27
http://cps.sagepub.com/ Comparative Political Studies http://cps.sagepub.com/content/40/6/665 The online version of this article can be found at: DOI: 10.1177/0010414006288969 2007 40: 665 Comparative Political Studies Michelle Kuenzi and Gina M. S. Lambright Voter Turnout in Africa's Multiparty Regimes Published by: http://www.sagepublications.com can be found at: Comparative Political Studies Additional services and information for http://cps.sagepub.com/cgi/alerts Email Alerts: http://cps.sagepub.com/subscriptions Subscriptions: http://www.sagepub.com/journalsReprints.nav Reprints: http://www.sagepub.com/journalsPermissions.nav Permissions: http://cps.sagepub.com/content/40/6/665.refs.html Citations: What is This? - May 9, 2007 Version of Record >> at NATIONAL CHUNG HSING UNIV on March 28, 2014 cps.sagepub.com Downloaded from at NATIONAL CHUNG HSING UNIV on March 28, 2014 cps.sagepub.com Downloaded from

Voter Turnout in Africa's Multiparty Regimes

  • Upload
    g-m-s

  • View
    214

  • Download
    0

Embed Size (px)

Citation preview

http://cps.sagepub.com/Comparative Political Studies

http://cps.sagepub.com/content/40/6/665The online version of this article can be found at:

 DOI: 10.1177/0010414006288969

2007 40: 665Comparative Political StudiesMichelle Kuenzi and Gina M. S. Lambright

Voter Turnout in Africa's Multiparty Regimes  

Published by:

http://www.sagepublications.com

can be found at:Comparative Political StudiesAdditional services and information for    

  http://cps.sagepub.com/cgi/alertsEmail Alerts:

 

http://cps.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

http://cps.sagepub.com/content/40/6/665.refs.htmlCitations:  

What is This? 

- May 9, 2007Version of Record >>

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Voter Turnout in Africa’sMultiparty RegimesMichelle KuenziUniversity of Nevada, Las VegasGina M. S. LambrightGeorge Washington University, Washington, D.C.

Many people consider voting the most important form of political participationin a democracy. To this point, no aggregate-level, cross-national studies havebeen done on the factors that affect voter turnout in Africa. This article seeks tofill the gap in the literature by examining the factors that influence voter turnoutin sub-Saharan Africa’s multiparty regimes that have had two consecutive elec-tions since the democratic transitions in the 1990s. Many of the central findingsof the research on voter turnout in other regions reappear in the examination ofvoter turnout in Africa. The authors find that two institutional variables—typeof electoral formula and concurrency of presidential and legislative elections—have significant effects on electoral turnout. Media exposure has a significantpositive relationship with voter turnout. The number of elections a polity hashad also appears to affect levels of voter turnout.

Keywords: voter turnout; democracy; Africa; elections; political institutions

It is through voting and the notion of “one person, one vote” that the idealof political equality is truly manifested in democracies. Lijphart’s (1997)

work reveals that low voter turnout often translates into “unequal participa-tion,” in which participation is systematically biased in favor of citizens ofhigher socioeconomic status. Because democracy in its most fundamentalsense is “rule by the people,” the proportion of the citizenry expressingits preferences through voting is of interest. Elections are at the core ofmodern democracy, and low voter turnout rates might indicate that peopledo not see elections as central to political life. Although some people mayquestion the meaningfulness of voting in the African context, Jenkins andKposowa (1992) found that high levels of voter turnout in the preindependence

Comparative Political StudiesVolume 40 Number 6

June 2007 665-690© 2007 Sage Publications

10.1177/0010414006288969http://cps.sagepub.com

hosted athttp://online.sagepub.com

665

Authors’ Note: We thank David Damore, Ted G. Jelen, Mark Jones, and the anonymous refereesfor their helpful comments and suggestions.

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

elections helped stave off coup activity in the postindependence period.Because voting appears to be important in Africa, just as it is elsewhere in theworld, understanding what factors affect voter turnout in Africa is critical.

Although some recent studies on voter turnout have focused on differentregions of the world in which emerging democracies are concentrated, noone has systematically studied the factors that influence voter turnout inAfrica at the cross-national level.1 An emerging literature examines votingbehavior in Africa using individual-level data (see, e.g., Bratton, Mattes, &Gyimah-Boadi, 2005; Kuenzi & Lambright, 2005; Posner & Simon, 2002;Youde, 2005). Some of these studies use survey data to explore the deter-minants of voting. Others focus on the determinants of political support forthe government or ruling party. This study differs from the existing litera-ture on voting in Africa in a couple of ways. First, we examine aggregate-level turnout and seek to understand cross-national differences in levels ofturnout. Second, unlike the economic voting literature, this article seeks toexplain voter turnout as opposed to voter choice. What factors affect thelevel of voter turnout in Africa’s electoral regimes?

Like Fornos, Power, and Garand (2004) in their study of turnout in LatinAmerican countries and Kostadinova (2003) in her study of voter turnout inpostcommunist countries, we employ an “eclectic approach” in attemptingto answer this question and look at the influence on turnout of variablesassociated with several theoretical approaches. We look at the effects ofmany of the institutional, socioeconomic, and contextual factors that havebeen identified as important in the literature. We find that two institutionalvariables—type of electoral formula and concurrency of presidential andlegislative elections—have significant effects on voter turnout. Media expo-sure, measured as the number of radios per person, consistently has a strong,significant positive relationship with voter turnout. The number of multi-party elections a polity has had also appears to affect levels of voter turnout.

Examining this subject in the context of Africa is important for theoret-ical and practical reasons. Testing the findings generated from studies of theadvanced industrial democracies with data from African polities allows usto assess the external validity of these findings. Seeing what factors willaffect voter turnout in Africa’s democracies is important because of theirfragility. Although many people lament the low turnout rates in the UnitedStates, the United States is a long-standing democracy with a very low riskof democratic reversal. In contrast, political institutions are in flux in manyAfrican countries, and thus if certain institutional arrangements are likelyto enhance political integration, such institutions could potentially beadopted.

666 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Theory

In his pioneering study of 30 democracies, Powell (1980) finds compul-sory voting laws and automatic registration to be the institutional factorsexercising the most influence on voter turnout. Jackman (1987) finds strongsupport for the institutional perspective. According to Jackman, institu-tional arrangements that lead voters to conclude that their votes matterencourage participation. Based on his analysis of legislative turnout data for19 advanced industrial democracies for the 1960s and 1970s, Jackman(1987) finds that unicameralism and nationally competitive districts areassociated with higher turnout, whereas electoral disproportionality andmultipartism suppress turnout. Studies of the emerging democracies sup-port the contention that institutions influence voter turnout. In their studyof voter turnout in the legislative elections that took place in Latin Americafrom 1980 to 2000, Fornos et al. (2004) find that unicameralism, the con-currency of legislative and presidential elections, and compulsory votinglaws all have significant effects on turnout.

Bratton and van de Walle (1997) observe that the “nature of politicalauthority” and political institutions differ in Africa from other areas of theworld. They note that neopatrimonialism was the defining characteristic ofthe pretransition regimes.2 Neopatrimonialism has survived democracy, and“big man democracy,” in which the chief executive towers over the politicalsystem, has emerged in many African countries (see also Schatzberg, 2001,for a description of “big men” and patrimonial systems in Africa). Given theneopatrimonial nature of the regimes under study, will voter turnout beaffected in predictable ways by formal institutional arrangements?

We hypothesize that institutions affect political behavior in Africa inmuch the same way they affect behavior in other parts of the world. Forexample, because the executive is indirectly derived from the vote for thelower house in parliamentary systems, we expect there to be a negative rela-tionship between turnout in legislative elections and presidentialism.Similarly, potential voters will likely perceive presidential elections to bemore significant than legislative elections (e.g., van de Walle, 2003) and bemore likely to vote in presidential elections. Holding presidential and leg-islative elections concurrently is therefore likely to boost turnout in legisla-tive elections. We also expect elections in bicameral systems to elicit lowerturnout rates than those in unicameral systems, because bicameralismdisperses legislative power.

The expectations in the literature regarding the effect of multipartism onvoter turnout are contradictory. Institutionalists, such as Jackman (1987),

Kuenzi, Lambright / Voter Turnout in Africa 667

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

have argued that disproportionality has a negative association with turnoutbecause it decreases the effect each vote has on the electoral outcome. Ofcourse, proportionality is positively associated with multipartism. However,Jackman argues that multipartism has a negative relationship with turnout.Scholars argue that multipartism decreases the likelihood that one will votefor a number of reasons. For example, the larger the number of parties con-testing an election, the greater the information costs imposed on the poten-tial voter. Also, the existence of multiple legislative parties decreases theeffect a person’s vote will have on governmental outcomes. Siaroff andMerer (2002) find a positive relationship between “polarized two-party sys-tems” and turnout in their study of legislative elections in Europe since1990. Similarly, Kostadinova (2003) finds a negative relationship betweenturnout and multipartism in postcommunist Europe.

Other scholars claim that multipartism is likely to increase electoralturnout. In a multiparty system, parties are apt to focus on a limited numberof issues or perhaps just one issue and are thus likely to draw single-issuevoters to the polling stations. Brockington (2004) has helped reconcile thetwo perspectives. In his study of the 15 advanced industrial democracies,Brockington finds that once one controls for the rate of non-minimal win-ning coalitions, multipartism appears to stimulate electoral turnout.

Political parties may serve as mobilization agents in the more democra-tic and in less democratic multiparty systems of Africa. For example, inSenegal, the members of entire villages are sometimes mobilized to vote ina particular way. One payoff from voting may be the social approbation onereceives for demonstrating loyalty to village notables or family members(see Schaffer, 1998). One may also receive more tangible payoffs such as abag of rice or a small sum of money. Although political parties themselvesare weak in Africa (Bratton & van de Walle, 1997; Kuenzi & Lambright,2001), they may still be effective mobilization agents. In their study of 12African countries, Norris and Mattes (2003) find that ethnicity influencesparty support. Thus, multipartism probably reflects a situation in whichmore identity groups are represented. People are more likely to vote whenthere are parties to which they feel directly connected than when such is notthe case. This expectation is consistent with Powell’s (1980) finding thatparty linkages with specific cleavage groups boosted turnout. Therefore, weexpect multipartism to have a positive effect on turnout in Africa.

Reflecting the concerns of the modernization theorists, many peoplehave claimed that socioeconomic variables influence turnout rates. At theindividual level, several U.S. studies have connected socioeconomic vari-ables, such as income and education, to voting (e.g., see Teixeira, 1987;

668 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Verba, Schlozman, & Brady, 1995; Wolfinger & Rosenstone, 1980). In theirstudy of 324 legislative elections conducted in 91 electoral democraciesfrom 1972 to 1995, Blais and Dobrzynska (1998) find that the socioeco-nomic environment has a strong effect on voter turnout. In contrast, Fornoset al. (2004) did not find any of the socioeconomic variables to be signifi-cantly related to turnout in Latin America.

At a more macro level, economic performance and voter turnout havebeen linked in the literature. Radcliff (1992) finds that there is a positiverelationship between economic performance and turnout in the advancedindustrial democracies. Kostadinova (2003) expects higher turnout in thecase of poor economic performance in transitional countries. AlthoughRadcliff does find this to be the case when he looks only at his sample ofdeveloping countries, Kostadinova does not find a relationship between eco-nomic performance and turnout in the countries of postcommunist Europe.

Posner and Simon (2002) report that declining economic conditionsreduce public support for incumbent governments in Africa, but public dis-satisfaction with the government is most often expressed through absten-tion rather than a vote for the opposition. We might therefore expect thatstagnating or negative growth rates translate into lower turnout in Africa.

Another factor of interest to modernization theorists was media expo-sure (e.g., Lipset, 1959). Exposure to media is thought to increase individ-uals’ desire and ability to vote. We expect level of media exposure to havea positive relationship with turnout in Africa.

Others point to contextual variables that are likely to affect voter turnout.For example, Fornos et al. (2004) find that founding elections in LatinAmerican countries are likely to produce higher turnout rates than thoseelections that follow. In contrast, people are also likely to acquire the habitof voting as they gain more and more experience with democracy. Rustow(1970) contends that democratization includes a “habituation” phase inwhich citizens and politicians begin to acquire the habits and practices ofdemocracy. Gerber, Green, and Shachar (2003) claim the results of theirU.S. field experiment regarding 25,200 registered voters support the notionthat “Voting and abstention. . . are habit-forming” (p. 540). Thus, wemight also expect a positive relationship between the number of elections apolity has had and voter turnout.

The manner in which previous elections were conducted and the level ofpolitical rights and civil liberties available to citizens are likely to influencevoter turnout. Where governments restrict citizens’ civil liberties and polit-ical rights, elections are unlikely to be truly free and fair or competitive. Insuch environments, citizens may be less likely to expend the resources

Kuenzi, Lambright / Voter Turnout in Africa 669

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

required to vote. Fornos et al. (2004) find that level of democracy has asignificant positive impact on voter turnout in Latin America. We expect tofind a similar relationship in Africa.

Research Method

Given the small number of cases available for a study of this nature andthe fact that many of our variables do not change across time, we chose torun cross-sectional analysis for two sets of elections—the most recent elec-tions (n = 32) and the penultimate elections (n = 32)—rather than conductpooled cross-sectional analysis.3 Our unit of analysis is the legislative elec-tion. Table 1 lists the countries and the years of the legislative electionsincluded in this study. Our data set includes the universe of sub-SaharanAfrican (SSA) multiparty electoral regimes for which turnout data areavailable for two legislative elections that have taken place since the demo-cratic transitions in SSA in the early 1990s. For each country, we includedata on the most recent and penultimate election.4 We also include two pairsof elections for Gambia: (a) the last two legislative elections held before thecoup in 1994 and (b) the two elections held since the resumption of multi-party elections in 1997.5

Like Jackman (1987) and others, we are particularly interested in thefactors associated with voting in legislative elections. Although the politi-cal systems in a majority of African countries are presidential or semipres-idential and historically power has been centralized in the office of thepresidency, one key feature of the constitutional reforms that characterizeddemocratic transitions in Africa was an effort to strengthen parliaments andother political institutions outside the executive branch (Bratton & van deWalle, 1997). Thus, identifying the factors associated with legislativeturnout is important.

Our cases include elections in Africa’s long-standing democracies (e.g.,Botswana and Mauritius) and elections in countries that are best characterizedas “hybrid regimes” (Diamond, 2002). We chose to include elections from allof the countries with multiparty electoral regimes in our study because we areinterested in exploring how the level of democracy affects turnout.

We focus on explaining turnout for Africa’s most recent legislative elec-tions but conduct analysis based on data from the penultimate elections asa check, because the dynamics of voter turnout may change with time aselections become more routinized and citizens become more accustomed tothe act of voting. We chose to look at turnout in the different sets of elections

670 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Kuenzi, Lambright / Voter Turnout in Africa 671

Table 1Voter Turnout in Africa

Official Turnout Official Turnout (Percentage of (Percentage of

Election Year Registered Voters) Voting-Age Population)

Benin 1999 70.1 65.92003 55.9

Botswana 1999 77.1 42.02004 76.2

Burkina Faso 1997 44.5 41.32002 64.1

Cameroon 1992 60.61997 75.6 46.7

Cape Verde 1995 76.5 79.5b

2001 54.1 65.9b

Central African 1993 67.9 50.3Republic 1998 58.5 53.6

Chad 1997 50.1 25.62002 52.8

Côte d’Ivoirea 1990 42.5 39.92000 31.5 23.8

Djibouti 1997 56.8 29.82003 48.4

Gambia I 1987 80.2 53.71992 55.8 55.3

Gambia II 1997 73.2 56.72002 56.4

Ghana 1996 65.0 68.0b

2000 61.2 64.6b

Guinea-Bissau 1994 45.0 32.11999 80.0 129.1b

Kenya 1997 65.5 46.72002 57.0

Lesotho 1998 71.8 61.72002 68.1

Madagascar 1998 60.1 40.62002 67.9

Malawi 1994 80.0 67.91999 92.3 105.9b

Mali 1992 21.1 21.9b

1997 21.6 21.5Mauritania 1996 52.1 47.7

2001 54.5

(continued)

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

separately rather than using an average of turnout across elections, becauseaveraging turnout across elections limits our ability to explore the dynamicnature of turnout across time. Time-series analysis would enable us to gofurther in understanding changes in turnout across time, but we are unableto employ time-series methods because most of the countries in SSA havehad only two or three multiparty legislative elections since their transitionsto democracy in the early 1990s. We took appropriate precautions to ensurethat our results are not driven by a few influential cases. The robustness andstability of the results are checked in numerous ways.

672 Comparative Political Studies

Table 1 (continued)

Official Turnout Official Turnout (Percentage of (Percentage of

Election Year Registered Voters) Voting-Age Population)

Mauritius 1995 76.7 77.12000 80.9 79.6

Mozambique 1994 87.9 66.41999 68.1 58.2

Namibia 1994 76.1 63.81999 62.8 61.7

Nigera 1995 35.0 37.2b

1999 99.4 96.2Nigeria 1999 84.8 93.1b

2003 50.0Saõ Tomé 1998 64.7 50.5

2002 67.0Senegal 1998 39.3 27.5

2001 67.3 40.6Seychelles 1998 86.7 90.2b

2002 87.0South Africa 1999 89.3 63.9

2004 76.7Tanzania 1995 76.5 47.9

2000 72.8 45.7Togoa 1994 65.1 69.1b

2002 67.4Zambia 1996 78.5 39.8

2001 68.5Zimbabwe 1995 30.8 26.0

2000 57.0

a. Indicates elections that are not sequential.b. Indicates that turnout as a percentage of voting-age population is greater than turnout as apercentage of registered voters.

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Operationalization

Dependent Variable: Voter Turnout

We use the percentage of registered voters that voted in legislative elec-tions for the lower house of parliament to measure voter turnout.6 The mea-sure is calculated using the total number of votes, which includes validballots, invalid ballots, and blank ballots.7 Like Powell (1980), we feel itmakes sense to include all votes, given that there are different ways a bal-lot’s validity is determined across countries and the fact that turning out tovote—whether the ballot is judged valid or not—is an expression of politi-cal participation. Although past studies have largely measured turnout as thepercentage of the voting-age population (VAP) that turned out to vote, welook at the percentage of the registered voters who turned out to vote for afew reasons. First, and most important, data on turnout as a percentage of theVAP were available for only 15 of the 32 most recent legislative elections(see Table 1). Moreover, there are problems with the data on VAP in severalcases, such as Guinea Bissau, where the reported VAP falls far below thenumber of reported votes.8 For 10 countries, the reported turnout as a per-centage of the VAP is actually higher than turnout as a percentage of regis-tered voters (see Table 1). This is not surprising, given the problems inherentin estimating the size of the VAP from census data (see Radcliff, 1996).9

Second, using turnout as a percentage of registered voters allows us toassess those factors that drive people to vote once they are registered. Thus,our analysis bypasses the issue of how difficult it is to register to vote andfocuses on the other factors that affect turnout. Despite the differencesnoted between the two measures of turnout and the limited number of datapoints we have for turnout as a percentage of VAP, the two measures arehighly correlated.10

Political Institutions

We are interested in the relationship between political institutions andturnout in Africa and therefore include a number of institutional variablesin our analyses: concurrency of presidential and legislative elections, elec-toral formula, bicameralism, presidentialism, and multipartism. (None ofthe African countries covered in our study have compulsory voting laws,and therefore we cannot include this variable in our analysis.) We use adummy variable to measure the concurrency of presidential and legislativeelections. Countries in which legislative and presidential elections are held on

Kuenzi, Lambright / Voter Turnout in Africa 673

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

the same day were coded as 1. Elections held in countries with parliamentarysystems (Mauritius, Botswana, Lesotho, Ethiopia, and South Africa) werealso coded as 1, because the executive is derived from the Parliament.Countries in which presidential and legislative elections are held at differ-ent times were coded as 0.

We also created a dummy variable to distinguish countries with majori-tarian electoral formulas from those with more proportional electoral rules.Elections held using plurality or majority electoral formulas were coded as1, whereas those conducted using some form of proportional representation(PR) or mixed electoral formula were coded as 0. Norris (2002) finds that“the key distinction is between PR systems and all others.” The way weoperationalize this variable is largely consistent with Norris. There are onlyfive mixed systems among our cases, and nearly all of them have very pro-portional elements. Thus, we distinguish the clearly majoritarian formulasfrom the more proportional formulas.

To measure bicameralism, we created a third dummy variable. Countrieswith a bicameral legislature were coded as 1, whereas those with a uni-cameral legislature were coded as 0.11 Additionally, we created a dummyvariable to differentiate countries with a presidential system from thosewith a parliamentary system. Countries with a parliamentary system werecoded as 0, whereas countries with an independently elected executive werecoded as 1.

To examine the impact of multipartism on voter turnout, we use Laaksoand Taagepera’s index of the effective number of parties to calculate theeffective number of legislative parties (ENLP) for each election. The for-mula calculates the effective number of parties by dividing one by the sumof the squared seat shares for each party.12 We use ENLP rather than theeffective number of electoral parties, because data on vote shares of all ofthe parties that participated in an election were unavailable for many of theelections of study.

Contextual Variables

We use Freedom House (FH) total scores for political rights and civilliberties to measure the level of democracy in the period directly precedingthe legislative elections. FH scores include many different indicators thatdeal directly with the conduct of elections, such as whether elections arecharacterized by “fair polling and honest tabulation of ballots.”13 Thus,these scores capture the most important features of the electoral processthat may influence voters’ decisions to participate. We use the total FH

674 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

scores from the year before the election. We use combined polity scoresfrom the year prior to the legislative election to cross-check the results.14

We examine the effect of executive dominance on legislative turnout byincluding a measure to capture the concentration of power in the execu-tive—the number of years the president has held office at the time of thelegislative election.

We are also interested in the effects of electoral experience on turnout.Our measure of electoral experience is the number of elections held in eachcountry since the reintroduction of multiparty elections up to the electionincluded in our study. For the few countries that have held multiparty elec-tions without interruption since independence, our measure is the numberof elections held in each country since independence up to the electionincluded in our study. We also tested whether electoral experience underauthoritarian rule influences turnout in current elections.15 (See Model 8 inTable 2).

Finally, we include several measures to capture important elements ofthe social and economic context. We operationalize urbanization as the per-centage of the population living in urban areas.16 In addition, we includemeasures of per capita income and growth in per capita income to captureeconomic performance within each country.17 Both measures are laggedand measured for the year preceding the election. We chose not to use lit-eracy as a measure of development or knowledge, because it is highly cor-related with other variables of interest.18 Furthermore, it is not significantlycorrelated with electoral turnout in Africa.

Media exposure is measured by the number of radios per capita in acountry, which we believe is the best measure of media exposure for SSA.19

As Bratton and van de Walle (1997) note, literacy rates are not high in manyAfrican countries, and newspapers are relatively expensive. In addition,radios are much more widely used and accessible than televisions.

Descriptives

Electoral turnout in Africa resembles turnout in other regions that haverecently undergone transitions to democracy. Mean turnout for legislativeelections in 17 Latin American countries from 1980 to 2000 ranges from alow of 35.8 in Colombia to a high of 95.7 in Uruguay (Fornos et al., 2004,p. 935). Average turnout is slightly higher in postcommunist Europe whereKostandinova (2003) reports a mean turnout rate of 73% during the 1990sand a range of 43.2 to 96.8 (p. 749). On average, turnout looks similar for

Kuenzi, Lambright / Voter Turnout in Africa 675

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Tabl

e 2

Ord

inar

y L

east

Squ

ares

Est

imat

es o

f E

lect

oral

Tur

nout

Mod

el 2

M

odel

4

Mod

el 5

M

odel

6

Mod

el 8

(E

xclu

des

Mod

el 3

(E

xclu

des

(“E

lect

oral

(I

nclu

des

Eig

ht(I

nclu

des

the

GD

P an

d(P

enul

timat

e “N

ot F

ree”

Dem

ocra

cies

”A

dditi

onal

N

umbe

r of

Ele

ctio

ns

Mod

el V

aria

ble

Mod

el 1

SEPr

esid

entia

lism

)SE

Ele

ctio

ns)

SEC

ases

)SE

Onl

y)SE

Cas

es)

SEM

odel

7SE

thro

ugh

1989

)SE

Con

stan

t21

.01

16.9

726

.04*

13.4

553

.91*

*22

.13

30.0

919

.84

16.8

413

.88

33.7

2**

13.9

548

.68

25.1

138

.91*

*17

.79

Whe

ther

ele

ctio

ns

16.0

7***

5.14

15.1

3***

3.75

15.4

5*7.

9616

.00*

**4.

0318

.54*

**4.

9313

.20*

**3.

9718

.12*

8.38

15.1

0***

4.40

held

con

curr

ently

(1

=Ye

s)W

heth

er e

lect

oral

−1

6.50

***

5.84

−15.

89**

*5.

26.5

677.

55−1

0.35

*5.

71−1

3.33

**4.

71−1

4.17

***

4.24

−35.

49**

*7.

69−1

3.46

**5.

18ru

les

are

maj

orita

rian

(1

=Ye

s)W

heth

er le

gisl

atur

e −2

.23

3.45

−3.0

63.

122.

057.

91−1

.72

3.34

−7.3

55.

92−1

.74

3.72

−23.

06**

7.93

−1.7

54.

63is

bic

amer

al

(1 =

Yes)

Whe

ther

pol

itica

l 2.

548.

82sy

stem

is

pres

iden

tial

(1 =

Yes)

Num

ber

of y

ears

−.

127

.401

−.12

0.3

37−.

214

.374

.096

.330

−.87

3*.4

37−.

198

.292

.193

.484

−.08

9.3

54cu

rren

t pre

side

nt

in o

ffic

ePe

rcen

tage

of

−.13

8.1

72−.

161

.165

.098

.226

−.15

5.1

72−.

224

.208

−.19

2.1

24−1

.27*

**.2

69−.

128

.172

popu

latio

n liv

ing

in u

rban

are

asM

edia

exp

osur

e.0

88**

*.0

27.0

81**

*.0

19.0

65**

.028

.077

***

.022

.108

***

.027

.072

***

.016

.161

**.0

58.0

79**

*.0

23G

DP

per

capi

ta

−.00

1.0

02(l

agge

d)

676

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Ann

ual p

erce

ntag

e .0

50.3

28.0

50.3

07−.

334

.481

−.12

1.1

97−.

116

.240

.031

.215

−.62

7.3

50.0

52.2

75ch

ange

in G

DP

(lag

ged)

Eff

ectiv

e nu

mbe

r 2.

852.

912.

952.

70.8

612.

492.

392.

765.

87*

2.94

3.75

*2.

0122

.88*

**4.

262.

832.

48of

legi

slat

ive

part

ies

Free

dom

Hou

se

2.48

**1.

122.

48**

1.08

−.61

31.

291.

741.

683.

63**

1.29

2.62

***

.915

−1.9

21.

041.

791.

22sc

ores

(la

gged

)N

umbe

r of

mul

tipar

ty

3.95

**1.

803.

47**

1.61

−1.7

52.

572.

701.

793.

182.

521.

781.

802.

812.

00le

gisl

ativ

e el

ectio

nsN

umbe

r of

ele

ctio

ns

.289

3.98

sinc

e in

depe

nden

ce

thro

ugh

1989

Num

ber

of e

lect

ions

3232

3226

2040

1532

R2

.641

0.6

378

.463

9.6

916

.819

7.6

057

.963

7.5

515

SEE

12.0

911

.55

15.8

910

.97

10.4

012

.58

10.4

812

.86

F s

tatis

tic12

.27

7.51

3.50

4.33

7.95

6.14

3.67

Sign

ific

ance

of

.000

0.0

001

.007

5.0

056

.002

3.0

001

.005

8F

stat

istic

Not

e:D

epen

dent

var

iabl

e is

vot

er tu

rnou

t as

a pe

rcen

tage

of

regi

ster

ed v

oter

s in

legi

slat

ive

elec

tions

,exc

ept i

n M

odel

7,i

n w

hich

the

depe

nden

t var

iabl

e is

turn

out a

s a

perc

enta

ge o

f th

e vo

ting-

age

popu

latio

n. R

obus

t sta

ndar

d er

rors

(SE

) ar

e pr

ovid

ed in

the

colu

mn

to th

e ri

ght o

f th

e co

effi

cien

ts.

*p≤

.10.

**p

≤.0

5. *

**p

≤.0

1 fo

r tw

o-ta

iled

test

s.

677

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

the most recent elections and the penultimate elections in Africa’s multi-party regimes. The mean voter turnout for the penultimate elections is64.1% compared to 63.9% for the most recent elections. Yet the correlationbetween our measures of turnout for the last and penultimate electionsis smaller than expected (r =.36, p ≤ .05). Within the set of penultimateelections, there are 12 founding elections, such as the 1995 election inTanzania. The two sets of elections exhibit similar levels of variation inturnout rates. The standard deviation for the most recent legislative elec-tions is 15.8, whereas that for the previous elections is 17.9. Voter turnoutranges from a low of 21.6% in Mali’s 2002 election to a high of 99.4% inNiger’s 1999 election.20

Results

Table 2 reports the results of our analysis using ordinary least squaresregression to estimate the effects of institutional, socioeconomic, and contex-tual variables on voter turnout in Africa’s legislative elections. Model 1 includesnearly all of the variables identified in the literature as factors likely to influ-ence voter turnout. The model includes five institutional variables: presiden-tialism, ENLP, concurrency of presidential and legislative elections, electoralformula, and bicameralism. The model also includes three measures of thesocioeconomic context: urban population, GDP per capita, and annual growthrates, and a measure of media exposure (number of radios per person). Weinclude the measure of electoral experience and the measure of the level ofdemocracy.

Like Jackman (1987), Jackman and Miller (1995), and Fornos et al.(2004), we find that institutions have a significant effect on voter turnout.Concurrency appears to exercise a strong effect on turnout in Africa. Basedon the unstandardized regression coefficient reported in Table 1, we can seethat predicted turnout increases by around 16% when presidential and leg-islative elections are held concurrently if all other variables remain con-stant. This is not surprising, given the costs associated with voting,especially in rural Africa, where voters may have to travel some distance toreach a polling station. Prior to Liberia’s most recent elections, one officialobserved that many eligible voters in the River Gee County would have towalk for 5 to 7 hours to reach the nearest polling station.21

Our findings also support Jackman’s contention that the electoral for-mula matters. We find that the relationship between electoral formula andturnout in Africa more closely resembles what Jackman (1987) reports forthe advanced industrial democracies than what Fornos et al. (2004) report

678 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

for Latin America. In Africa, voter turnout is consistently lower in electionsconducted using majoritarian electoral formulas. As we can discern fromthe unstandardized regression coefficient in Model 1, majoritarian electoralrules reduce predicted turnout by around 17%. Among the 32 recent leg-islative elections in Africa, 18 were conducted using proportional or mixedelectoral rules, and 14 were conducted using majoritarian electoral rules.The mean turnout for the more proportional systems is 71%, compared toan average of 59% for elections conducted using majoritarian rules. Noneof the elections using proportional formulas had electoral turnout rates ofless than 50% of registered voters.

On the other hand, multipartism does not have a strong effect on turnoutin Africa, which is also the case in Latin America (Fornos et al., 2004;Pérez-Liñán, 2001). The coefficient for ENLP is not significant, but it isworth noting that the sign of the coefficient for multipartism is positive,given the expectations in the literature that more parties will serve todepress turnout (Jackman, 1987; Kostadinova, 2003). Given the closenessbetween ENLP and other measures of political competition, the weak rela-tionship between multipartism and turnout could indicate that the competi-tiveness of electoral races exercises less influence on voter turnout in Africathan may be expected from literature on turnout in industrialized countries(Franklin, 2004).22

The influence of bicameralism on voting in Latin America is mixed:Pérez-Liñán (2001) finds that bicameralism is not related to turnout,whereas Fornos et al. (2004) report that turnout is systematically higher incountries with unicameral legislatures. We find that bicameralism does notexert much influence on voter turnout in Africa. The coefficient for bicam-eralism is not significant, although the sign, which is negative, is in theexpected direction. The limited influence of bicameralism on turnout maybe a result of the weak nature of bicameralism in Africa. In Lijphart’s(1999) terms, the power of the legislative bodies in Africa’s bicameral leg-islatures tends to be asymmetrical. Often, the members of the second cham-ber are not directly elected and may comprise traditional chiefs or stalwartsof the party in power.

As we alluded earlier, the marginalization of legislatures in Africa mayact as an institutional feature that depresses citizens’ motivations to partic-ipate in legislative elections. The coefficient for our measure of executivedominance, the number of years the current president has been in office, isnot significant, although it is negative, as predicted.

We find evidence to support the ideas of political learning or habituationin that voter turnout is higher in countries with greater electoral experience.

Kuenzi, Lambright / Voter Turnout in Africa 679

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

The coefficient for the number of legislative elections is positive and sig-nificant in Model 1.23 Previous experience with elections boosts turnout inAfrica rather than depressing turnout, which is just the contrary of whatKostandinova (2003) found in her study of the new democracies in EasternEurope.24

The results presented in Model 1 also indicate that the level of democ-racy matters, although not in expected ways. These results suggest thatvoter turnout in legislative elections is higher in countries with higher FHscores, which indicates less democracy. These results are somewhat sur-prising. The literature suggests that citizens will have little incentive to votewhen elections are not free and fair, because their votes are unlikely to havemuch effect under such conditions.25

Media exposure has a consistent significant, positive effect on voterturnout in Africa. One can imagine why such is the case. People with reg-ular access to radios are likely to be more knowledgeable about the elec-toral campaigns. They are also likely to be more informed about the timingand logistics of the election—where and when to vote. The number ofradios per person is highly correlated with different development indica-tors, such as GDP per capita. Thus, it is possible that the reason that thenumber of radios has such a powerful effect is that it captures both mediaaccess and level of development. We substituted adult literacy rate in placeof the number of radios, however, and found that it was not significantlyrelated to turnout, which suggests that our measure of media exposure cap-tures something more than development.

The size of the urban population does not systematically influence thelevel of voter turnout in Africa. Fornos et al. (2004) also find that urbaniza-tion is not related to legislative turnout in Latin America. Moreover, the signof the coefficient is consistently negative for the last elections, suggesting thatturnout may actually be lower where populations are more urbanized. As inthe case of Senegal’s former ruling party, Africa’s ruling parties often receivea great deal of support in rural areas and thus focus their mobilization effortsoutside the urban areas. The ability of African parties to mobilize voters mayalso be easier in rural Africa, where the threat of sanctions for not voting maybe more effective and resource scarcity increases the impact of party effortsto buy votes. This finding supports results of research on individual partici-pation, which indicates that rural voters are more likely to vote than urbanresidents (Bratton et al., 2005; Kuenzi & Lambright, 2005).

As can be seen in Model 1, one of the principal measures of economicperformance—annual change in per capita income—is not significantlyrelated to turnout.26 Although, as noted earlier, studies have proffered evidence

680 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

of economic voting in Africa at the individual level, our results suggest thatvoters in Africa are not systematically motivated to vote during times ofincreasing prosperity or periods of economic stagnation and decline. Oneexplanation for these divergent results is that the influence of economic per-formance on political participation is felt at the individual level in thatvoters respond to changes in their immediate economic conditions—changesnot always captured by aggregate economic measures.

Similarly, GDP per capita is not significant in Model 1. However, thereis a potential problem of collinearity in Model 1, which is discussed below,as GDP per capita is significantly related to a number of other variables inthe model. The correlation between radios per person and per capita incomeis .68 (p < .01) for the most recent legislative elections. When the numberof radios is excluded from the model, the coefficient for GDP per capita issignificant and positive, indicating that turnout is higher in more developedcountries (results not shown). Thus, we find some support for the socio-economic model. With regard to legislative turnout in Latin America, Fornoset al. (2004) find essentially no support for the socioeconomic model.

The correlations between independent variables and the variance inflationfactor scores suggest a potential problem of multicollinearity in Model 1.27

For example, GDP per capita and presidentialism are highly correlated witheach other and other variables in the model, including the number of previ-ous elections.28 For this reason, we reran the analysis and excluded thesetwo variables (see Model 2, Table 2).29 The results are essentially the same.The standard error of the estimate of Model 2 drops slightly to 11.55 from12.09 for Model 1.

We ran the model with data from the penultimate elections as a check onthe robustness of the results. The results displayed in Model 3 are generallysimilar to those of Model 2, although there are some differences. The coeffi-cients for concurrency and media exposure are still positive and significant.The coefficient for majoritarian electoral formula, however, is neither negativenor significant. The coefficient for level of democracy is negative, as we orig-inally hypothesized it would be, and insignificant. Moreover, the coefficientfor electoral experience is neither positive nor significant. The nature of theresults makes intuitive sense. We would expect that people need a little expe-rience with new institutions before their behavior is consistent with the incen-tives provided. Thus, we see stronger relationships between the institutionalvariables and turnout in the last elections than in the penultimate elections.

As noted above, results presented in Table 2 for Models 1 and 2 suggesta negative relationship between voter turnout and the level of democracy.To further explore this curious finding and perform another check on the

Kuenzi, Lambright / Voter Turnout in Africa 681

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

robustness of our results, the cases for which FH scores are greater than orequal to 11 prior to the election are excluded from the analysis. Countriesthat earn total scores greater than 10 are classified by FH as “not free.” Sixelections met this criterion and were excluded from analysis presented inModel 4: Cameroon (1997), Chad (2002), Kenya (2002), Mauritania (2001),Niger (1999), and Zimbabwe (2000). When these elections are excluded,the effect of level of democracy on turnout is no longer significant,although the sign of the coefficient remains positive.30 Other than that, theresults change little, although electoral experience is not significant.

In Model 5, the analysis is restricted to the elections conducted in onlythose African polities deemed electoral democracies by FH.31 The odd rela-tionship between FH scores and turnout is unchanged. The results of thisanalysis are quite consistent with results previously discussed.32 The coef-ficient for ENLP is significant, however, in the analysis for the electoraldemocracies. Perhaps smaller parties, which one finds in multiparty sys-tems, are able to operate more effectively in more democratic environ-ments, whereas the ruling parties dominate in the less democratic systems,preventing the benefits of multipartyism from being manifested. Thisanalysis confirms that the relationships between turnout and the variablesof interest are generally the same whether the sample includes elections ofthe less democratic multiparty electoral regimes or is restricted to the elec-tions of those regimes that are electoral democracies.

Each of the different models estimating turnout for Africa’s legislativeelections explains a substantial amount of the variance in electoral turnout.All the models estimating turnout for the most recent legislative electionsaccount for at least 60% of the overall variance in turnout, whereas themodel for the penultimate elections accounts for nearly 46% of the varia-tion in turnout. Moreover, each adjustment to Model 1 (for example,excluding per capita income and presidentialism to address the problem ofcollinearity in Model 2) lowers the standard error of the estimate of themodel, slightly suggesting a good overall fit. For example, the standarderror of Model 4 is 10.97 compared to 12.09 for Model 1. Moreover, theresults of each model remain the same when outliers with Cook’s D scorebeyond the critical threshold are excluded from the analysis (Cook &Weisberg, 1982).33 All of the variables remain significant using eitherrobust or regular standard errors in the models presented in Table 2,although in a few cases coefficients are significant at the .10 level ratherthan the .05 or .01 level when regular standard errors are used.34

Table 2 displays the results of several additional models that were run toagain check the robustness of our results. We expanded our set of cases to

682 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

include data from an additional eight countries (see Model 6 in Table 2).35

Included in this group are countries that have held only a single electionsince the transition to democracy (Angola, 1992; Guinea, 2002; SierraLeone, 2002; and Rwanda, 2003) and countries for which turnout data wereunavailable for more than a single election (Congo, 1993; EquatorialGuinea, 1999; Ethiopia, 2000; and Gabon 2001). The results are largely thesame. The most significant difference is that the coefficient for ENLP ispositive and significant at the .10 level. Model 7 displays results of theanalysis when turnout as a percentage of VAP is used as the dependent vari-able. Once again, our results are largely confirmed. Actually, the coeffi-cients for three other variables (bicameralism, urbanization, and ENLP) arein the same direction as those in Model 2, but they achieve significancehere. Given the small number of cases, it is hard to draw any firm conclu-sions about why such might be the case. Finally, in Model 8, the effect ofthe number of elections from independence through 1989 on turnout is alsotested and found not to be significant (and the coefficient is actually nega-tive). Thus, it appears that it is the number of elections held in multipartyregimes that matters for turnout. In sum, our results appear to be generallyrobust.

Implications and Conclusions

The above analysis indicates that it is possible to test within the Africancontext the theories generated to explain political outcomes outside Africa.Many of the central findings of the research on electoral turnout in otherregions reappear in our examination of voter turnout in Africa. Similarly,some of the central findings of the literature on voting in Africa usingindividual-level analysis are also confirmed by this study. Most important,our analysis provides evidence of the importance of political institutions inshaping political behavior in Africa. Turnout is systematically higher whenelections are conducted using proportional electoral rules and legislativeand presidential elections are held concurrently.

Moreover, our findings suggest that there may be a lag between theestablishment of institutions and the perception and reaction on the part ofcitizens to the incentives to which the institutions give rise. First, citizensmust have confidence in the stability of institutions before they begin toreact to the incentives provided by the institutions. We would also expect thatpeople need a little experience with new institutions before their behavior isconsistent with the incentives provided. Thus, we see stronger relationships

Kuenzi, Lambright / Voter Turnout in Africa 683

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

between the institutional variables and turnout in the last elections than inthe penultimate elections.

Institutions, however, certainly do not explain all of the variation in voterturnout rates across African countries. As we can see from the models dis-played in Table 2, although the models perform very well, a fair amount ofvariation remains to be explained. Moreover, socioeconomic and contextualvariables also influence turnout. Media exposure, measured as the numberof radios per capita, has a strong positive effect on turnout in Africa.Moreover, we do not explain all of the variance in turnout rates, so otherfactors clearly affect voter turnout as well. The political systems of Africaare well known for being characterized by personal rule, and multipartyelections are still a relatively new innovation in most African countries.Perhaps as the multiparty, electoral regimes become more institutionalized,institutional arrangements will explain a larger proportion of the variationof turnout rates across African countries. In this article, we test a number ofdifferent theories about the determinants of turnout and make a significantcontribution toward explaining turnout in Africa, but there are certainlyother idiosyncratic or country-specific factors yet to be uncovered. We werealso unable to test cultural explanations for turnout in this article. Perhapsthe next step in the research program is to look at the effect of some of thecultural variables, such as levels of social trust and participation in civilsociety, on turnout in Africa.

The findings of our study also have practical implications for the con-duct of elections in Africa. Holding presidential and legislative electionsconcurrently will boost voter turnout in legislative elections and probablyeliminate some redundant costs. Adopting proportional representation isalso likely to promote political participation among the citizenry. Finally,increasing people’s level of media exposure by expanding the number ofradios available to the population seems like a straightforward way to pro-mote voting. Donors and civil society organizations concerned with pro-moting political participation should consider this strategy.

The finding that turnout is higher in Africa’s less democratic regimes isunexpected but perhaps augurs well for the future. Research by Lindberg(2006) demonstrates that the act of holding elections, with time, contributesto improvements in civil liberties in African countries. Whether citizens arevoting in higher rates in Africa’s less democratic regimes in the hope ofchanging the system or simply as a result of strategic mobilization by rul-ing parties may be less important than the fact that citizens are voting.Lindberg’s (2006) work suggests that turning out to vote in these countriescould contribute to enhanced democracy in the future.

684 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Our analysis reveals a positive relationship between electoral experienceand voter turnout. This finding provides some evidence of democratic learn-ing or the acquisition of the habit of voting by African citizens. It remainsuncertain whether this trend will continue if future elections fall far short ofdemocratic standards of what constitutes a free and fair election. In a studyof the quality of elections in Africa, Lindberg (2003) finds evidence of improve-ment in the quality of elections with time. Stable or increasing voter turnoutwith time could contribute to this trend, which bodes well for the survivalof democracy in Africa’s multiparty regimes.

Notes

1. Bratton and van de Walle (1997) do report the correlation between competitiveness(percentage of the vote obtained by the winning candidate) and voter turnout in the foundingelections of 28 of the countries that embarked on transitions to democracy in the early 1990s,which is -.499 (p = .007).

2. Bratton and van de Walle (1997) define neopatrimonialism as “those hybrid politicalsystems in which the customs and patterns of patrimonialism co-exist with, and suffuse, rational-legal institutions” (p. 62).

3. See Wilson and Butler’s (2004) discussion of the problems related to analysis of panelcross-sectional data as laid out in Beck and Katz (1995).

4. For three of our cases (Togo, Côte d’Ivoire, and Niger), turnout data were not availablefor the penultimate election. We use data from the previous election. For two of our cases(Cameroon and Mali), turnout data were not available for the most recent election, and weinclude data from the previous two elections.

5. In a sense, we are looking at turnout in Gambia’s two democratic regimes: that whichwas in place before the democratic reversal caused by the coup and that which was constructedafter the democratic reversal.

6. Turnout data are from the International Institute for Democracy and Electoral Assistance(IDEA) Web site, (www.idea.int). Data were obtained from the Inter-Parliamentary Uniondatabase (www.ipu.org), if unavailable from IDEA’s Web site, which was the case for 12 elec-tions. Turnout for Zimbabwe’s 2000 election was obtained from IFES’s election guide(www.electionguide.org).

7. See http://www.idea.int for more information.8. IDEA reports that turnout as a percentage of the voting-age population is 129% in

Guinea Bissau’s 1999 election. The reported VAP in 1999 is 311,674 compared to the numberof total votes (402,400) in 1999. A surprising finding was that the reported voting-age popu-lation for Guinea Bissau in 1999 is considerably lower than that reported for 1994, when556,500 citizens were reported to have been eligible for vote.

9. As Radcliff (1996) argues, “. . . registration rolls themselves may not be entirely accurate,but they are likely to be at least as accurate as our tortuous efforts at estimating the eligiblepopulation” (p. 719).

10. Pearson’s r = .7607, p < .01, for the last parliamentary elections and r = .7709, p < .01,for the penultimate elections.

11. Because the preponderance of the second chambers is less powerful than the corre-sponding first chambers in African countries, we use this simple binary variable. We did, however,

Kuenzi, Lambright / Voter Turnout in Africa 685

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

also use two other measures of bicameralism. First, we drew on Lijphart’s (1999) measure ofthe level of bicameralism, which ranges from 1 to 4, depending on whether the power of thechambers is symmetric and the selection of the members is incongruent. Second, we createda simple measure that ranges from 0 to 2. Unicameral legislatures are coded as 0. Bicamerallegislatures that are characterized by any type of power asymmetry are coded as 1, and bicam-eral legislatures in which the two houses are equally powerful are coded as 2. Each of thesemeasures produces almost exactly the same results when they are used in the model, and thesemeasures are extremely highly correlated.

12. Laakso and Taagepera’s (1979) formula for calculating the effective number of partiesis N = 1/Σsi

2.13. See the methodology section of www.freedomhouse.org.14. For elections held in 2003 and 2004, we used the Polity scores from 2001, because

2001 is the last year covered by the Polity dataset. Combined Polity Scores are from the PolityIV dataset.

15. Data on electoral experience prior to democratic transition were obtained from Bratton,M. & van de Walle, N., with K. Butler, S. Chan, K. Ludwig, & Y. Wang. (1996). Politicalregimes and regime transitions in Africa: A comparative handbook (Afrobarometer WorkingPaper No. 14). Michigan State University: East Lansing.

16. Data on the size of the urban population were obtained from the United Nations, Worldpopulation prospects Web site: The 2002 revision and world urbanization prospects: The 2001Revision, available from http://esa.un.org/unpp. Data on urban population are tracked to theyear of the election or as close to the year of the election as possible.

17. We used data from the IMF World Economic Outlook (April 2003) to measure percapita GDP in current prices (U.S. dollars) and the annual percentage change in real gross GDP.

18. Literacy rates are positively and significantly correlated with the number of radios(r = .5960), per capita income (r = .5026), and the size of the urban population (r = .3545) forthe most recent set of elections.

19. Data on the number of radios were obtained from www.abacci.com/atlas.20. As one of the anonymous reviewers noted, one might want to take this turnout figure

for Niger “with a grain of Saharan salt.” It seems quite reasonable to expect that governmentsmay try to inflate the voter turnout rates, but the reported turnout rates do seem in line withthe actual level of turnout in most countries.

21. See “Hundreds May Not Vote in River Gee,” by Martin Nyeka, published in TheAnalyst, September 7, 2005. Accessed September 30, 2005, from www.allAfrica.com.

22. The effective number of legislative parties (ENLP) is highly correlated with the differ-ence in seat shares between the largest and second largest parties (r = −.6638, p < .01). Whenthe difference in seat shares between the two largest parties is substituted into the model forENLP, this measure of competitiveness is also not significantly related to turnout (results notshown). Fornos et al. (2004) use a similar measure of electoral competition but look at the dif-ference in vote shares of the two largest parties. We use the seat share given the difficulty inobtaining vote shares of parties in many African countries. We also tested the effects of analternative measure of competition—the seat share of the party with the largest share of theseats. This measure is also unrelated to turnout when substituted for ENLP in our analysis.

23. The number of elections is no longer significant in Model 1 if Africa’s four parlia-mentary systems are excluded.

24. We tested other measures that capture electoral experience before transitions to democ-racy in the 1990s and find that they are not systematically related to turnout. For example, a mea-sure of the total number of elections held since independence through 1989 is not significantly

686 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

related to turnout in recent legislative elections in either bivariate or multivariate analysis (seeTable 2). We also looked at the effect of previous experience with different regime types onturnout but could not display the model because including them left too few degrees of freedom.There did, however, appear to be a positive relationship between years under personal dictator-ship and turnout.

25. The regression results differ substantially from the bivariate relationships revealed incorrelations. Freedom House (FH) scores are unrelated to turnout in the most recent elections(r = -.0744) and significantly and negatively related to turnout in the penultimate elections(r = −.3519, p < .05).

26. Growth rates are not correlated with per capita income in Africa (r = -.0760 for the lastelections).

27. We refer to Hamilton (2002) to determine the value of the mean variance inflationfactor (VIF) score for the model or value of the VIF score for a single independent variable thatindicates multicollinearity. If a single VIF score is greater than 10 or if the mean VIF score isgreater than 1, multicollinearity is likely to be present and affecting the regression results.

28. The correlation between GDP per capita and presidentialism is r = −.4763, p < .01,whereas presidentialism is also correlated with the number of elections (r = −.5287, p < .01)for the last elections. GDP per capita is also correlated with the number of elections (r = .4058,p < .05) for the last elections.

29. The mean VIF for Model 1 is 1.83, and the VIF scores for GDP per capita and presi-dentialism are 3.23 and 2.67, respectively. The mean VIF for Model 2, when these two vari-ables are excluded, drops to 1.27. Finally, presidentialism is not significantly related to turnouteven in different configurations of the model in which GDP per capita and the number ofradios are excluded individually or at the same time.

30. When we substitute Polity scores for Freedom House (FH) scores, the results are sim-ilar. The Polity scores are significantly and negatively related to turnout, which means thatturnout is lower in more democratic regimes. The effect of Polity scores on turnout, however,disappears when outliers are excluded or when countries with Polity scores below -5 areexcluded, but the coefficient remains negative.

31. FH considers specifically the conduct of national elections and whether elections are“free and fair.”

32. We also ran our analysis for 13 of the 14 African countries that Posner and Simon(2002) identify in FN 3 (p. 314) as having held elections that were free and fair. When werestrict our analysis to these 13 countries, the results are essentially the same. The only dif-ference is that the measure of majoritarian electoral formula is not statistically significant butcome close to reaching significance at the .10 level.

33. Five cases were identified as potentially influencing the results of Model 1 given aCook’s D value greater than the critical threshold: Botswana, 2004; Djibouti, 2003; GuineaBissau, 1999; Malawi, 1999; and Niger, 1999. The results of Model 1 are the same when thesefive cases are excluded. Four of these cases were identified as potential outliers in Model 2, andthe results of the analysis are the same when these cases are excluded. Five cases were identi-fied as potential outliers in Model 4. The results are similar when these cases are excluded fromModel 4. The results are the same when influential cases, identified by |DFBETA scores| >2/sqrt (n), are excluded. When 12 cases with DFBETA scores beyond the critical threshold areexcluded, all of the variables significant in Models 1 and 2 remain significant.

34. For example, in Model 1, electoral experience is significant at the .10 level with regu-lar standard errors. In Model 3, concurrency, bicameralism, and level of democracy are eachsignificant at the .10 level if regular standard errors are used.

Kuenzi, Lambright / Voter Turnout in Africa 687

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

35. Excluded from our analysis are countries that have not held multiparty elections as aresult of ongoing civil wars and/or because the political system precludes any formal opposi-tion (e.g., Sudan, Liberia, Democratic Republic of Congo, Somalia, Eritrea, Burundi, Uganda,and Swaziland). Comoros is also excluded because turnout data were not available for eitherthe 1996 or the 2002 parliamentary elections.

References

Beck, N., & Katz, J. N. (1995). What to do (and not to do) with time-series cross-section data.American Political Science Review, 89, 634-647.

Blais, A., & Dobrzynska, A. (1998). Turnout in electoral democracies. European Journal ofPolitical Research, 33, 239-261.

Bratton, M., Mattes, R., & Gyimah-Boadi, E. (2005). Public opinion, democracy and marketrecovery in Africa. Cambridge, UK: Cambridge University Press.

Bratton, M., & van de Walle, N. (1997). Democratic experiments in Africa: Regime transitionsin comparative perspective. Cambridge, UK: Cambridge University Press.

Brockington, D. (2004). The paradox of proportional representation: The effects of party sys-tems and coalitions on individuals’ electoral participation. Political Studies, 52, 469-490.

Cook, R. D., & Weisberg, S. (1982). Residuals and influence in regression. London: Chapmanand Hall.

Diamond, L. (2002). Elections without democracy: Thinking about hybrid regimes. Journal ofDemocracy, 13, 21-35.

Fornos, C. A., Power, T. J., & Garand, J. C. (2004). Explaining voter turnout in Latin America,1980-2000. Comparative Political Studies, 37, 909-940.

Franklin, M. (2004). Voter turnout and the dynamics of electoral competition in establisheddemocracies since 1945. Cambridge, UK: Cambridge University Press.Gerber, A. S., Green, D. P., & Shachar, R. (2003). Voting may be habit-forming: Evidence

from a randomized field experiment. American Journal of Political Science, 47, 540-550.Hamilton, L. C. (2002). Statistics with stata. College Station, TX: Stata Corporation.Jackman, R. W. (1987). Political institutions and voter turnout in the industrial democracies.

American Political Science Review, 81, 405-424.Jackman, R. W., & Miller, R. A. (1995). Voter turnout in the industrialized democracies dur-

ing the 1980s. Comparative Political Studies, 27, 467-492.Jenkins, J. C., & Kposowa, A. J. (1992). The political origins of African military coups: Ethnic

competition, military centrality, and the struggle over the postcolonial state. InternationalStudies Quarterly, 36, 271-291.

Kostadinova, T. (2003). Voter turnout dynamics in post-communist Europe. European Journalof Political Research, 42, 741-759.

Kuenzi, M., & Lambright, G. (2001). Party system institutionalization in 30 African countries.Party Politics, 7, 437-468.

Kuenzi, M., & Lambright, G. (2005). Who votes in Africa: An examination of electoral turnoutin 10 African countries (Afrobarometer Working Paper No. 51). Michigan StateUniversity, East Lansing.

Laasko, M., & Taagepera, R. (1979). Effective number of parties: A measure with applicationto West Europe. Comparative Political Studies, 12, 3-27.

688 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Lijphart, A. (1997). Unequal participation: democracy’s unresolved dilemma. AmericanPolitical Science Review, 91, 1-14.

Lijphart, A. (1999). Patterns of democracy: Government forms and performance in thirty-sixcountries. New Haven, CT: Yale University Press.

Lindberg, S. I. (2003, November). The ‘democraticness’of multiparty elections: Participation,competition, and legitimacy in Africa. Paper presented at the annual meeting of the AfricanStudies Association, Washington, DC.

Lindberg, S. I. (2006). The surprising significance of African elections. Journal of Democracy,17, 139-151.

Lipset, S. M. (1959). Some social requisites of democracy: Economic development and polit-ical legitimacy. American Political Science Review, 53, 69-105.

Norris, P. (2002). Democratic phoenix: Reinventing political activism. Cambridge, UK:Cambridge University Press.

Norris, P., & Mattes. R. (2003). Does ethnicity determine support for the governing party?(Afrobarometer Working Paper No. 26). Michigan State University, East Lansing.

Pérez-Liñán, A. (2001). Neoinsitutional accounts of voter turnout: Moving beyond industrialdemocracies. Electoral Studies, 20, 281-297.

Posner, D. N., & Simon, D. J. (2002). Economic conditions and incumbent support in Africa’snew democracies: Evidence from Zambia. Comparative Political Studies, 35, 313-338.

Powell, G. B., Jr. (1980). Voting turnout in thirty democracies: Partisan, legal, and socio-economic influences. In R. Rose (Ed.), Electoral participation: A comparative analysis.(pp. 5-34). Beverly Hills, CA: Sage.

Radcliff, B. (1992). The welfare state, turnout and the economy: A comparative analysis.American Political Science Review, 86, 444-454.

Radcliff, B. (1996). Turnout and the economy: A comment on Jackman and Miller’s ‘voterturnout in industrialized democracies in the 1980s.’ Comparative Political Studies, 29,718-723.

Rustow, D. A. (1970). Transitions to democracy: Toward a dynamic model. ComparativePolitics, 2, 337-363.

Schaffer, F. C. (1998). Democracy in translation: Understanding politics in an unfamiliar culture.Ithaca, NY: Cornell University Press.

Schatzberg, M. G. (2001). Political legitimacy in middle Africa: Father, family, food.Bloomington: Indiana University Press.

Siaroff, A., & Merer, J. W. A. (2002). Parliamentary election turnout in Europe since 1990.Political Studies, 50, 916-927.

Teixeira, R. A. (1987). Why Americans don’t vote: Turnout decline in the United States, 1960-1984. Westport, CT: Greenwood.

van de Walle, N. (2003). Presidentialism and clientelism in Africa’s emerging party systems.Journal of Modern African Studies, 41, 297-321.

Verba, S., Schlozman, K., & Brady, H. (1995). Voice and equality: Civic voluntarism inAmerican politics. Cambridge, MA: Harvard University Press.

Wilson, S. E., & Butler. D. M. (2004). A lot more to do: The promise and peril of panel datain political science. Unpublished manuscript.

Wolfinger, R., & Rosenstone, S. (1980). Who votes? New Haven, CT: Yale UniversityPress.

Youde, J. (2005). Economics and government popularity in Ghana. Electoral Studies,24, 1-16.

Kuenzi, Lambright / Voter Turnout in Africa 689

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from

Michelle Kuenzi is an assistant professor at the University of Nevada, Las Vegas. Much of herresearch has focused on public opinion and political behavior in Senegal. She has publishedon these subjects in Democratization and Community Development Journal and has an articleforthcoming in Political Behavior. She has also conducted research on party systems and theconsolidation of democracy in Africa and has published on this subject in Party Politics (withGina Lambright).

Gina Lambright is an assistant professor of political science and international affairs at GeorgeWashington University in Washington, D.C. Her research interests include political reform,decentralization, and the politics of development in Africa. She has conducted and publishedresearch (with Michelle Kuenzi) on party systems in Africa’s new multiparty regimes in PartyPolitics. She is also currently working on a project exploring local government performance inseveral African countries.

690 Comparative Political Studies

at NATIONAL CHUNG HSING UNIV on March 28, 2014cps.sagepub.comDownloaded from