Explaining the African Vote

Explaining the African Vote Karen Ferree Department of Political Science University of California, San Diego Clark Gibson Department of Political Science University of California, San Diego Barak D. Hoffman1 Center for Democracy and Civil Society Georgetown University James D. Long Department of Political Science University of California, San Diego February 2009 In a shocking victory in Ghana’s 2008 presidential election, the opposition National Democratic Congress (NDC) prevailed over the incumbent New Patriotic Party (NPP) by one-half of one percentage point, after trailing far behind the NPP according to most opinion polls. What were the attributes of voters for each party? Standard theories about elections in Africa suggest that they are little more than ethnic headcounts and that parties typically are a thin cover for ethnicity. Data from a survey we conducted shortly before the 2008 seriously challenges this view. The NDC and the NPP drew support from a range of ethnic groups and there was little evidence of ethnic block voting. Moreover, while supporters of each party do not fit a clear ethnic profile, they possess strong beliefs about the parties. In addition, in the final weeks of the election, a sizable proportion of the electorate remained undecided and was amenable to voting for either party. For these reasons we contend that the perceptions of the NDC and the NPP shaped the outcome of Ghana’s 2008 election far more than the ethnic identity of its candidates. 1 bdh29@georgetown.edu Ghana’s Undecided Voters Introduction Ghana’s 2008 presidential election caught the world’s attention. In a close presidential run-off, the opposition National Democratic Congress (NDC) defeated the incumbent New Patriotic Party (NPP) by less than one-half of a percentage point after the NPP came in first place in the first round by just over one percent.2 International and domestic observers praised the Electoral Commission of Ghana (EC) for its professionalism in counting the ballots in a very close set of races and the parties for their adherence to democratic procedures during the campaign. In his speech to the Ghanaian parliament in July 2009, US President Obama remarked that “[t]he people of Ghana have worked hard to put democracy on a firmer footing, with peaceful transfers of power even in the wake of closely contested elections.” Despite the recognition due to the EC and parties for a well-managed election, analysts have focused less attention on why voters chose the party they did. In particular, few have focused on the extent to which ethnicity versus attributes of the parties determined for whom Ghanaians cast their vote. Did Ghanaians have firm opinions of parties? Did NDC and NPP voters form distinct ethnic blocks or did each party attract votes from a diverse range of ethnic groups? Did the NDC prevail because of ethnic appeals or because it appealed to a wide range of voters? Such questions are relevant because they examine a core debate about elections in sub-Saharan Africa: are elections little more than ethnic headcounts or do beliefs about parties shape vote choice in ways far more complex that ethnic group membership? The data from the 2008 election in Ghana are clear: beliefs about parties, not ethnicity, determined vote choice. While supporters of the NPP and NDC were ethnically heterogeneous within and across ethnic groups, supporters for each party possessed very strong sets of beliefs about the parties, including the incumbent NPP’s performance. Thus, party attributes, not the ethnicities of candidates, formed the primary axis of political competition during the election. In the next section, we review standard approaches to voting in Africa that highlight the importance of ethnicity to of voting, and contrast them with those on institutional performance of parties. Subsequently, we analyze the findings of existing studies of voting patterns in Ghanaian election. In the fourth section, we describe our approach to studying Ghana’s 2008 election. Next, we describe the hypotheses we test to determine party support. In the following section we show the empirical results. The last section concludes. Theoretical Foundations This study engages two broad approaches to voting. First, a large literature on democracy in multi-ethnic and developing countries suggests ethnicity is the prime predictor of voting patterns. This general literature employs a number of different channels and mechanisms through which ethnicity structures vote choice. One broad approach states that binds of identity are strong enough to cause anxiety of out-group members and positive evaluations of in-group members, regardless of what those members do in office. A second one highlights that the distribution of 2 The NPP and NDC received 49% and 48% of the vote in the first round, respectively. In Ghana, if no party wins at least 50% in the first round, the two top parties compete in the second round. 1 goods, including patronage and services, in poor democracies is primarily determined follows ethnic lines. We contrast these approaches to voting in multi-ethnic democracies with standard approaches to voting in more established democracies that examine party attributes, behavior, and performance. Ethnic Voting Scholars maintain that ethnicity provides a strong predictor of vote choice in multi-ethnic and nascent democracies and that elections produce ethnic headcounts where electoral results overlap the census as voters line-up with their co-ethnic candidates. Existing studies point to two broad channels that may link ethnicity and voting behavior: strong assertions of identity and expectations of patronage or policy favoritism. First, voters may select co-ethnic candidates because they hold nearly uniform positive evaluations of in-group members and receive personal gratification from choosing people like them. Voting is thus an expression of social solidarity deriving from group affinity and voters use their ballots to assert and affirm their ethnic identity (Horowitz 1985). These strong in-group attachments are reinforced by negative evaluations of out-group members, which may produce ethnic fear, anxiety, and animosity (Horowitz 1985; Geertz 1963; Oliver and Mendelberg 2000; Key 1949; Snyder 2000;). Voters will select co-ethnics to avoid what they perceive to be the negative consequences of governing from another ethnic group. This may produce voting for coethnic candidates or multi-ethnic parties and coalitions that contain a person’s group and bloc voting. Thus elections in divided societies do little more than produce ethnic head counts, where patterns of voting match the ethnic census (Horowitz 1985, Rabushka and Shepsle 1972; Dickson and Scheve 2006). As a result, in ethnically polarized societies, voters may prioritize winning elections by fraudulent means over democratic process. Second, voters may select co-ethnics to receive individually targeted goods or club goods for their group. The existence, depth, and importance of patronage are central to understanding African politics (Bates 1973; Cruise O’Brien 1971, 2003; Bayart 1993; Chabal and Daloz 1999; Schaffer 2000). Scholars argue that clientelism is a strategy politicians employ to channel goods towards co-ethnics in exchange for electoral support. Politicians generally use clientelism to motivate voters within their own group and reward them with private goods or benefits such as public service jobs if successful (Posner 2005; Wantchekon 2003). A variant of the patronage hypothesis expands beyond the transfer of individualized goods and includes ethnic bias in service delivery. Politicians use favoritism - such as the construction of schools and roads - to benefit co-ethnic supporters at the expense of other groups. While patronage politics is not exclusive to Africa, or even to poor, under-developed, or multiethnic societies (Kitschelt and Wilkinson 2007), scholars studying Africa commonly assert that leaders distribute patronage along ethnic lines. Patronage and policy favoritism may follow lines of ethnicity because shared identity provides better methods of coordination and mobilization (Bates 1974). Coordination may be easier within groups than between them because similarities of language and culture provide denser networks (Hardin 1995, Fearon and Laitin 1996, Habyarimana et al. 2009). Co-ethnicity also produces norms of reciprocity (Bates 1974, Taylor 1988) that allow for easier sanctioning of in-group members (Miguel and Gugerty 2005). 2 Consequently, ethnicity provides an important source of information to voters about what they can expect from their elected officials (Chandra 2004; Mattes 1995; Posner 2005; Ferree 2006; Birnir 2007). Party Attributes and Performance These two broad ethnic approaches assume that what drives party support is who parties are, not what they do. But an extensive literature from more established democracies focuses the attraction of candidates and parties on how well they have performed in the past and other attributes. Theories of performance and issue voting focus on the importance of politicians’ accomplishments and candidate evaluations by voters. According to these lines of reasoning, partisan identification is a strong predictor of the vote and voters form attachment to a parties based on their ideologies and performance in office. Downs (1957), for example, contends that voters make an investment decision in parties that render them the highest utility relative to the other options on offer, including the policy outcomes the citizen desires and the party provides. Party labels therefore help voters decide with informational uncertainties. Fiorina (1981) asserts that when incumbents perform well, they are returned to office. When they perform poorly, by contrast, opposition candidates are more successful. These retrospective, or backwards looking, models typically rely on economic voting, where voters choose the party that best represents their economic interests (Duch and Stevenson 2007). For Fiorina, party identification reflects a running tally of past evaluations of a party’s performance and is malleable over time. Voting is not purely about the past or how parties have historically behaved, however. Voters must also choose between the policies that potential rivals propose if elected (Goldberg 1966, Hartwig et al. 1980). Thus, voters are also prospective or forward-looking (Fearon 1999; Downs 1957; Lockerbie 1991; MacKuen, Erikson, and Stimson 1992). Spatial models of voting, for example, assert that voters select candidates that are closest to them on policy issues (Downs 1957, Page and Jones 1979, Riker and Ordeshook 1973, Stokes 1963). In addition, voters may rely upon other candidate qualities, such as age or experience, to assess challengers who do not have records to evaluate (Popkin 1994). While existing studies of performance and issue-voting focus primarily on industrialized democracies, recent work has begun extending this line of research to sub-Saharan Africa. Youde (2005) demonstrates the importance of economic performance in predicting government support in Ghana, while Lindberg and Morrison (2008) argue that voters in Ghana care more about performance than ethnicity or patronage. Multiple studies (Mattes 1995; Mattes and Piombo 2001; Ferree 2006, 2008) have explored economic voting in South Africa. Bratton and Kimenyi (2008) argue that economic performance considerations strongly shaped voting in the 2007 Kenyan elections, even in a context of heightened politicization of ethnicity. Using exit poll data from Kenya’s 2007 election, Gibson and Long (2009) also find that Kenyan voters preference performance over ethnicity. Norris and Mattes (2003) offer one of the first ever crossnational studies of voting in Africa and suggest that both ethnicity and economic performance affect vote choice. 3 However, matching purely retrospective or prospective model to voters and parties in Africa poses many challenges. First, African parties are fairly young, and therefore prove fluid both in labels and membership following recent transitions to electoral democracy. Second, voters may have a difficult time assessing the performance and policies of incumbents given information scarcities (Posner 2005). Third, scholars have argued that inchoate African parties do not yet divide along lines of ideology, making it difficult for citizens to place their own ideological commitments relative to a party’s (Burnell 2001; Di Lorenzo and Sborgi 2001; Nugent 2001; Posner 2005). Fourth, the continued dominance of hegemonic parties in many countries have left an imbalance in perceptions of party credibility: opposition promises may ring hollow since they have rarely if ever held power. Both ethnic and performance theories therefore present a priori insights and challenges in application to the study of African voting behavior. Furthermore, they present important predictions about the existence of swing voters, people who do not align with any specific party and are amenable to voting for more than one party. Ethnic theories maintain that swing voters should not exist as individual voters. Rather, voters align themselves with their co-ethic or vote in a bloc if a co-ethnic is not a candidate for office. Alternatively, theories of party attributes presume that candidates for office can persuade individual voters and expect a diverse range of preferences within ethnic groups. Voting in Ghana The three main hypotheses that emerge from existing studies of voting behavior specifically in Ghanaian elections concern ethnicity, demographics, and the characteristics of swing voters. We distill the most common themes for vote choice that emerge from these analyses before considering the limitations of these studies and how our survey data in this paper improves upon on them. Like most countries in Africa, scholars who have examined Ghanaian elections tend to find strong patterns of ethnic voting. In Ghana, this does not just involve the ethnicity of candidates but also the ethnic labels of parties. The most basic hypothesis is that the Asante ethnic group votes for the NPP and the Ewe support the NDC (Fridy 2007, Morrison and Woo-Hong 2006, Nugent 2001). The reason the former support the NPP is because it is the successor to the independence-era Asante party, the United Gold Coast Convention (Lindberg and Morrison 2005, Morrison 2004). The Ewe, by contrast, support the NDC as it is the party that Jerry Rawlings, an Ewe, founded prior to Ghana’s 1992 democratic transition (Lindberg and Morrison 2005, Morrison 2004). The table below that shows the size of the main ethnic groups in Ghana as a share of the total population demonstrates that these patterns form a theory of ethnic voting that is highly incomplete to explain the outcomes of Ghana’s elections. Ethnic Groups in Ghana Group Akan of which: Asante Fante Share 45% 15% 10% 4 Other Ewe Ga MoleDagbani Others 20% 12% 7% 15% 21% The first problem the table demonstrates is that combined, the Asante and the Ewe account for only approximately 30% of the population, leaving 70% of Ghanaians without a clear ethnic party. Scholars of Ghanaian election address this problem by asserting that the Asante and the Ewe possess reliable ethnic alliances. However, there is a lack of consensus on which groups support each party, apart from agreement that the Mole-Dagbani tend to ally with the Ewe. More specifically, a synthesis of existing studies shows that scholars find that all regions of the country except Asante, Upper East, Upper West, and Volta, have switched their voting between the NDC and the NPP across elections (e.g., Fridy 2007, Morrison and Woo-Hong 2006, Nugent 2001). As the regions containing strong NDC and NPP supporters only account for approximately onethird of the total electorate, potential swing regions contain twice as many voters as regions that are not competitive according to these studies. The second problem is that the Asante are one member of a larger ethnic group, the Akan. While it is logical to presume that Akan would choose to support a near co-ethnic, they have never congealed politically as a coherent ethnic group (Fridy 2007). More problematic, in the 2008 election, the NDC and NPP candidates came from groups that are small subsets of the Akan. The NDC candidate in the 2008 election, John Atta-Mills was a Fante, and the NPP candidate, Nana Akufo-Addo, was an Akim. Consequently, the only Akan that possessed a clear ethnic candidate in the 2008 election were the Akim, Asante, and Fante.3 These three groups account for about half of the total Akan population. The second most common hypothesis about voting in Ghana concerns demographic predictors of voting other than ethnicity. The NPP supposedly attracts well-educated and urban voters while the NDC appeals to the less well-educated voters, especially those living in rural areas (Jeffries 1998, Morrison 2004). Scholars advance two main arguments for why voting patterns occur along these lines. The first addresses the different policies each party advocates. According to these scholars, the NPP is a market-oriented party while the NDC espouses a populist economic program (Morrison 2004). The second suggests that differences in the policies reflect the dissimilar ethnic bases of the parties (Jeffries 1998). Specifically, the NPP represents the interests of the Asante, the wealthiest ethnic group in the country, while the NDC stands for those of the Ewe in Volta Region and groups living in northern Ghana, economically marginal parts of the country. The NDC’s populism thus reflects the economic interests of these groups. The extent of swing voters in Ghana has also received prior attention. Based on patterns of vote switches between the NDC and the NPP, scholars contend that approximately 15% of Ghanaian 3 We consider the Asante to have a clear ethnic choice in the NPP because of their historical ties with the party and because many of the senior leaders of the NPP are Asante. 5 voters are open to voting for either party (Lindberg and Morrison 2008, Morrison 2004).4 The controversy surrounds the motivations of swing voters. Some scholars argue that they vote as ethnic blocs (Fridy 2007, Morrison and Woo-Hong 2007). The competing hypothesis argues that voter characteristics, not ethnicity, are the defining feature of swing voters (Jeffries 1998, Lindberg and Morrison 2005). Specifically, more educated voters and those who are most concerned about the quality of democracy in Ghana are more likely to switch votes.5 Existing studies have trouble adjudicating between the aforementioned hypotheses for two main reasons. First, hypotheses about ethnic, demographic, and party determinants for voting can each be true simultaneously as voters may possess a diverse range of preferences and have subjective perceptions of performance. It is entirely plausible that while the NPP and NDC both have ethnic bases, the parties can only win an election by expanding their support through appeals beyond their ethnic groups since the Asante and the Ewe account for only a small minority of Ghanaian voters. One mechanism could be highlighting policy and performance differences between the two parties. Additionally, weak methodologies from extant studies make identifying the factors driving party support difficult to isolate. Some studies match census data with election returns (e.g., Anebo 2001, Nugent 2001). There are three problems with this approach. First, because many districts in Ghana are ethnically homogeneous (especially rural ones), these studies run the risk of finding a spurious correlation between regional party strength and ethnicity deriving from unobservable factors, such as perceptions of government performance. Second, ecological fallacy, attributing individual behavior from group-level data, inhibits our capacity to determine the veracity of these studies as they are unable to isolate the motivations of individual voters. These problems are especially acute in ethnically and demographically mixed areas, primarily cities. Third, it is very difficult for these studies to determine how the performance of the incumbent party affects voting due to their inability to isolate voters’ perceptions of it. Studies that attempt to explain voting at the district or regional level as a consequence of government performance in those areas implicitly make two assumptions that are difficult to sustain. One, voters evaluate the effectiveness of the government based on objective measures of performance, not subjective ones. Two, voters determine performance by comparing their own district or region relative to others. It is equally as plausible that voters evaluate party performance based on how their own situation has changed relative to what it was in the past, or others they know. Scholars have also employed two sets of surveys, the Afrobarometer and Lindberg and Morrison (2008) to attempt to derive patterns of support from individual-level data. Both these surveys have flaws that inhibit their effectiveness in this task. Lindberg and Morrison’s survey contains voters from only six constituencies out of 200 and is not a random sample of constituencies or voters in them. As a result, it is impossible to project their findings to the national level. They also ask voters directly about the logic structuring their individual vote choice, which may elicit socially desirable (i.e., non-ethnic) responses rather than sincere ones. Second, only one survey (Afrobarometer 2005) occurred close to an election, but it did not ask people for whom they voted. The Afrobarometer Survey 2008 took place nine months before that year’s election and 4 It is important to note that this definition is somewhat tautological as it states the potential swing voters are those who have switched parties in the past. 5 Notice, however, that this theory contradicts, in part, the theory that educated voters are more likely to vote NPP. 6 Lindberg and Morrison conducted their survey a year and a half before the 2004 election. Youde (2005) uses Afrobarometer data from 1999 to measure levels of government support in nonelectoral periods. At the times these surveys occurred, the parties had not identified the issues they would raise and voters did not know the ethnicities of candidates for office. It is plausible that these perceptions may change significantly once voters encounter specific candidates and issues. Direct tests for the effects of ethnicity and performance can only result from extensive polling during the electoral period to uncover the microfoundations of choice. There are several reasons why polls occurring at the time of the election possess advantages over surveys farther from them when seeking the determinants of vote choice. First, unlike pre-election and post-election polls, surveys conducted more than a few months before an election cannot present respondents with the accurate list of parties, candidates or issues they will likely choose from in an election. A key component to measuring the importance of ethnicity is to present voters with the actual candidate choices, including a candidate’s ethnicity. In most African countries, such information is not available until right before elections (this is especially true of parliamentary and local races, and where parties and coalitions are ephemeral and frequently change). 6 Moreover, voters often do not have solid opinions in election surveys that occur far before an election. Crucially, these surveys can downplay the salience of ethnicity as the heated rhetoric and polarization around ethnicity is often most evident during voting, as Posner, Eifert, and Miguel (2007) demonstrate. Therefore, electoral periods should be the hardest cases to test performance theories as prior evidence shows ethnic attachments increase at this time. Surveys occurring more than a few months after an election also suffer weaknesses. Evidence from surveys conducted in the United States demonstrates that voters often forget for whom they voted (particularly for downthe-ballot races) and/or misreport how they voted after they know the outcome of the election. Our Approach We synthesize prior approaches to ethnic voting and performance voting to build a more general model amenable to empirical testing. We do so with particular reference to the role that information plays in an individual’s vote choice. We argue that Ghanaians vote as a function of the information they have about the performance, policy differences, and ethnicities of candidates and parties. A recent strand of research on voting in ethnically divided countries employs informational theories of voting (Downs 1957; Popkin 1994; Dawson 1994) to argue that voters use ethnicity as an information shortcut about the likely behavior of politicians. Mattes (1995), Chandra (2004) and Posner (2005) argue that ethnicity becomes the basis of politics in nascent multiethnic democracies because it is one of the few sources of reliable information available to voters and sends a strong signal about the likely distribution of patronage and goods .7 Ferree (2006 and 2008) suggests that South African voters rely on the racial credentials to distinguish credible from non-credible promises. 6 7 Parliamentary candidates were nominated about one month before elections in Ghana. Chandra (2004) studies India, but argues that patronage-democracies are the norm in Africa and Asia. 7 Ethnicity, however, does not always provide sufficient or clarifying information to determine an individual’s vote. In Ghana’s 2008 presidential election, for example, the ethnic groups of the two major candidates represented each only approximately 15% of the total population. Consequently, the vast majority of Ghanaians could not use shared ethnicity as the motivating factor determining whom they would support. In addition, ethnicity can confuse as well as inform. For example: consider the Fante, typically an ally of the Asante and the NPP. The presidential candidate that was their co-ethnic in the 2008 election, John Atta-Mills, ran for the NDC, a party that typically represents the Ewe, a rival ethnic group. Simply knowing the ethnic identity of the NDC presidential candidate thus did not convey clear information to Fante voters. In contrast to theories that stress the primacy of ethnic cues, we argue that voters often require and acquire additional information about the parties beyond the ethnicities of their candidates. In particular, we assert that voters employ information about the policies of the NDC and NPP, and the performance of the incumbent NPP government to determine for whom they would vote. Based on models that predict voting behavior in more developed democracies, we contend that Ghanaian voters use retrospective and prospective evaluations when deciding for whom to vote. But unlike models for developed nations that employ aggregate economic indicators as a measure of performance, we instead focus on subjective measures of performance, voters perceptions of parties’ abilities to provide public services. We also look at various characteristics of parties and how voters perceive they will behave if elected, including their trustworthiness, credibility, and ability to make policy. Adjudicating between ethnic and performance voting move prove difficult if voters only think of performance in terms of how well as politician has delivered to their group at the exclusion of other groups. In particular, it requires distinguishing ethnically-targeted patronage from local public services, which is not always easy. For example, building a school in an ethnically homogeneous area can be both a public service and an ethnically-targeted good. Since it is nearly impossible to discriminate these according to any firm set of criteria, we ask voters what factors they think determine how elected officials deliver services to see whether or not they think the parties discriminate based on ethnicity. In addition, a voter’s perceptions of why governments supply services are likely to be more relevant to their voting calculation than any objective criteria employed by politicians. Some scholars assert that voters discount the importance of policy, have low levels of relevant information to judge it, and are unable to monitor performance of parties in office (e.g., Posner 2005). But evidence from Ghana’s election is at odds with this view. Candidates differed over the issues they highlighted in their campaigns as well as the alternatives they proposed precisely to appeal to voters who lacked a clear ethnic choice, particularly candidates without records. Specifically, the NPP campaigned as a more market-oriented party while the NDC campaigned on a more economically populist platform. Voters proved sensitive to policy issues and challengers purposively draw policy differences from incumbents precisely in order to attract swing voters and those who know little about them. Moreover, Ghanaian voters, like those in the US, are likely to require only general policy dispositions, not detailed platforms to know which candidates best represent their self interest (Popkin 1994). 8 We contend that Ghanaians gleaned relevant information not just from ethnicity, but also the attributes of parties, including incumbent performance, policies, and other characteristics to determine their vote. Individuals, we maintain, are most likely to vote on the basis of ethnicity when they have few sources of information about candidates, see few differences on performance and policy, and/or face a vote that includes a co-ethnic. Alternatively, we contend that ethnicity will recede in importance when it cannot discriminate between candidates and/or voters possess information about the policies and performance of candidates. Data and Hypotheses The three hypotheses we test in this paper derive from those which other scholars of Ghanaian politics have suggested account for voting patterns in Ghana’s elections, including our argument about the importance of performance and policy. We list them below. Subsequently, we describe the data we employ to test them. Hypotheses As we described above, there are three basic hypotheses forwarded to explain voting behavior in Ghanaian elections: ethnicity, demographics, and party attributes. • Ethnicity. The basic ethnic hypothesis is that NPP voters are primarily Asante and some (but not all) other Akan, while NDC supporters are Ewe and groups from the north, primarily the Mole-Dagbani. More generally, ethnic hypotheses maintain that both strong in-group attachments and expectations of patronage/policy favoritism should produce ethnic bloc-voting, as opposed to variation in party support within ethnic groups. • Demographics. The demographic hypotheses are that NPP voters are likely to be urban and well-educated, while NDC voters are more likely to live in rural areas and possess low levels of education. Swing voters are also likely to have higher levels of education. • Party Attributes. Our third general hypothesis suggests that party attributes determine for whom voters cast their ballot. We consider a range of mechanisms that could link party attributes with vote choice, such as people’s perceptions about the parties, their views on the importance of electoral institutions, and their evaluations of the incumbent party’s performance. They seek to determine whether voters had a clear evaluation of the stances and performance of each party beyond the ethnic composition of its leadership. Data The data we employ in this paper capture the views of undecided voters compared to supporters of the NDC and NPP come from a pre-election survey we conducted approximately three weeks 9 prior to the election. The poll we conducted was a nation-wide random sample of 2,033 eligible voters. Respondents could take the survey in the language of their choice. 8 Dependent Variable Overall, in the pre-election survey 49%, 32%, and 19% of the respondents said they supported NPP, NDC, and other/undecided, respectively. In the event, the NPP and NDC received 49.7% and 50.2% of the vote in the second-round run-off, respectively. The dependent variable we test in this paper is whom the voter supports for president. We have coded these as Nana AkufoAddo (NPP), John Atta-Mills (NDC), and swing/undecided. The last category is people who gave an answer other than NPP or NDC. The crucial point to understand about these voters and the reason we can place them in one category is that nearly all of them eventually voted for the NDC. The reason we can be fairly certain that almost all swing/undecided voters chose the NDC is that while we understate the NDC’s eventual vote count in our pre-election survey, the poll correctly predicts the percent of the vote the NPP won, 49%. While it is possible that an equal number of NPP and NDC supporters in our survey changed their minds and party-hopped, such an outcome is unlikely given the intensity of the party preferences these voters demonstrated, as the results will show. The table below examines party support within ethnic groups as well as the number of effective ethnic groups within each category of the dependent variables (NPP, NDC, and swing). A number of interesting patterns emerge. First, while the Asante overwhelmingly support the NPP, a non-trivial amount was undecided (13%). Second, approximately 19% of Ewe supported the NPP. Third, even though the NDC candidate was a Fante, 51% of this group supported the NPP. Fourth, both parties are diverse ethnic coalitions. These data present strong evidence against ethnic bloc-voting. Rather, the NPP and the NDC received support from sizable amounts of almost all ethnic groups. While ethnicity is an important correlate of voting, it is far from absolute, especially amongst groups other than the Asante and, to a lesser extent, the Ewe. Party Support by Ethnic Group NPP NDC Swing 67% 16% 17% Asante 83% 4% 13% Fante 51% 28% 22% Other Akan 67% 17% 17% Ewe 19% 59% 22% Ga 32% 48% 20% Mole 34% 49% 17% Other 35% 37% 26% 8 The survey was nationally representative using the registry voters for its sampling frame; conducted in all 10 regions using multi-stage cluster sampling proportionate to size, with random selection of households and respondents. 10 Effective Number* 3.4 5.5 4.5 *Excluding other Explanatory Variables The explanatory variables fall into three categories, ethnic, demographic, and party attributes. Below we describe the questions we employed to solicit answers for each. • Ethnicity. We did not ask respondents directly if ethnicity was the motivating factor that determines whom they supported. Rather, we asked them to self-indentify their ethnic group. We later aggregated these into the larger categories that the table above reports. • Demographics. The two demographic questions we employed are level of education (none, primary, secondary, or tertiary) and whether the respondent lives in a rural or an urban area. • Party attributes. We asked a number of questions to ascertain Ghanaians’ views of the main political parties, the performance of the NPP (the incumbent party), and the importance of the electoral process. The question on performance is, “Thinking about the money it has to spend on services: does the central government do an excellent, good, just fair or poor job with the money it has to spend on services?” We dichotomize the answer as 1 for “excellent” and “good”, and 0 for “just fair” and “poor”. This question attempts to ascertain respondents’ opinions about the performance of the incumbent NPP party at the national level. We call this variable “Quality of Services.” To understand the importance of the electoral process, we asked respondents to prioritize process versus outcome. Our question is, “Thinking about elections in Ghana, which of the following do you prefer? Please choose A or B: A) That your candidate wins, even if the election is not completely honest B) That the election is completely honest, even if your candidate does not win.” We code the variable 0 for A and 1 for B and label it “Candidate Wins.” We also asked a series of questions about respondents’ views of the NPP and the NDC - “Do you feel very close, somewhat close, or not very close to the [NPP/NDC]?” We call this variable “Close NPP/NDC” and code it as 1 for very close or somewhat close and zero for not very close or the respondent choose a different party.9 - “Does the [NPP/NDC] serve the interest of only one group of people, more than one group of people, or haven't you heard enough to say?” We call this variable 9 This question was the follow-on question to “Regardless of how you intend to vote, do you feel close to any particular party?” Respondents could choose one party or no party. 11 “NPP/NDC Exclusive” and code it as 1 for only one group and 0 for more than one group. The purpose of the question is to ascertain if respondents view the party to favor only one group or serve broad interests. - “Do you think the [NPP/NDC] is very likely, somewhat likely, somewhat unlikely, or not at all likely to fulfill their promises if elected, or haven’t you heard enough to say?” We call this variable “NPP/NDC Promises” and code it 1 for very likely and somewhat likely and 0 for somewhat unlikely and very unlikely. - “Do you think the [NPP/NDC] is very trustworthy, somewhat trustworthy, not very trustworthy, or haven't you heard enough to say?” We call this variable “NPP/NDC Trust” and code it 1 for very trustworthy or somewhat trustworthy and 0 for not trustworthy. Results Below we show the results of the tests. We ran a Probit with robust standard errors for each type of respondent: NPP, NDC, and swing/undecided.10 We ran each model twice, once with only ethnic identification and demographic variables, and once with these variables as well as those for policy, performance, and process. The results confirm some of the hypotheses about Ghanaian voters and do not confirm others. They also shed considerable light on why swing voters chose the NDC. The models with only ethnic and demographic variables confirm most hypotheses about patterns of ethnic voting, but not most demographic ones. The Asante and Akan more broadly support the NPP, with the exception of the Fante. The Ewe, Fante, Ga and Mole-Dagbani, by contrast, support the NDC. That the Fante show strong support for the NDC is not surprising as the party’s candidate, John Atta-Mills, was their co-ethnic. Swing voters, by contrast, are less likely to belong to any of the major ethnic groups. These models do not support the demographic hypotheses for the NDC or NPP, but do support those for swing voters. Specifically, NPP supporters are not more likely to have higher levels of education or live in urban areas. Moreover, the evidence that NDC voters are more likely to have lower levels of education and live in rural areas has only weak support. Alternatively, the demographic models support that swing voters are more likely to live in urban areas and have a high level of education. 10 It is important for us to discuss why causality runs from perceptions of parties to vote choice rather than the other direction. For example, it is possible to presume that a voter will trust a party when it picks a candidate he or she likes. Nonetheless, this forces us to ask the question why the voter likes the candidate. Reasonable responses to this question are candidate attributes, such as experience, party affiliation, and/or ethnicity. Hence, since voters need information on the parties/candidates before making their choices, while party perceptions can inform vote choice, the opposite is not plausible. 12 The models with party perceptions and performance show that demographic and ethnic factors are far less important than respondents’ beliefs about the parties, especially if they feel close to a party and if they trust it. They also show that voters possess strong views regarding the integrity of the electoral process versus the outcome of it, and that these beliefs vary systematically by party support. Moreover, adding these variables improves the fit of the models considerably. These results make clear that policy and performance were far more relevant factors in the election than shared ethnicity and/or expectations of patronage. • NPP Supporters. Not surprisingly, NPP supporters are more likely to hold strongly positive views of the NPP and negative ones of the NDC. They also have a strongly positive view of the performance of the existing (NPP) government. NPP voters seem to have no preference for their candidate winning over the fairness of the process. After adding in these variables, the only ethnic and demographic ones that remain significant at the 5% level are NPP supporters are more likely to be Akan and less likely to be Ewe. • NDC Supporters. NDC supporters are close to the mirror image of NPP supporters in their respective views of the party and the NPP’s performance. Unlike in the NPP models, the significance of most ethnic and demographic factors remain significant after accounting for their views on policy and performance, although many of the coefficients recede in magnitude considerably. While Asante, Ewe, Ga, remain significant at the 5% level, Akan and Mole drop to significance at the 10% level, and Fante falls from statistical significance. That Fante is not significant is extremely surprising as this is the sub-ethnic group of the NDC candidate. The most interesting result, however, is that many NDC supporters suggest that they prefer their candidate to win rather than have a fair election. This may result from anxiety at the fairness of the electoral process managed by the incumbent NPP. • Swing Voters. Party perceptions of swing voters shed a substantial amount of light as to why they chose the NDC over the NPP. First, they feel close to neither party. Second, they have no strong opinions about the NDC’s exclusivity, credibility of promises, or trustworthiness. Third, while they do not feel the NPP is exclusive and have no opinion on the credibility of its promises, they do not trust it. In combination, these factors suggest that swing voters chose the NDC because they viewed it as the less inferior of the two parties. Fourth, they care more that the election is honest than that their candidate wins, possibly because they feel close to neither party. Finally, unlike NPP voters who have a positive view of the incumbent government’s performance and NDC voters who possess a negative one, undecided voters showed no consistent pattern in their beliefs on performance. This is not a surprising result since if the group in aggregate had strong perceptions of the performance of the NPP they would have more firm party preferences. Rather, since they are undecided, we should expect that they would not have consistent views about the effectiveness of the existing government. 13 NPP Ethnicity Akan Asante Fante Ewe Ga Mole Demographics Primary Education Secondary Ed. Tertiary Ed. Urban Swing 0.767 (9.07)*** 0.543 (4.26)*** -0.417 (3.73)*** -0.494 (4.16)*** -0.102 (0.90) -0.056 (0.50) 0.411 (3.50)*** 0.051 (0.31) -0.108 (0.65) -0.322 (1.98)** -0.092 (0.59) -0.068 (0.43) -0.623 (6.98)*** -0.835 (4.35)*** 0.407 (3.39)*** 0.571 (5.20)*** 0.284 (2.57)*** 0.263 (2.39)** -0.206 (1.67)* -0.536 (2.04)** 0.186 (0.97) 0.393 (2.57)*** 0.501 (3.32)*** 0.313 (1.92)* -0.345 (3.75)*** -0.153 (1.11) 0.154 (1.23) -0.207 (1.74)* -0.247 (2.03)** -0.294 (2.33)*** -0.265 (2.45)** -0.101 (0.65) 0.047 (0.34) -0.180 (1.36) -0.318 (2.24)** -0.250 (1.80)* 0.148 (1.37) 0.061 (0.60) -0.116 (1.01) 0.006 (0.10) 0.006 (0.04) -0.081 (0.58) -0.254 (1.69)* -0.094 (1.08) -0.179 (1.59) -0.177 (1.68)* -0.151 (1.28) -0.128 (1.96)* 0.000 (0.00) -0.086 (0.61) 0.056 (0.36) -0.184 (1.99)** 0.021 (0.17) 0.140 (1.19) 0.316 (2.49)** 0.143 (2.04)** 0.022 (0.16) 0.215 (1.65)* 0.276 (1.93)* 0.190 (2.35)** -0.104 (0.94) 2033 0.12 -0.166 (1.82)* 0.589 (4.24)*** -1.022 (5.58)*** 1.689 (13.39)*** 0.482 (4.07)*** -0.317 (1.58) -0.667 (5.61)*** 0.626 (4.48)*** -0.361 (3.06)*** 0.856 (5.73)*** -1.116 (5.47)*** 2033 0.59 -0.850 (6.77)*** 2033 0.01 -0.133 (1.60) -0.460 (2.96)*** -1.533 (12.29)*** -1.417 (10.16)*** -0.380 (3.51)*** 0.135 (1.17) 0.025 (0.22) -0.159 (1.57) -0.428 (3.55)*** 0.081 (0.77) 0.041 (0.23) 2033 0.22 Party Attributes Quality of Services Candidate Win Close NPP Close NDC NPP Exclusive NDC Exclusive NPP Promises NDC Promises NPP Trust NDC Trust Constant NDC -0.390 (3.55)*** Observations 2033 Pseudo R Squared 0.12 * p <0.10, ** p < 0.05, *** p <0.01 0.295 (3.39)*** -0.059 (0.43) 1.589 (14.00)*** -1.093 (5.81)*** -0.279 (1.90)* 0.113 (0.77) 0.693 (5.38)*** -0.244 (2.24)** 0.993 (7.09)*** -0.599 (5.63)** -1.171 (5.82)*** 2033 0.57 The table below shows the effect of each of the variables in the full models in the table above on the probability that a survey respondent was an NPP supporter, an NDC supporter, or 14 swing/undecided. The entries in the table that are bold are those variables that are significant at the 5% level. The table makes one point very clear. In general, views about party attributes are more important for voting choices than demography and ethnicity. In addition, the table reinforces that swing voters chose the NDC because they felt marginally less far away from it than they did the NPP and because they distrust the NPP much more than the NDC. NPP NDC Swing Ethnicity Akan Asante Fante Ewe Ga Mole 0.16 0.02 -0.04 -0.12 -0.04 -0.02 -0.05 -0.10 0.05 0.11 0.14 0.09 -0.05 -0.02 0.01 -0.03 -0.05 -0.04 Demographics Primary Education Secondary Ed. Tertiary Ed. Urban 0.01 -0.03 -0.10 -0.04 0.01 -0.02 0.02 -0.04 0.01 0.05 0.06 0.04 Party Attributes Quality of Services Candidate Win Close NPP Close NDC NPP Exclusive NDC Exclusive NPP Promises NDC Promises NPP Trust NDC Trust 0.12 -0.02 0.56 -0.38 -0.11 0.05 0.26 -0.10 0.36 -0.23 -0.04 0.17 -0.20 0.54 0.13 -0.06 -0.18 0.13 -0.09 0.18 -0.03 -0.07 -0.23 -0.18 -0.07 0.03 0.01 -0.03 -0.10 0.02 Conclusion The NDC’s victory in Ghana’s 2008 presidential election came as a surprise to many. Far behind in the polls just weeks before the election, the party managed to eke out a narrow win in the second round of the election. This paper seeks to understand vote choice in that election. Standard theories about elections in sub-Saharan Africa presume that ethnic block voting is pervasive in the region’s elections and that parties serve as little more than a cover for ethnicity. Our results seriously challenge these views. First, supporters of the NDC and NPP formed a diverse range of ethnicities and a number of ethnic groups slit their votes across parties. Second, attributes of parties possess substantially more explanatory power than ethnic group. Third, approximately 20% of the population from a range of ethnic groups remained undecided just weeks before the election. These findings cast serious doubt on the sufficiency of ethnicity to predict vote choice in Ghana’s 2008 election. While there is no question that ethnicity was important for many voters, especially the Asante and the Ewe, it is an insufficient explanation for the electorate as a whole. Instead, party attributes, such as their performance, policies, and voters’ perceptions of them mattered a great deal. Rather than a simple ethnic headcount, party 15 competition for persuadable voters was an important part of Ghana’s 2008 election. Ghana’s 2008 election was far closer to political competition in a democracy than an ethnic census. In closing, it is important to recognize that Ghana’s ethnic composition is typical for sub-Saharan Africa. Very few countries have a small number of ethnic groups, each of which can govern alone. Rather, most countries possess a large number of small ethnic groups, thus rendering multi-ethnic coalitions the norm. Before accepting ethnic bloc voting as a sufficient explanation for vote choice in the region’s electoral contests, we need to understand more about voters’ motivations. 16 References Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain Wacziarg. 2004. “Fractionalization,” Journal of Economic Growth 8(2): 155-194. Anebo, Felix. 2001. “The Ghana 200 Elections: Voter Choice and Electoral Decisions,” African Journal of Political Science 6(1): 69-88. Arriola, Leonardo. Forthcoming. “Ethnicity, Economic Conditions, and Opposition Support: Evidence from Ethiopia’s 2005 elections,” Northeast African Studies 10(2). Bates, Robert. 1974. “Ethnic Competition and Modernization in Contemporary Africa,” Comparative Political Studies 6(4): 457- 477. Bayart, Jean-Francois. 1993. The State in Africa: the Politics of the Bell. New York: Longman. Block, Steven A. 2002. "Political Business Cycles, Democratization, and Economic Reform: The Case of Africa," Journal of Development Economics 67: 205-228. Block, Steven A., Karen Ferree, and Smita Singh. 2003. “Mulitparty Competution, Founding Elections and Political Business Cycles in Africa. Journal of African Economies 12: 444468. Bratton, Michael, Robert Mattes, and E. Gyimah-Boadi. 2005. Public Opinion, Democracy, and Market Reform in Africa. New York: Cambridge University Press. Bratton, Michael and Mwangi Kimenyi. 2008. Voting in Kenya: Putting Ethnicity in Perspective. Afrobarometer Working Paper No. 95. Chabal, Patrick and Jean-Paul Daloz. 1999. Africa Works: Disorder as Political Instrument. Bloomington: Indiana University Press. Chandra, Kanchan. 2004. Why Ethnic Parties Succeed. New York: Cambridge University Press. Cruise O’Brien, Donal. 1971. The Mourides of Senegal: the Political and Economic Organization of an Islamic Brotherhood. Oxford: Clarendon Press. Cruise O’Brien, Donal. 2003. Symbolic Confrontation. London: Hurst. Dawson, Michael. 1994. Behind the Mule: Race and Class in African-American Politics. Princeton: Princeton University Press. Dickson, Eric and Kenneth Scheve. 2006. “Social Identity, Political Speech, and Electoral Competition,” Journal of Theoretical Politics 18(1): 5-39. Downs, Anthony. 1957. An Economic Theory of Democracy, New York: Harper and Row. 17 Fearon, James. 1999. “Electoral Accountability and the Control of Politicians: Selecting Good Types versus Sanctioning Poor Performance,” in Democracy, Accountability, and Representation, Bernard Manin, Adam Przeworski, and Susan Stokes, eds. Cambridge: Cambridge University Press. Fearon, James and David Laitin. 1996. “Explaining Interethnic Cooperation,” American Political Science Review 90(4): 715-735. Ferree, Karen. 2006. “Explaining South Africa’s Racial Census,” Journal of Politics 68(4): 803815. Ferree, Karen. 2008. The Political Origins of South Africa’s Racial Census. Cambridge: Cambridge University Press. Fiorina, Morris. 1981. Retrospective Voting in American Politics. New Haven: Yale University Press. Fridy, Kevin. 2007. “The Elephant, Umbrella, and Quarrelling Cocks: Disaggregating Partisanship in Ghana’s Fourth Republic,” African Affairs 106(423): 281-305. Geertz, Clifford. 1963. Old Societies and New States: the Quest for Modernity in Asia and Africa. New York: Collier-Macmillan. Geertz, Clifford. 1973. The Interpretation of Cultures. New York: Basic Books. Goldberg, Arthur S. 1975. “Discerning a Casual Pattern Among Data on Voting Behavior,” American Political Science Review 60: 913-922. Habyarimana, James, Macartan Humphreys, Daniel Posner, and Jeremy M. Weinstein. 2007. “Does Ethnic Diversity Undermine Public Goods Provision?” American Political Science Review 101(4): 709-725. Hardin, Russell. 1995. One for All: the Logic of Group Conflict. Princeton, NJ: Princeton University Press. Hartwig, Frederick, Jenkins, William R., and Earl M. Temchin. 1980. “Variability in Electoral Behavior: the 1960, 1968, and 1976 Elections,” American Journal of Political Science 24: 553-558. Horowitz, Donald. 1985. Ethnic Groups in Conflict. Berkeley: University of California Press. Huntington, Samuel. 1968. Political Order in Changing Societies. New Haven: Yale University Press. 18 Huntington, Samuel. 1991. The Third Wave: Democratization in the Late Twentieth Century. Norman: University of Oklahoma Press. Jacobson, Gary. 2006. A divider, not a uniter: George W. Bush and the American people, New York: Pearson Longman. Jeffries, Richard. 1998. “The Ghanaian Elections of 1996: Towards the Consolidation of Democracy?” African Affairs 97(387): 189-208. Kaplan, Robert D. 1994. “The Coming Anarchy,” The Atlantic Monthly, February. Kasara, Kimuli. 2007. “Tax Me If You Can: Ethnic Geography, Democracy, and the Taxation of Agriculture in Africa,” American Political Science Review 101: 159-172. Kinder, Donald R. and D. Roderik Kiewiet. 1981. “Sociotropic Politics: The American Case,” British Journal of Political Science 11(1): 129-161. Kramer, Gerald H. 1971. “Short-Term Fluctuations in US Voting Behavior,” American Political Science Review 65: 131-143. Kitschelt, Herbert and Steven we. Wilkinson, eds. 2007. Patrons, Clients, and Policies, New York: Cambridge University Press. Levi, Margaret. 2005. Achieving Government Legitimacy - Good Government. Presented at World Bank Conference on New Frontiers of Social Policy. Arusha, Tanzania, December 12-15, 2005. Levi, Margaret and Audrey Sachs. 2007. Legitimating Beliefs: Sources and Indicators. Afrobarometer Working Paper No. 74 (2007). Levy, Mark R. 1983. “The Methodology and Performance of Election Day Polls,” The Public Opinion Quarterly 47(1): 54-67. Lindberg, Staffan we. 2006. Democracy and Elections in Africa, Baltimore: Johns Hopkins University Press. Lindberg, Staffan and K.C. Morrison. 2005. “Exploring Voter Alignments in Africa: Core and Swing Voters in Ghana,” Journal of Modern African Studies 43(4): 565-586. Lindberg, Staffan and K.C. Morrison. 2008. “Are African Voters Really Ethnic or Clientelistic?: Survey Evidence from Ghana,” Political Science Quarterly 123: 95-122. Lockerbie, Brad. 1991. “The Temporal Pattern of Economic Evaluations and Vote Choice in Senate Elections,” Public Choice 69(3): 279-294. 19 MacKuen, Michael B., Erikson, Robert S. and James A. Stimson. 1992. “Peasant or Bankers? The American Electorate and the US Economy,” American Political Science Review 86(3): 507-611. Mattes, Robert. 1995. The Election Book: Judgement and Choice in South Africa’s 1994 Election. Cape Town: Idasa. Mattes, Robert and Jessica Piombo. 2001. “Opposition Parties and the Voters in South Africa’s General Election of 1999,” Democratization 8(3): 101-128. Miguel, Edward and Mary Kay Gugerty. 2005. “Ethnic Diversity, Social Sanctions, and Public Goods in Kenya,” Journal of Public Economics 89(11): 2325-2368. Morrison, K.C. 2004. “Political Parties in Ghana through Four Republics,” Comparative Politics 36(4): 421-442. Morrison, K.C. and Jae Woo Hong. 2006. “Ghana’s Political Parties: How Ethno/Regional Variations Sustain the National Two-Party System,” Journal of Modern African Studies 44(4): 623-647. Norris, Pippa and Robert Mattes. 2003. “Does Ethnicity Determine Support for the Governing Party?” Afrobarometer Working Paper, no. 26, Harvard University. Nugent, Paul. 2001. “Winners, Losers, and Also Rans: Money, Moral Authority, and Voting Patterns in the Ghana 2000 Election,” African Affairs 100(400): 405-428. Page, Benjamin we. and Calvin Jones. 1979. “Reciprocal Effects of Policy Preferences, Party Loyalties and the Vote,” American Political Science Review 73: 1071-1089. Platteau, Jean-Philippe. 1994. “Behind the Market Stage: Where Real Societies Exist (Parts we and II),” Journal of Development Studies April/July: 533-577, 753-817. Popkin, Samuel. 1991. The Reasoning Voter. Chicago: Chicago University Press. Posner, Daniel N. and David J. Simon. 2002. “Economic Conditions and Incumbent Support in Africa’s New Democracies: Evidence From Zambia,” Comparative Political Studies 35(3): 313-336. Posner, Daniel. 2004. “Measuring Ethnic Fractionalization in Africa,” American Journal of Political Science 48(4): 849-863. Posner, Daniel. 2005. Institutions and Ethnic Politics in Africa, Cambridge: Cambridge University Press. Posner, Daniel. 2007. “Political Sources of Ethnic Identification in Africa” (with Benn Eifert and Edward Miguel) Afrobarometer Working Paper No. 89 (2007). 20 Rabushka, Alvin and Kenneth Shepsle. 1972. Politics in Plural Societies: a Theory of Democratic Instability. Columbus, OH: Merrill. Riker, William H. and Peter C. Ordeshook. 1973. An Introduction to Positive Political Theory. Englewood Cliffs, NJ: Prentice-Hall. Schaffer, Frederic Charles. 2000. Democracy in Translation: Understanding Politics in an Unfamiliar Culture. Ithaca, NY: Cornell University Press. Snyder, Jack L. 2000. From Voting to Violence: Democratization and Nationalist Conflict. New York: WW Norton. Stasavage, David. 2005. “Democracy and Education Spending in Africa,” American Journal of Political Science 49(2): 343-358. Stokes, Donald E. 1963. “Spatial Models of Party Competitions,” American Political Science Review 57(2): 368-377. Taylor, Michael. 1988. “Rationality and Revolutionary Collective Action,” in Rationality and Revolution, Michael Taylor, ed. New York: Cambridge University Press. Tucker, Joshua. 2006. Regional Economic Voting: Russia, Poland, Hungary, Slovakia, and the Czech Republic, 1990-1999. Cambridge: Cambridge University Press. Wantchekon, Leonard. 2003. “Clientelism and Voting Behavior: Evidence from a Field Experiment in Benin.” World Politics 55(3): 399-422. Wilkinson, Steven. 2004. Votes and Violence: Electoral Competition and Ethnic Riots in India. Cambridge: Cambridge University Press. Youde, Jeremy. 2005. “Economics and Government Popularity in Ghana,” Electoral Studies 24: 1-16. 21 View publication stats