers feel ambiva-lent: they agonize whether to comply with the wishes ofthe vote buyer or to act according to conscience. To resolvethis dilemma, individuals sometimes avoid voting alto-gether. In other words, they engage in a form of defection.An alternative interpretation is that, in some cases, votebuyers succeed in their aim of preventing their opponents’supporters from casting a ballot. This possibility – whichthen must be interpreted as compliance – arises wherevervoters report that they have sold their voter registrationcards in return for a payment (Vicente, 2007).

But the negative effect of vote buying on voter turnout issmall and, in the current analysis, does not attain statisticalsignificance. Thus, the Nigeria data do not validate anyclaim to the effect that vote buying ‘‘works,’’ at least inthe limited sense of boosting voter turnout.

Much more powerful is the effect of threatened cam-paign violence. As Table 3 shows, the effect is again negative,but now it is strong and statistically significant. For an aver-age Nigerian (say, a female rural dweller) and with other

Table 2Partisan choice, Nigeria, February 2007

Incumbentpartisan

Non-partisan(inc. ‘‘don’t know’’)

Oppositionpartisan

National President 33.3 23.0 43.7National Assembly 34.1 27.8 38.1State Governor 36.0 23.0 41.0State Assembly 32.5 28.6 38.9Mean 34.0 25.6 40.4

Cell entries are percentages of survey respondents, i.e., adult Nigerians(n¼ 2410).

variables controlled at their mean level, a threat ofviolence reduces the odds of intending to vote by 52%.14

Moreover, intimidation’s effect seems to be long lasting,since the model works almost as well if exposure toviolence is measured for 2003 rather than 2007.15 To all ap-pearances, Nigerians who encounter a threat against votingfreely often withdraw from the electoral process entirely,that is, they abstain from voting, perhaps persistently. So po-litical intimidation apparently has an intended effect: itmakes citizens so fearful that they abandon their right tovote.

The question arises whether female voters feel espe-cially vulnerable to electoral violence. We already knowthat, across all African countries where Afrobarometer sur-veys have been conducted, women place high value on so-cial peace and political unity; they are more likely than mento express concern that multiparty competition will lead to‘‘conflict and confusion’’ (Logan and Bratton, 2006). To testwhether this tendency carries over to campaign violence, Iinserted an interaction term that represents the mediatingeffect of gender on intimidation’s impact on turnout (notshown). Although the relevant regression coefficientdisplayed the expected negative sign, it was not statisticallysignificant. Thus, while exposure to campaign violence isa deterrent to electoral participation, it is not a distinctivelygendered one.

6.2. Partisan choice

If violence undermines voter turnout, does it also com-pel partisan choice? In other words, do illicit campaignmethods affect not only whether people vote, but alsohow they vote? And, with reference to vote buying, canwe now discern a clear and substantial effect on voterbehavior?

14 The predicted probability is 0.5186 with a confidence interval rangingfrom 0.3667 to 0.6704. Thanks are due to Wonbin Cho and Tse-Hsin Chenfor assistance in calculating predicted probabilities.

15 Not shown. The raw correlations with voting in 2007 are r¼�0.152for exposure to a threat of violence in 2007 and r¼�0.240 for exposurein 2003. A caveat: the data for 2003 may have measurement problemsdue to recall bias.

M. Bratton / Electoral Studies 27 (2008) 621–632 627

A further distinction is necessary for this inquiry. Weneed to add information on the valence of campaign mal-practices. In other words, from where do illegal interven-tions originate: from incumbent or opposition parties? Iassume that both sides are implicated. Take vote buying:in total, the Afrobarometer recorded 485 cases of votebuying in 2007 as experienced by 296 respondents out ofa random sample of 2410 Nigerians. Because there weremore crimes than victims, it follows that many individuals(196, or about two-thirds of all victims) received more thanone offer from more than one partisan group. The largestpolitical parties were the most active in vote buying.According to the survey respondents, the ruling PDPmade 40% of all reported attempts to buy votes, followedby the leading opposition groups: the ANPP at 31% andthe Action Congress (AC) at 10%.

Table 4 displays factors that determine whether an indi-vidual chooses the incumbent party at the ballot box. Theobject to be explained is partisan choice, made operationalwith a dummy variable; it is scored as 1 if the individual,whether intending to vote or not, preferred a PDP candi-date in the April 2007 National Assembly election.

In this case, being rural or older – or especially if poor orfemale – had no significant impact on whether an individ-ual preferred the candidate of the incumbent party. Theonly significant demographic effect was education, whichincreased the odds of voting for the PDP.

Interestingly, the incumbent party was apparently muchmore successful than opposition parties in building supportamong eligible voters. It is unclear whether this advantagewas due to an explicit effort on the part of incumbents toconcentrate voter registration drives in their own electoralstrongholds. Alternatively, well-known candidates mayhave found it easier to generate popular enthusiasm forvoting than opposition candidates, some of whom werefresh faces newly arrived on the political scene. Whateverthe reason, the supporters of Nigerian opposition partieswere significantly less likely to plan to turn out on theday of the election.

But the issue under review is the effectiveness of votebuying and violence. As Table 4 shows, violence was coun-terproductive for the ruling party. People who felt threat-ened by political intimidation were consistently less likely

Table 4Determinants of choosing the incumbent party, Nigeria 2007

B S.E. Sig.

Constant �1.323 0.315 0.000

ControlsPoverty �0.005 0.053 0.925Education 0.053 0.023 0.019Rural 0.182 0.097 0.060Age �0.007 0.004 0.069Female �0.017 0.094 0.852Intended voter 0.476 0.102 0.000

Campaign malpracticesVote buyingoffer from incumbent party

1.633 0.203 0.000

Vote buyingoffer from opposition party

�0.063 0.199 0.754

Experience of threat of violence �0.559 0.155 0.000

Nagelkerke r2¼ 0.077 (n¼ 2265).

to vote for the PDP. We do not have the data to determinewhether particular incidents of intimidation originatedfrom incumbents or opposition. But I note the observationof a senior Nigerian political scientist that the PDPunleashed campaign repression in good part as a meansto control its own members.16 Since all political parties inNigeria have short histories and shallow institutional roots,there is reason to believe that party leaders might easily re-sort to heavy-handed measures as a means of overcomingweak party discipline, even in their own electoral strong-holds.17 But, contrary to leaders’ presumed intentions, vio-lence was associated with lower levels of electoral support.

Importantly, partisan choice is also strongly associatedwith vote buying, but this time positively. In Nigeria in2007, citizens who received a vote-buying offer from the in-cumbent party were significantly more likely to express anintention to vote for the PDP in the April elections.18 For anaverage Nigerian (say, a female rural dweller) and withother variables controlled at their mean level, vote buyingby an incumbent was linked to an increase in voting forthe ruling party by 38%.19 In other words, efforts by rulersto reward loyalists and attract others to the ruling partyfold apparently paid off. In this regard, we can interpretvote buying as one aspect of the larger phenomenon ofpatronage politics in which leaders exchange materialrewards in return for political allegiance. It is unclear, how-ever, whether campaign inducements were essential tocement the loyalty of citizens who already felt an affinitywith the PDP or whether these individuals would havevoted for the PDP anyway for other reasons, including eth-nic solidarity or policy performance.

Vote buying by incumbents was apparently more effec-tive than vote buying by opposition parties.20 It is true thata vote-buying bid by an opposition party is connected to anopposition vote when the latter is used as the dependentvariable (not shown).21 But in Table 4, the effect of opposi-tion vote buying looks weak. To be sure, the negative signon the relevant coefficient correctly predicts the reducedprobability that a citizen will vote for the incumbent. Butthe effect is not statistically significant. One obvious inter-pretation of this result is that office holders enjoy the ad-vantage of political incumbency. Rulers are able to makemore credible and binding offers to voters than the cash-starved opposition because they enjoy access to a largerpool of resources, including the public purse controlledby the state. The resource edge of incumbency is borneout by supporters of the ruling party, who are more likely

16 Statement by Professor John Ayaode at a workshop on ‘‘Institutionsfor Pro-Poor Growth’’ at the Centre for the Study of African Economies,Oxford University, June 2007.

17 My own observations in Zimbabwe confirm that the ruling ZANU-PFintimidates its own followers as well as those of the opposition MDC.

18 Some 62% of persons who received an offer from the incumbent party(versus 38% who received no offer) said they would vote for the PDP.

19 The predicted probability is 0.3819 with a confidence interval rangingfrom 0.3043 to 0.4596.

20 Only half (50% exactly) of those who received an offer from an oppo-sition party (versus the other half, who received no offer) said they wouldvote against the PDP.

21 The relevant statistics are B¼ 0.728, sig.¼ 0.000.

Table 5Popular reactions to vote buying, Nigeria 2007

You Other people

Comply 8 8Defect 42 28Refuse 41 24Other (inc. don’t know) 9 40

Cell entries are percentages of survey respondents, i.e., adult Nigerians(n¼ 2410).

M. Bratton / Electoral Studies 27 (2008) 621–632628

than opposition partisans to estimate the price of a vote at10,000 naira or more.22

7. Popular reactions: comply, defect, or refuse?

Politicians manipulate campaigns in order to maximizevotes. So far, I have shown that material inducements area more effective means to this end than political intimida-tion. Because threats of violence suppress voter turnout, in-timidation is not a useful campaign tactic except perhaps tocounteract an impending electoral loss. By contrast, al-though vote buying also suppresses turnout slightly, it ap-pears to boost partisan support, and therefore can beconsidered – morality and legality aside – as a campaigntactic that ‘‘works.’’

But these judgments make sense mainly from the prag-matic perspective of a politician who seeks to obtain orhang onto office. What courses of action are available tovoters in the face of vote buying and violence? Do specificforms of campaign manipulation invite different re-sponses? Under what circumstances do citizens comply,defect or refuse?

I concentrate the analysis on vote buying because fullersurvey data are available on this subject. The Afrobarome-ter asked, ‘‘what would you do if a candidate or party offi-cial offered you money for your vote in April 2007?’’ Wouldyou ‘‘take the money and vote for him/her’’ (that is, com-ply), ‘‘take the money and vote for the candidate of yourchoice’’ (that is, defect), or ‘‘refuse the money and vote forthe candidate of your choice’’ (that is, refuse)? Using thesame question with the same response categories, the sur-vey also asked respondents to judge the reactions of ‘‘otherpeople in your neighborhood or village.’’

The distribution of responses is shown in Table 5. Onlya small minority said that they would comply (8%) by tak-ing the money and then casting a ballot for the vote buyer’sparty. We can have confidence in the reliability of this esti-mate since respondents attributed exactly the same level ofcompliance to other people in their locality. Most people,however, said they would defect (42%) by taking the moneybut voting according to their conscience. A similar propor-tion said they would refuse from the outset to enter anyvote buying agreement (41%). One respondent elaboratedthat he would ‘‘drive that person away’’ and another saidthat she would ‘‘call the police.’’

Understandably, citizens were much less certain aboutresponses among the members of their residential commu-nity, with a plurality saying that they didn’t know whatother people would do. Some even admitted that theywere unsure of how they themselves would respond (6%).Finally, a small proportion volunteered that they would‘‘take the money and not vote at all.’’

As might be expected, popular reactions to vote buyingdepend in part on an individual’s socioeconomic status.Poor people are slightly more likely to comply and edu-cated people, if approached, are slightly more likely to re-fuse. But residential location again makes the largest

22 However, there is no significant difference between partisans of dif-ferent parties for lower price estimates.

difference, with rural dwellers being markedly more likelyto comply than urban dwellers (10 versus 6%). Concomi-tantly, rates of refusal are significantly higher in townsthan in the countryside (54 versus 43%).23 These data sug-gest that, if there is a culture of vote buying in Africa – inwhich votes are exchanged for campaign rewards – it ispredominantly (though not exclusively) a ruralphenomenon.

But the fact that fully 85% of urban dwellers wouldeither refuse a vote buying agreement or defect from itsuggests that, outside of pockets of reciprocity in the coun-tryside, vote buying may not be a very effective strategy foramassing electoral support.

As Table 6 shows, effectiveness of vote buying also de-pends in part on the source of the offer. If the PDP in Nigeriais at all representative, then incumbent political parties areapparently more effective at inducing voter compliance. Bya small but significant margin (15% versus 11%), the PDPwas more likely to get a Nigerian voter to say that he orshe would cast a ballot for a vote buyer. Again, the incum-bents’ ability to offer more attractive rewards than their op-ponents may be part of the explanation, though we shouldnot discount mass political loyalty to ruling parties, whichtend to have deeper social roots than insurgent opposi-tions. Moreover, while citizens are equally likely to refusevote-buying offers from incumbent and opposition alike,they are also somewhat more likely to defect from the op-position by promising support but then voting freely. Espe-cially where the opposition is unlikely to win, the voter hasless reason to fear that defection will result in subsequentretribution from powerful office holder.

I hypothesize that defection is most likely when would-be voters receive (and perhaps even accept) vote-buyingoffers from more than one party. Under these conditions,voters are faced with cross-pressures. They find themselvesin the uncomfortable position of being unable to simulta-neously comply with the preferences of both sides. Butthe data do not reveal this regularity. As Table 6 shows,voters are equally likely to defect whether they receiveoffers from one, or from more than one, political party(58%). Instead, voters tend to engage in other behaviors. Ei-ther they comply with the wishes of only one party (prob-ably the party they judge most likely to win), hoping thatthe other party will not be able to punish them. Or they re-fuse all offers, knowing that it is impossible to keep morethan one party happy at the same time.

But there is a third option. As some respondents told us, itis feasible for voters to take the money and not to vote at all.

23 Chi-square¼ 36.12, p¼ 0.000.

Table 6The source of vote buying by compliance and turnout, Nigeria 2007

Offer fromincumbentparty

Offer fromoppositionparty

Offer fromonly oneparty

Offer frommore thanone party

Comply 15 11 14 9Defect 56 59 58 58Refuse 29 30 29 34Intend to vote 58 49Intend to abstain 42 51

Cell entries are percentages of survey respondents who received vote-buying offers (n¼ 368).

M. Bratton / Electoral Studies 27 (2008) 621–632 629

The data reveal that this outcome is especially likely if votersaccept inducements from more than one party. We alreadyknow that entering a vote buying agreement has a suppres-sive effect on voter turnout. We now discover that this effectis particularly large when voters face cross-pressures fromcompeting vote buyers. Whereas, in January 2007, 66% ofNigerian survey respondents said they intended to vote inthe April elections, just 58% did so if they had receiveda vote-buying offer from a political party. But the intentionto vote falls even more precipitously – to less than half ofall eligible voters (49%) – when individuals entertain vote-buying offers from more than one political party.

In short, the Afrobarometer survey provides compellingevidence to the effect that, when citizens are caught in thecross-pressures of competitive vote-buying, their principalresponse is to abstain from voting at all.

24 This validity of the proxy is strengthened by the observation that in-dividuals who feel a need to ‘‘be careful about what you say’’ are also con-cerned about declining opportunities of ‘‘freedom to choose who to votefor without feeling pressured’’ (r¼ 0.179, p¼<0.001).

8. Correlates of defection

In a democracy, a responsible citizen would refuse to en-ter a vote-buying agreement. Most Nigerians acknowledgethis moral precept and many adopt it in practice. Citizencompliance with the wishes of vote buyers may be a pathof least resistance but it is ethically and legally fraught. Italso undermines the development of democratic citizen-ship. Perhaps the most rational response – though hardlythe most honest one – is defection, when citizens takeany money that may be on offer but vote as they wish any-way. Some civic educators even encourage this course ofaction (Schaffer, 2007, pp. 161–79). But its implicationsfor democratization are mixed: while citizens retain andexercise their right of free choice at the polls, they also im-plicate themselves in an electoral malpractice. One possiblesaving grace is that if enough citizens repeatedly defect,politicians will learn that vote buying does not work.

Because defection is the most ambiguous and interest-ing option – not to mention the most common one inNigeria – I conclude the discussion of vote buying by delv-ing into possible determinants. If, as shown above, theextent of cross-pressure does not drive defection, thenwhat does? I propose three hypotheses.

First, the prospect of defection from a vote buying agree-ment raises a collective action problem. In order to avoidrevealing that they have acted alone – thus exposing them-selves to punishment – citizens will seek strength in num-bers. They will only violate the agreement if they thinkothers will do so too. Hence, one would expect to find a pos-itive relationship between an individual’s own reaction to

a vote buying offer and his or her estimate of what ‘‘otherpeople in your neighborhood or village’’ would do underthe same circumstances. Specifically, they will defect onlyif they think others will defect too.

Second, it would seem logical that people are morelikely to defect if they think that the ballot is secret. If pol-iticians cannot discover how individuals or small groupsvoted, then the possible costs of defection are greatlyreduced. The Afrobarometer currently contains no directmeasure of whether voters regard the ballot as secret. Buta proxy measure can be constructed from a question abouthow often ‘‘people have to be careful of what they say aboutpolitics.’’ Those who say ‘‘often’’ or ‘‘always’’ are deemed toexpress political fear (66% in Nigeria); this group probablyalso worries that the ballot may not be secret.24

Third, citizens who are committed to democracy as theirpreferred political regime are unlikely to surrender theright to vote lightly. It is reasonable to expect that mostwill refuse offers to buy their votes. But, even if such per-sons succumb to the temptation of campaign inducements,they are still likely to want to make a free choice in the pri-vacy of the voting booth. I therefore propose that commit-ted democrats are more likely to defect than individualswho harbor nostalgia for authoritarian rule. Commitmentto democracy is measured in various Afrobarometer studiesby a standard index that combines an expressed preferencefor democracy with rejection of several alternative author-itarian regimes (Bratton et al., 2005). By this criterion,about half (49%) of adult Nigerians could be characterizedas ‘‘committed democrats’’ in 2007.

These correlates of defection from vote buying areentered, along with the usual controls, into the logistic re-gression model in Table 7.

Let us dispense quickly with the demographic controls.As expected, poverty, age and female gender are all signifi-cantly associated with a reduced likelihood that voters willdefect. To all appearances, older and poorer women are themost compliant constituency for vote buyers. But, notwith-standing what was said earlier about the sale of votes beinga predominantly rural phenomenon, defection rates aresome 7% higher among rural dwellers. Thus, politiciansprobably face meaningful obstacles when they try to mon-itor the behavior of the electorate in outlying areas, espe-cially if, as is common, their party organizations are weak.Nor do we find from the present African case any compen-satory support for the argument that ‘‘tightly knit, stable,and small communities in which everyone knows one an-other can generate good information about voter prefer-ences’’ (Lehoucq, 2007, p. 43; see also Stokes, 2007, p. 87).

But we wish to know how rural and other citizens mayarrive at a decision to defect from vote buying. Each of thehypotheses presented above contributes to an ecumenicalaccount. I will treat them in reverse order of importance.

To begin with, we can confirm the impact on electoral be-havior of a citizen’s personal commitment to democracy.

Table 7Determinants of defection from vote buying, Nigeria 2007

B S.E. Sig.

Constant �0.357 0.378 0.344

ControlsPoverty �0.146 0.061 0.017Education �0.028 0.026 0.291Rural 0.272 0.110 0.014Age �0.014 0.004 0.001Female �0.219 0.108 0.042

DeterminantsThink thatothers will defect

2.811 0.130 0.000

Express political fear �0.166 0.053 0.002Committed to democracy 0.235 0.099 0.017

Nagelkerke r2¼ 0.371 (n¼ 2166).

M. Bratton / Electoral Studies 27 (2008) 621–632630

A person who sees democracy as the best available regime andalso rejects military, one-party and one-man rule is unlikely tosell his or her vote lightly. Even if some of these committeddemocrats accept payments from vote buyers, they still reportthat they vote according to conscience. This explanation doesnot preclude, of course, that such voters might choose thevote buyer’s party; but they assert that the preferences under-lying this partisan choice are entirely their own.

Moreover, people who are fearful of expressing them-selves have a reduced propensity to defect. Especially ifvoters suspect that the ballot is not secret – an impressionthat unscrupulous politicians are in no hurry to dispel – de-fection will be seen as too risky. The biggest danger is thatvote buyers will discover that voters have not kept theirpart of the bargain, an outcome that invites retaliationand punishment. So it stands to reason that those whodoubt that freedoms of expression and voting will beprotected are unlikely to become defectors.

Finally, I can report a definitive result: defection fromvote buying depends on solving the collective action prob-lem. Voters defect if they have some assurance that othersin their locality will do so too. As revealed in Table 7, thestrongest and most significant relationship with defectionis voters’ own expectations that they are partaking in thecollective behavior of a larger group. For an average Niger-ian (say, a female rural dweller) and with other variablescontrolled at mean levels, the expectation that others willdefect is associated with an increase in one’s own odds ofdefecting by 59%.25 I do not know for sure whetherwould-be defectors have reliable information about theplanned voting behavior of their friends and neighbors.But such subjects are surely a topic of communal conversa-tion. As such, citizens can probably figure out whether theycan subsume their own behavior within that of a largergroup, and thus avoid being singled out for retribution. Un-der these circumstances, they are very much more likely totake the money and run.26

25 The predicted probability is 0.5922 with a confidence interval rangingfrom 0.5543 to 0.6300.

26 According to these data, the counter hypothesis that collective actionwould invite collective punishment (for example by the subsequent de-nial of development benefits to an entire village or neighborhood) isnot supported. But further research is called for.

9. Complementary or alternative strategies?

In this article, I have treated vote buying and political in-timidation as if these electoral strategies were separable.This discrete approach assumes that political campaignerssee such interventions as alternatives. Accordingly, politi-cians who lack sufficient resources to buy off voters willbe prone to resort to heavy-handed methods. Or if violencebackfires, then they will try to tempt voters with rewards. Athird possibility is that voters will experience promises ofreward from one side of the partisan divide and threats ofviolence from another.

In reality, of course, political campaigns reflect a multi-tude of strategies, often applied simultaneously. It is there-fore unexpected that, in Nigeria in 2007, overlap betweencampaign strategies was empirically quite limited. Amongadult Nigerians, just 3% reported being victimized by botha vote buying offer and a threat of negative consequencesduring the 2007 campaign. This of course exempts thosewho saw vote buying as an offer they couldn’t refuse be-cause it contained an implied threat of extortion. The re-mainder of the population was divided between the 11%who experienced just one of these malpractices and the86% who reported no direct personal experience of eitherone.

Table 8 teases out the links between, on one hand, a com-bination of vote buying and violence and, on the otherhand, Nigerian voting behavior. It displays a cross-tabula-tion of two ordinal scales: a scale of exposure (to none,one, or both violations) and a scale of intended compliance(running from refuse, through defect, to comply). It showsthat combining vote buying and violence had little appre-ciable association with the likelihood that a voter would re-fuse to enter an agreement (to vote a certain way) orcomply with such an agreement (if they did enter). In otherwords a threat of violence or an offer of a material induce-ment alone would be just as effective as a combination ofthese influences.

But a voter who experienced both vote buying and vio-lence was more likely to defect, either by not voting at all orby exercising a free vote choice. The probability of this con-nection rises by six percentage points when he or she issubjected to both treatments.27 In this regard, we confirmthat, from a politician’s perspective, employing a comple-mentary campaign strategy is self-defeating. There is noapparent advantage to be had in terms of controlling voterbehavior by supplementing vote buying with a threat of vi-olence, or supplementing political intimidation with mate-rial rewards. Instead, when faced with a combination ofboth irregularities, voters are significantly more likely tochoose to go it alone. The most extreme scenario involvesa voter who is induced to vote one way (say by an incum-bent party) but threatened to vote another way (say by anopposition group). According to Table 8, this harsh combi-nation of cross-pressures is also much more likely to belinked to defection than compliance.

27 The reader is cautioned that this result rests on a small number ofcases, just 183. Gamma¼ 0.212, p¼ 0.000.

Table 8Combined effects of vote buying and violence on voter behavior, Nigeria2007

Experienced NEITHERvote buying NORthreat of violence

Experienced EITHERvote buying ORthreat of violence

Experienced BOTHvote buying ANDthreat of violence

Refuse 51 30 29Defect 42 57 63Comply 8 13 9

Cell entries are percentages of survey respondents, i.e., adult Nigerians(excluding ‘‘don’t know’’/‘‘other’’) (n¼ 2252).

M. Bratton / Electoral Studies 27 (2008) 621–632 631

10. Conclusion and implications

This article has shown that vote buying and political in-timidation are characteristic dimensions of Nigerian elec-tion campaigns. According to survey-based estimates,almost one out of five Nigerians is personally exposed tovote buying and almost one in ten experiences threats ofelectoral violence. But when, as commonly happens, cam-paign irregularities are targeted at the rural poor, effectsare concentrated. These effects are as follows: violence re-duces turnout; and vote buying enhances partisan loyalty.But, perhaps because most citizens condemn campaignmanipulation as wrong, compliance with the wishes of pol-iticians is not assured. Defection from threats and agree-ments is more common than compliance, especiallywhere voters are cross-pressured from both sides of thepartisan divide.

That vote buying and violence affect relatively few peo-ple and rarely work well does not mean that these malprac-tices are without consequence. As others have noted, theintrusion of money and violence into election campaignsdamages the quality of democracy (Schedler, 2002;Schaffer, 2007). These transgressions undermine demo-cratic norms of political liberty (by depriving voters offree choice) and political equality (by benefiting the richat the expense of the poor). They diminish the legitimacyof electoral outcomes by giving ‘‘losers,’’ usually oppositionparties, reason to think that the vote was fraudulent. Evenwithout other methods of manipulation – such as ballotstuffing, ballot stealing and tampering with vote tallies –Nigeria’s disastrous April 2007 elections suffered preciselythis fate.

Negative consequences may be long lasting because de-fective election campaigns set the stage for governance bycorruption. As a defeated gubernatorial candidate said inNigeria: ‘‘anyone who is willing to steal a ballot box will(also) steal public money.’’28 Vote buying and violence en-able the elevation into elected office of cronies, criminalsand strong-arm ‘‘godfathers’’ who are singularly unfit forpublic service. Low caliber leaders rarely offer political rep-resentation to the marginalized majorities whose voicesare seldom heard in African politics. Instead, policy debateremains distorted: the views of the poor are muffled andthe preferences of the rich are amplified on key issues,such as the desired balance between taxation and services(Stokes, 2007; Acemoglu and Robinson, 2006). All told,

28 Kayode Fayemi, quoted in The New York Times, April 20, 2007.

irregular elections reduce the institutionalization of politi-cal accountability.

Campaign irregularities may also infect the quality ofdemocratic citizenship. It is encouraging that most Niger-ian citizens see vote buying and electoral violence aswrong. But morality, as well as behavior, may be commu-nally defined. We know that people who enter vote-buyingagreements are more likely to defect if they think otherswill do so too. The normative dimension of this solutionto the collective action problem, however, is that partici-pants in vote buying and violence are also more likely toregard these infractions as ‘‘wrong but understandable’’or ‘‘not wrong at all.’’ If participation in electoral malprac-tices reduces critical citizenship, it can hardly be healthyfor the development of democracy.

So what is to be done? Some requirements are contex-tual: it makes little sense to convene elections in an envi-ronment of political insecurity where armed factionsstand ready to intimidate opponents. Disarmament mustcome first. Other requirements are institutional: there areno substitutes for an independent electoral commission,an honest bureaucracy, and a neutral police force that canguarantee a secret ballot.

But electoral and other institutional reforms are aloneinsufficient. Also essential is a culture of democratic citi-zenship that begins with a citizenry ready to insist on cleanelections. In this regard, it is encouraging that so many Af-rican voters have moral objections to violence and votebuying. They already feel revulsion and indignity when po-litical entrepreneurs and thugs try to manipulate their pref-erences. Thus there is less need for voter education than fora system of incentives and capabilities that will allow ordi-nary people to resist these misguided efforts to appropriatetheir votes. In this regard, the attainment of a measure ofsocioeconomic development that reduces existing inequal-ities between political elites and ordinary citizens wouldsurely be one good place to start.

References

Acemoglu, Robinson, 2006. The Economic Origins of Dictatorship andDemocracy. Cambridge University Press, New York.

Alliance for Credible Elections and CLEEN, March 2007. Summary Reportof National Public Opinion Survey on Preparations for 2007 Electionsin Nigeria, mimeo.

Banegas, R., 1998. Marchandisation du vote, cioyennete et consolidationdemocratique au Benin. Politique Africaine 69 (1), 75–88.

Bratton, M., July 2007. Formal versus Informal Institutions in Africa. Jour-nal of Democracy 18 (3), 96–110.

Bratton, M., Lewis, P., February 2007. The durability of political goods? Ev-idence from Nigeria’s new democracy. Commonwealth and Compar-ative Politics 45 (1), 1–33.

Bratton, M., Mattes, R., Gyimah-Boadi, E., 2005. Public Opinion, Democracyand Market Reform in Africa. Cambridge University Press, New York.

Bratton, M., Chu, Y.-h., Lagos, M., 2007. Who Votes? Implications for NewDemocracies?. http://www.globalbarometer.net/publication.htm.

Brusco, V., Nazareno, M., Stokes, S., June 2004. Vote buying in Argentina.Latin American Research Review 39 (2), 65–88.

Chabal, P., Daloz, J.-P., 1999. Africa Works: Disorder as Political Instru-ment. James Currey, Oxford.

Downs, A., 1957. An Economic Theory of Democracy. Harper and Row,New York.

The Economist, April 21, 2007. How to steal yet another election. 51–52.Hicken, A., 2007. How do rules and institutions encourage vote buy-

ing? In: Schaffer (Ed.), Elections for Sale: the Causes and Conse-quences of Vote Buying. Lynne Rienner Publishers, Boulder,Colorado, pp. 47–60.

M. Bratton / Electoral Studies 27 (2008) 621–632632

Hirschmann, A., 1970. Exit, Voice and Loyalty: Responses to Decline inFirms, Organizations, and States. Harvard University Press, Cam-bridge, Massachusetts.

Human Rights Watch, 2007. Criminal Politics: Violence, ‘‘Godfathers’’ andCorruption in Nigeria. HRW Report No A1916, October, New York.

Ibrahim, J., January 2007. Nigeria’s 2007 Elections: the Fitful Path to Dem-ocratic Citizenship. U.S. Institute for Peace, Washington D.C. USIP Spe-cial Report No. 182.

Keefer, P., March 2004. Clientelism, Credibility and Democracy. The WorldBank, Development Research Group, Washington D.C.

Kitschelt, H., Wilkinson, S. (Eds.), 2007. Patrons, Clients and Policies: Pat-terns of Democratic Accountability and Political Competition. Cam-bridge University Press, New York.

Krishna, A., September 2005. Do poor people care less for democracy? Pa-per presented at the Annual Meeting of the American Political Sci-ence Association, Philadelphia.

LeBas, A., July 2006. Polarization as craft: explaining party formation andstate violence in Zimbabwe. Comparative Politics 38 (4).

Lehoucq, F., 2007. When does a market for votes emerge? In: Schaffer (Ed.),Elections for Sale: the Causes and Consequences of Vote Buying. LynneRienner Publishers, Boulder, Colorado, pp. 33–45.

Lindberg, S., Summer 2003. ‘It’s our time to chop’: do elections in Africafeed Neopatrimonialism rather than counteract it? Democratization10 (2), 121–140.

Logan, C., Bratton, M., 2006. The Political Gender Gap in Africa: Similar At-titudes, Different Behaviors. Afrobarometer Working Paper No. 58.www.afrobarometer.org.

Mohammed Salih, M.A., Nordlund, P., 2007. Political Parties in Africa:Challenges for Sustained Multiparty Democracy. International IDEA,Stockholm.

Mustapha, R., December 2006. Nigeria since 1999: revolving door syn-drome or the consolidation of democracy? Paper presented at a Con-ference on ‘‘Africa Since Democratization, University of Oxford.

Ramsayer, J.M., Rosenbluth, F., 1993. Japan’s Political Marketplace. Har-vard University Press, Cambridge, Massachusetts.

Rotberg, R., April 2007. Nigeria: Elections and Continuing Challenges.Council on Foreign Relations Special Report No. 27.

Scott, J., 1985. Weapons of the Weak: Everyday Forms of Peasant Resis-tance. Yale University Press, New Haven, Connecticut.

Schaffer, F. (Ed.), 2007. Elections for Sale: the Causes and Consequences ofVote Buying. Lynne Rienner Publishers, Boulder, Colorado.

Schedler, A., April 2002. Elections without democracy: the menu of ma-nipulation. Journal of Democracy 13 (2), 36–50.

Shugart, M., Nielsen, D., 1999. Constitutional change in Colombia: policyadjustment through institutional change. Comparative Political Stud-ies 32 (3), 313–341.

Sklar, R., Onwudiwe, E., Kew, D., July 2006. Nigeria: completing Obasan-jo’s legacy. Journal of Democracy 17 (3), 100–115.

Stokes, S., 2007. Is vote buying undemocratic? In: Schaffer (Ed.), Electionsfor Sale: the Causes and Consequences of Vote Buying. Lynne RiennerPublishers, Boulder, Colorado, pp. 81–99.

Transition Monitoring Group, 2007. An election programmed to fail. Pre-liminary Report on the Presidential and National Assembly Electionsheld on Saturday, April 21, 2007.

Vicente, P., March 2007. Is Vote Buying Effective? Evidence from a Ran-domized Experiment in West Africa. Center for the Study of AfricanEconomies, Oxford University.

van de Walle, N., June 2003. Presidentialism and clientelism in Africa’semerging party systems. Journal of Modern African Studies 41 (2),297–319.

van de Walle,N.,2007. Meet thenew boss, same astheold boss? The evolutionof political clientelism in Africa. In: Kitschelt, Wilkinson (Eds.), Patrons,Clients and Policies: Patterns of Democratic Accountability and PoliticalCompetition. Cambridge University Press, New York, pp. 50–67.

Wang, C.-S., Kurzman, C., 2007. The logistics: how to buy votes. In: Schaffer(Ed.), Elections for Sale: the Causes and Consequences of Vote Buying.Lynne Rienner Publishers, Boulder, Colorado, pp. 61–79.

Wilkinson, S., 2004. Votes and Violence: Electoral Competition and EthnicRiots in India. Cambridge University Press, New York.

Yadav, Y., 2000. Understanding the Second Democratic Upsurge: Trendsof Bahujan Participation in Electoral Politics in the 1990s. In:Frankel, F., Hasa, Z., Bhargava, R., Arora, B. (Eds.), Transforming India:Social and Political Dynamics of Democracy. Oxford University Press,Oxford.