Abstract
What determines the vote choice in Ghana? This paper examines this question using a data set on voting behaviour generated from a nationwide survey of 2042 voters from the ten regions of Ghana based on six national elections held between 1992 and 2012. The results based on descriptive analyses and logistic regressions indicate that political parties tackle substantive issues during African elections, but generally voice them through valence appeals rather than by staking out distinct positions. Further, the results show that campaign message and the ethnicity of the presidential candidate matter in the vote choice. It is worth noting, however, that of the seventeen factors identified by respondents as most important, ethnicity ranked only 14th. This finding suggests that the declarative information of respondents is not consistent with their real intentions. Overall, the findings of the study show that both patronage politics and programmatic policies matter in winning elections. Accordingly, politicians need to focus on enhancing the credentials they bring to the political arena and improve their image in terms of how best they are able to achieve the public good.
Introduction
In the last few decades, many developing countries, and African countries in particular, have embarked on democratization of their political system. This is associated in a large measure with the political economy literature that suggested that the absence of democratic institutions accounted for the poor growth in developing countries (Hall and Jones, 1999; Rodrik and Wacziarg, 2005). Indeed, many studies do report that the quality of government could not be improved as long as authoritarian regimes remained in power, because of their tendency towards corruption and clientelist behaviour. By the end of the 1980s, most of Sub-Saharan Africa’s (SSA) authoritarian regimes were in dire economic conditions and, this helped to give a boost to the democratic reforms that swept the entire region. For example, in 1983 6% and 59% of SSA countries were rated as free and not free respectively, and by 2013 this improved to 20% as free and 41% as not free (Freedom House, 2014).
The remarkable drive towards political institutionalization is also evidenced by the failed attempts of leaders to amend constitutions in Zambia, Nigeria and Malawi (Posner and Young, 2007). By the end of the 1990s, many of Africa’s ruling elite did not only recognize but also accepted the maxim that political legitimacy required a popular vote in multiparty elections (Bratton, 2013). The trend towards democratization in Ghana and Africa provides opportunity to engage in an in-depth study of voting behaviour. Obviously, the Africa of today is very different from that of three decades ago when Horowitz (1985) described African elections as nothing more than ‘ethnic census’.
In this paper, democracy is defined as a political system characterized by popular participation, competition for executive office, and providing an institutional check on how power is used (Minier, 1998). Voting behaviour is defined as a set of personal electoral activities, including participation in electoral campaigns, turnout at the polls, and choosing who to vote for (Bratton, 2013). In this paper, however, the focus is on the determinants of the vote choice. The core institution of modern liberal democracy, whereby the right of the people to self-government can be exercised, is competitive and participatory elections (Lindberg and Morrison, 2008). The extent to which elections fulfil this mission is to a significant extent dependent on citizens’ rationale for how they behave at the polls.
Despite the importance of elections as instruments of democracy, experimental and independent surveys of voters’ behaviour and rationale in multiparty elections in new and transitional democracies remain scarce. This study represents an effort in that direction. The choice of Ghana is appropriate as it is the first country in the region to pass the threshold for the consolidation of democracy (Huntington, 1991). Also, looking at its economy, politics and culture, it could be considered as an Upper Middle Class African country (Bossuroy, 2011), which makes an interesting case for the study of politics in Africa. The study seeks to examine whether the deepening of democracy in Ghana has resulted in a shift from clientelist or patronage politics to programmatic policy formulation.
It is important to note, however, that other studies have examined voting behaviour in Ghana, for example: Anebo (1997, 2001) and Lindberg and Morison (2008) looked at the 1996 and 2000 elections for 690 voters; Fridy (2007) studied the 2004 elections; and Ichino and Nathan (2013) and Harding (2015) examined the 2004 and 2008 elections. This study differs from the previous studies as it uses data over a longer time frame (1992, 1996, 2000, 2004, 2008 and 2012) and also a larger dataset of 2,042 respondents from all the ten regions in the country; therefore, the results are more likely to represent the actual situation on the ground. The period under study (1992–2012) reflects all the elections held after the coming into effect of the 1992 Fourth Republican Constitution, to examine presence or otherwise of swing voters and particular trends, if any, over the study period. This is important because each of the two major parties, New Patriotic Party (NPP) and National Democratic Congress (NDC), received at least 40% of the vote in the last four elections.
In the sections that follow, the theoretical and empirical bases of voting behaviour are presented. The methodology is then described and the data collected are analysed. The findings of the study are discussed, policy implications given and a concluding remark is offered.
Theoretical and empirical literature
Generally, three main theoretical perspectives have been used to explain voting behaviour in the literature: Sociological theory or the Columbia School (Converse, 1944; Lazarsfield et al., 1944; Lipset, 1959, 1960); Psychosocial or the Michigan School (Campbell et al., 1960); and the Rational Choice perspective (Downs, 1957; Fiorina, 1981, 2002; Key and Cummings, 1966). The sociological perspective shows that long-held factors (social characteristics) such as socio-economic variables, religion, and location are key determinants of voting behaviour (Lazarsfeld et al., 1944). The assumption here is that the majority of people vote according to their original political predisposition. The problem with this theory, however, is that if vote choice was determined solely by stable sociological factors, election results will remain unchanged for a long time.
The Michigan School or the psychosocial model is an updated version or complementary to the Columbian model, as it helps to explain how electoral change occurs. It provides a framework that combines sociological and psychological factors (so-called psychosocial approach) to explain the vote choice. The psychological model relates to the need of individuals to belong to or be identified with a group or the collective, in this case the political party. The central concept of the Michigan School of thought is political affiliation or partisanship, where party identification is loosely understood as a socio-psychological product of family and social group ties (Dalton, 2013). Party identification is therefore expected to shape the evaluation of candidates, issues, and the expected capacity of parties to solve problems (Erdmann, 2007a).
The rational theory of voting focuses on the fact that human beings behave the same way in both political and economic settings. In sum, the model predicts that voters derive the most utility from the candidate or party closest to them on some ideological or policy continuum. On one hand, it establishes a direct analogy between consumers and voters, and on the other, between enterprises and political parties. It can be inferred that voters will seek to maximize the utility of their vote as the parties act to maximize electoral gains obtained from their electoral campaigns. From the preceding discussion, it can be said that the sociological model centres on the influence of social actors, while the psychosocial assumes party identification as the key factor in determining voting behaviour, and the rational theory concentrates on rationality and choice.
Many empirical studies have been used to support the various theoretical perspectives, which support the view that voting behaviour is complex such that no one model can fully explain completely the determinants of voting behaviour (Converse, 1944; Roth, 1998). This is associated with the fact that the clarity of the domestic institutional context influences the relationship between economic perceptions and the vote choice (Devries et al., 2011). Vicente (2014), for example, claims that crude vote buying is a frequent practice during election time in many parts in Africa (with evidence from São Tomé and Príncipe). It is not surprising therefore that there is a pessimistic view of African elections as little more than contests in corruption and ethnic loyalty (Harding, 2015). Erdmann (2007a) in a study of Zambia reports that ethnicity provides the basic social cleavage for voting behaviour, and the formation of parties and politicians are more likely to direct resources (both public and private) to their co-ethnics.
Challenging the ethnic census view of African elections, Conroy-Krutz (2013) used Uganda to show that goods distribution and ethnicity become less important as constituents gain more political information. This finding is consistent with Dendere’s (2013) observation that the Africa of today is very different from what was described as ethnic census by Horowitz (1985) and Chandra (2007a, 2007b). In a study of Zimbabwe, Dendere (2013) reports that voters make their choice based on the policy and economic preferences underlying a preferred ideology. Similarly, Andrews and Inman (2009) using the 2005 Round three (3) Afrobarometer survey of seven African countries, report that while ethnic ties affect vote choice in Africa retrospective evaluations (evaluative voting) of economic performance are equally important .
In light of the recent findings in many African countries, many researchers suggest that ethnicity should not be abandoned as a determinant of the vote choice, but the way it is looked at must obviously be refined. In support of the Andrews and Inman (2009) study, Lindberg and Morrison (2008) report that clientelist and ethnic predisposed voting are minor features of the Ghanaian electorate. Lindberg and Morrison find that voters’ evaluative behaviour is related to the actual or expected performance of the candidates and not politicians or governments’ ability to provide private and public goods to their constituents. In a related study, André and Mesemplé-Somps (2011) demonstrate that public transfers seem irrational as the ruling government invests more in opposition districts (especially, where the leading opposition members are very powerful) to avoid political agitation originating in these districts.
Interesting though is Andrews and Inman’s (2009) assertion that clientelism persists today in roughly the same form as it did in the 1970s and 1980s (Chabal and Daloz, 1999; Wantchekon, 2003). The reason being that politicians gain and maintain their positions by dispensing public goods to their supporters and special treatment to co-ethnics (Erdmann 2007b). This accounts for the high inequality between politicians and the citizens in the region. However, Harding (2015) reports that when the provision of public goods (in this case roads) can be attributed to political action, then, especially in the rural areas, it is likely to be affected by electoral support. This is not surprising as studies (Bossuroy, 2011; Kopřiva and Varvažovská, 2011; Vicente, 2014) do show that rural voters differ in their understanding of democracy, policy preferences, and access to independent media. Conroy-Kutz (2013) shows that the informational environment impacts on the vote choice as the utility of the rural voter, for example, lies partly in the information that demographics provide about candidates. Beard (2006) and McClurg et al. (2013) demonstrate that gender and religion do matter in explaining voting behaviour in Africa
What makes voters in Ghana decide? Is it related to evaluative rationales such as characteristics and accomplishments of candidates, performance of government, and policy platforms of parties (Minns, 1984) or non-evaluative factors such as political affiliation, gift from candidates, ethnic or family ties? There has been virtually no long-term tracking of these fleeting issues, and therefore the current survey is dedicated to the collection of data about voting behaviour and reasons that respondents give for voting for one party or the other. This study thus complements other studies on Ghana, but it differs from the previous studies as it uses data over a longer time frame (1992–2012) and also a larger dataset of 2042 respondents from all the ten regions in the country. Obviously, seeking to identify trends is consistent with the political economy literature that links voting behaviour to economic business cycles.
Methodology
The study relies on triangulation of methods involving primary and secondary data. Creswell (2013) and Silverman (2005) support a combination of techniques and suggest that social scientists have to abandon the spurious choice between qualitative and quantitative data, and make use of the most valuable features of each. Therefore to obtain fairly accurate information, this study uses both qualitative and quantitative data. The study’s analysis builds on a data set on voting behaviour and rationales generated from a survey of 2042 voters in the 1992 to 2012 elections in Ghana.
Sample and data collection
The study is designed mainly as a representative nationwide sample survey. The first stage of sampling included all the 10 regions of Ghana. In each region, one-tenth of the districts were selected using a combination of probability and non-probability sampling techniques. A district in the regional capitals of all the 10 regions was purposively selected to represent the urban population of the sample and all major tribes in the region. The other districts outside the regional capital were randomly selected using a simple random technique.
Lindberg and Morison (2008) looked at the 1996 and 2000 elections for 690 voters; Ichino and Nathan (2013) studied the 2005 and 2008 Afrobarometer data (focus on Brong Ahafo Region); and Bossuroy (2011) examined the 2004 elections. Building on these prior studies, this study considered 2,042 voters from all the ten regions of the country to give ample scope for detailed sub-group analysis, and thereby making it possible to draw reasonable inferences about the population, at regional and national levels. The regional distribution of the sample was proportional to the results of the 2010 National Population and Housing Census (see Table 1). A total of 22 research assistants trained to administer the structured questionnaire from the 22 selected districts were recruited. In each of the selected districts, the research assistants visited public places such as offices, markets and schools to interview any person who voted over the years and who was willing to complete the questionnaire. Quantitative data alone was considered as inadequate, so it was decided to triangulate quantitative data with qualitative data by conducting some focus group discussions (FGDs).
Sample distribution of respondents.
Source: Survey Data (2014).
The FGDs allowed the study to capture verbatim statements from participants. Considering time and cost of the study only two FGDs were organized from urban (GIMPA) and rural (Afram Plains) areas. In all, 16 people with diverse background were carefully selected to participate in the FGDs (eight participants were selected for each group). To qualify to be selected as a participant of the FGDs, the person should have voted in at least four presidential elections.
Data analysis
Field data were edited, coded and analysed. The Statistical Package for Social Sciences (SPSS 20) was used in analysing the data. Also logistic regression was employed to examine the determinants of voting behaviour over the study period.
Findings of the study
The discussion of the findings of the study is divided into two main parts: descriptive and logistic regression results.
Descriptive results
The respondents are considered as rational and responsible actors who are knowledgeable about the reasons for their voting behaviour. The first set of substantive questions measures respondents’ opinions regarding the level of importance of 17 factors or variables that influence their vote. In each case, the respondents had to indicate how important (1 least important and 5 most important) they rated each of these factors. The summary mean scores for the total sample are presented in Figure 1.
Mean score on factors that influence voting behaviour.
Clearly, non-evaluative factors such as ethnicity, gender, regional backgrounds, religious backgrounds of the candidate and gifts from the candidates were rated below the mean score of three points. However, all the evaluative factors were rated above the mean score of three points. These factors include educational background, campaign message and ability to fight corruption. The respondents (2,042) were further asked to select the most important factor that determined their vote choice from the seventeen variables and the results are reported in Table 2.
The most Important factor that influenced voters’ choice.
Source: Survey Data, 2014.
The five most important factors of the vote choice are: campaign message (20.5%); human relations of the presidential candidate (15.5%); educational policy (12.8%); personality of candidate (9.9%); and performance of the ruling government (8.4%). The least important factors are: regional background of candidate (0.4%); gifts from candidate (0.7%); gender of candidate (0.8); and religious affiliation (1.0%). The findings from the respondents indicate that voters are more evaluative rather than being clientelist or ethnic heads, as suggested by Chandra (2007b). These initial findings are consistent with Lindberg and Morrison’s (2008) evidence that evaluative voting far exceeds non-evaluative voting in Ghana.
The important role of the campaign message is supported by Godbout and Belanger (2007) and Nordin (2014), who claim that a party’s campaign message helps to define issues about the economy and explain to the electorate which candidate is better able to manage the challenges confronting the citizenry and the economy at large. This observation therefore does not support studies that show lack of position taking on issues or an absence of substantive electoral debate in Africa (Bleck and van der Walle, 2013). Obviously, more informed voters are better at voting for their most preferred politicians and parties. The access to information has been used to explain the difference between the urban and rural vote in many studies. Thus, the rural voters who are usually poorer and less educated tend to vote for the left wing party (NDC), while the more educated and urban voters tend to vote for the right wing party (NPP). This supports a study of the Czech Republic by Kopřiva and Varvažovská (2011) who show that the less educated rural voters are more likely to vote for the left and the highly informed, urban and educated are more likely to vote for the right wing party.
Evidence from informal focus group discussions indicates that the two major political parties, the New Patriotic Party (NPP) and the National Democratic Congress (NDC) usually castigate the policies and programmes of the other party while in opposition. For example, the NPP’s campaign message on health in 2000 was intended to replace out-of pocket payments at the point of health service use. Similarly, the NDC campaigned on one-time premium in 2008 and described the NPP’s annual premium payment as denial of access to the health service for the poor. Many analysts claim that the NPP’s slogan ‘Hwe wa sitana mu na nu tu aba pa’, which simply means that ‘examine your life and vote appropriately’ (if your condition is good then vote for the ruling party but if not then vote for change) was successful in convincing voters to vote the party into power in 2000 (Boafo-Arthur, 2006; Ayee, 2011). Similarly, the NDC’s slogan of the need for the country to move forward was critical in helping the NDC to retain power in 2012. These results provide support to the Bleck and van der Walle (2013) observation that political parties tackle substantive issues during African elections, but generally voice them through valence appeals rather than by staking out distinct positions. In other words, the basic difference between the two major parties is more of how to achieve the objective rather than an ideological difference. In 2012, for example, the key campaign issue was on education, with the NDC boasting of increasing educational access and the NPP promising free education if it won power.
The respondents were also asked directly to indicate whether their vote choice is more evaluative or otherwise, and the results are presented on a regional basis in Table 3. On average, 95% of respondents claim that their vote choice is driven by evaluative criteria.
Summary of evaluative and non-evaluative voting.
Source: Survey Data (2014).
These findings, however, are not consistent with the evidence provided by respondents (field data) that show only about 23% of the respondents did change their voting behaviour (swing voters over the study period). A swing voter in this study is defined as a voter who is not exclusively committed to a party or candidate. Additionally, in the case of the Ashanti and Volta regions, the NPP and NDC do consistently receive more than 70% respectively, independent of which government is in power (Fridy, 2012). The responses on evaluative versus non-evaluative criteria therefore might be more indicative of what it ought to be rather than what it is. This gives credence to the findings of Bratton and Kimenyi’s (2008) study of Kenya, who report that over 90% of the respondents deny voting on ethnic grounds, and yet the majority of respondents believe that ethnicity plays a major role in Kenyan elections. The implication is that in many cases the declarative information of the respondents may be biased. The respondents fail to say what they really think or they may have a biased perception of their own views (Bossuroy, 2011). To deal with this bias, the respondents were asked to rate the importance or otherwise of the identified issues without categorizing them as evaluative or otherwise.
A comparison of the ranking of the factors and analysing the factors individually, provide more information about the factors (See Table 4). For example, only 14 respondents indicated that getting gifts from candidates and 27 respondents indicated ethnic background were the most important when they ranked the factors together. However, when they were asked to discuss the factors individually, the responses were very different; on the issue of gifts, the figure increased from 14 to 651 respondents, 1,009 for campaign message, 1,032 for education, 932 for health policy, and 918 for employment provision. Overall, rating the factors individually and collectively suggests that voting behaviour is mediated by many complex and dynamic factors (Goodman and Murray, 2007) as shown in Table 4.
Most important factors based on group and individual factors.
Source: Field Data (2014).
A further analysis of the voting patterns of card and non-card holding members also indicates that over 80% of card-holding members and about 70% of non-card holding members voted for the same party over the study period. This is relevant in light of the fact that 57% of respondents reported that they do identify with a political party. Also important to this discussion is the observation that 366 (17.6%) respondents each identified themselves with both the NDC and the NPP respectively. This helps to reduce any bias in the results reported. Further, the reliability of the results is seen in the consistency of the field data collected with the results declared by the Election Commission over the 1992–2012 period (See Table 5). Though there are slight variations in the numbers (percentages), it is interesting to note that the winner in all the cases (field data) is the same as in the numbers provided by the National Electoral Commission (NEC).
Trend of voting by respondents.
Field data showing winner of election.
These findings are indicative of the fact that sociological and psychosocial actors are still relevant in determining who wins elections in Ghana. The voting pattern since 2000 shows a little over 40% votes for each of the two major political parties. No doubt, Ghana has so far had two presidential run-offs in 2000 and 2008 elections, all of which led to a change of government. The question is whether there is a pattern in the trend of voting, which can be explained by the behaviour of the voters to date: two terms of the NDC (1993–2000), after which the NPP came to power (2001–2008), and then the NDC again (2009–to date). Does the data indicate that there will be a run-off in 2016 and that the NPP (the current major opposition party) will win the elections in 2016? It is expected that the findings of this research will help provide answers or at the least provide clues to this question.
Logistic regression results
This study used primary data which consisted of demographic variables and factors that influence voting behaviour. The variables are specified and defined in Table 6.
Variables of interest and their definitions.
Dependent variable
The outcome variables were categorized into respondents who voted for NDC and NPP due to the fact that they formed about 98% of the votes cast by respondents in the dataset. NDC was coded as 1 and NPP as 0.
Data processing
Some continuous variables were categorized before starting the analysis. Age of respondents was categorized into two age groups: up to 30 and above 30 years of age. The breakpoint was selected according to the distribution of the data. Level of education was categorized into secondary and post-secondary. Location was grouped into urban versus rural, and religion into Christians and non-Christians. Furthermore, the response of the rating factors for presidential and parliamentary candidates was categorized into important and less important (Table 6).Two levels of analyses – bivariate and multivariate – were conducted. Variable choice was determined by the focus of the paper. The bivariate analysis examined the relationship between the selected demographic characteristics and attitudes of the respondents on their voting behavior (Cohen and Chaffee, (2012). This was done through the calculation of the crude odd ratios using the bivariate logistic regression method. The multivariate analysis involves the selection of variables found to be significantly associated with voting behaviour. The Wald’s test was used to test the overall significance of the variables selected by the model at the end of each model fitting. The essence of the logistic regression is in its flexibility, and ability to adjust for many explanatory factors and controlling for many confounders at the same time.
Bivariate analysis
Bivariate results are reported in Table 7. The results show that the rural population, older people, females, singles, non-Christians and Ewes are more likely to vote for the NDC (see Table 7). Those above 30 years are nearly one and half times more likely to vote for the NDC than for the NPP with a multiplicative factor of 1.323. Females are also 73% more likely to vote for the NDC. The singles also vote by a multiplicative factor of 0.781 or are78% more likely to vote for the NDC. Non-Christians are one and half times more likely to vote for the NDC and the Ewes are eight times more likely to vote for the NDC than the Ashanti’s, and rural voters are about one and half times more likely to vote for the NDC. In a related study, Leigh (2005) examined the effect of individual, local and national factors on voting behaviour in Australia between 1966 and 2001 and reported that the poor and those unmarried were more likely to vote for the left wing party.
Bivariate analysis.
Note: *p<0.05.
Multivariate analysis
Controlling for other determinants of voting behaviour, the results show that although the variables maintained their sign of coefficients, only ethnicity remained significant (Table 8) with a slightly lower multiplicative factor (7. 82). It is interesting to note that even when the location of respondents was controlled for, the ethnic coefficient was significant for both rural and urban voters though the two groups differed in terms of the religious affiliation and marital status (results not shown here). This is consistent with the NEC results reported over the past 20 years. Also, on the ranking of the 17 factors that determine the vote choice; ethnicity ranks 14th (27 respondents) which is only higher than three other factors (regional background, religious background and sex of presidential candidate) as shown earlier in Table 3. The results also show that significant predictors of the vote choice for presidential candidates are personality, human relation, performance of ruling party, gifts from candidate, ability to provide employment and health policy of the candidate.
Results of multivariate analysis.
Note: *p<0.05.
There is nearly a 1.5 times advantage for having a personable character (Table 9). In the 2012 elections, for example, there was the perception that the NDC presidential candidate who won the elections was more likeable than the NPP candidate, who was considered elitist.
Results of multivariate analysis for rating factors of presidential candidates associated with voting behaviour (Model 2).
Note: *p<0.05, **p<0.01, ***p<0.001.
Interestingly, the regional background of the presidential candidate and corruption are not significant. The results also show that the determinants of vote choice are a mix of both evaluative and non-evaluative factors.
Implications and conclusion
This study sought to identify determinants of voting behaviour as well as quantify their impact with a view to identifying the relative importance of each variable. The results are interesting and could provide insights to stakeholders (political parties and citizens) as to the key determinants of the vote choice in Ghana. Overall, the findings of the study are consistent with the view that factors that explain why and how voters vote the way they do are complex, and that no one study can identify all factors. The bivariate analyses show that the rural voters, those below 30 years, females, singles, non-Christians and Ewes are more likely to vote for the NDC (See Table 8). However, level of education, income and employment status had no significance on voting behaviour.
The multivariate results indicate that ethnicity and human relations of the presidential candidate, educational background, and gifts from candidate, ability to provide jobs and health policy are key determinants of voting behaviour factors. The results of the multivariate results to some extent are supportive of Bossuroy’s (2011) claim that ethnic actors still play a critical role in Ghanaian politics. These findings have to be considered with some caution because many of the studies do indicate that ethnic cues are more pronounced in information scarce societies, which is not supported by our data with more than 60% of respondents having degrees. Indeed, only 237 respondents (11%) do not have a high school qualification.
Unlike Lindberg and Morrison (208), however, this study is unable to conclude that ethnicity is not critical in Ghanaian elections. What this study is able to infer is that over time what decides who wins elections in Ghana are the 20–25% so-called swing voters who continually change their votes for the party they support. Accordingly, in the context of studying voter behaviour in emerging democracies, this study is a significant research that reinforces the rational choice perspective as nearly 25% of the respondents changed their voting pattern over the study period. However, non-evaluative criteria issues in terms of the ethnicity, gender, religious affiliation and gifts from candidates or the party, do matter. On the other hand, the 70–80% of the respondents who voted consistently for the same party, independent of their performance in or outside of government, shows the important role of sociological and psychosocial factors as determinants of voting behaviour. This consistency in voting independent of the performance of the ruling government suggests strong bondage between politics and the citizenry (Branton, 2004). The question is whether the bondage is through mere ethnicity or political ideology or some other form of cultural or social identification. The study’s findings are suggestive of the fact that these could be interacting in various ways; further research is needed to clarify this.
Interesting to our study is that the pattern of vote is consistent with the national elections, thus the majority vote on all occasions is in alignment with the winning party (1992–2012), and more importantly the number of respondents who indicated their political affiliation (NDC and NPP) are the same (366). What is clear from the findings of the study is that the vote choice is not always rational; sometimes the sociological, and in certain cases the psychosocial factors dominate. The question is which factors dominate when?; further research is needed to clarify this. The findings of the study indicate that Ghanaian voters use multiple cues as well as their own complex social and ideological identities in deciding who to vote for, which is supportive of the instrumentalist view that both performance evaluations and ethnicity contribute to voter decision making (Weghorst and Lindberg, 2011, 2013; Dahl, 1995).
Moreover, while supporters of each party do not fit a clear ethnic profile, they possess strong beliefs about the parties. It is important to note that, for example, the Volta Region has consistently voted for the NDC (over 80% of the populace) over the last six elections, but it is only between the 1992 and 2000 elections that the presidential candidate was a co-ethnic. In the 2008 and 2012 elections, they won by the same margin but the candidates were from the Central and Northern Regions of the country respectively. These results are not surprising because Ottoway (1999) argues that the absence of ideological or programmatic differences between parties in Africa means that ethnicity becomes the dominant cleavage in African politics. At the same time, it is worth noting the idea that ethnicity acts as a bonding agent between politicians and voters (Knoesen, 2009). The findings of the current study, however, do not support the view that ethnicity provides a better fit for the politician and the citizenry than political parties do in other established democracies. As can be observed in the descriptive analyses, campaign message which is related to the political ideology of the party was ranked the most important (see Table 2). This is supportive of Piszczek and Kaminski’s (2010) assertion that voting is not always rational, and that people do not always act in their best economic interests.
Finally, the main implication of the current study is that the role of ethnicity cannot be ignored, and the overall campaign message of the parties in delivering both collective and clientele goods matter in who wins elections in Ghana. It is necessary for politicians to focus on enhancing the credentials they bring to the political arena and improve their image in terms of how best they are able to achieve the public interest. As the results of the study show, public goods provision in education and health are critical in determining the vote choice. Thus, who wins the elections may not depend so much on the ethnicity of the candidate but rather on other political factors driven by the ability to persuade or send credible signals to different constituents about the ideology of party, effectiveness of campaigns and convincing constituents (especially swing voters) that they understand their needs.
Using data based on six rounds of elections between 1992 and 2012, our study contributes to the literature on voting behaviour in Ghana through descriptive and statistical analyses to show that elections in Ghana cannot be described as ethnic census. The findings suggest that the deepening democracy has meant that elections in Ghana resemble those of major democracies in the world, where both patronage politics and programmatic policies matter in determining the vote choice. Whether and when patronage politics or programmatic policies dominate is an empirical matter, which further studies are needed to shed more light on. Additionally, further studies are needed on the 20–25% of the voters who constantly change their vote (swing voters) as to who they are, where they are located and whether they are driven by different factors.
Footnotes
Funding
The researchers are grateful to the management of the Ghana Institute of Management and Public Administration for the support provided for data collection.