Briefing Papers | BP103: Bratton, Michael and Eldred Masunungure. Voting Intentions in Zimbabwe: A Margin of Terror?. 2012
Voting Intentions in Zimbabwe: A Margin of Terror?
Afrobarometer Briefing Paper No. 103 VOTING INTENTIONS IN ZIMBABWE: A MARGIN OF TERROR? by Michael Bratton and Eldred Masunungure August 2012 Intermittently over the past decade, researchers have taken the political pulse of the general public in Zimbabwe. Public opinion surveys provide information on what ordinary citizens are thinking about the issues of the day. Among the most anticipated survey results are expressed party preferences and voting intentions. At any given time, Zimbabweans are understandably eager to know how their fellow citizens would vote if an el ection were held tomorrow. A debate has arisen, however, about the reliability of survey research under conditions of widespread political violence. Skeptics are right to ask whether citizens feel free enough to offer honest answers to sensitive survey questions if, by so doing, they risk losing life, limb or property. In the memorable words of the late Masipula Sithole, a margin of terror can distort the profile of public opinion. Concerns on this score surfaced in reaction to a report entitled Political Change and New Politics in Zimbabwe ied byssu Freedom House on August 20, 2012. The report used survey data gathered in June 2012 to indicate that if a presidential election were held tomorrow, Robert Mugabe, the candidate of the Zimbabwe National Union Patriotic Front (ZANU-PF), would garner 31 percent of the votes as compared to 19 percent for Morgan Tsvangirai, the candidate of the Movement for Democratic Change-Tsvangirai (MDC-T). This result represented a profound reversal of fortunes for both parties. It was met by a prompt and carefully worded response from MDC-T: ³We note that a lot of people interviewed refused to disclose their political preferences. This is obviously for fear of intimidation and the violence they have been subjected to by ZANU-PF and its military junta. The margin of terror fundamentally impugns the conclusion that can be derived from this repwowwrt .md(c.co.zw, Aug 22, 2012). The present briefing paper offers an alternative account of current voting intentions in Zimbabwe. 1 The analysis rests on data from the latest Afrobarometer survey of July 2012. 1 The Afrobarometer is an independent, non-partisan social science research project organized as an African-led international collaboration. It conducts regular surveys on topics related to democracy and governance in more than 30 African countries. The Mass Public Opinion Institute (MPOI) conducted fieldwork for the Round 5 Afrobarometer survey in Zimbabwe during July 16-30, 2012. A sample of 2400 adults of voting age was selected using a stratified, multistage, area design with probability proportional to size and randomization at every stage. Interviewers were carefully trained and closely supervised to conduct interviews in the language of the respondents choice. Respondents were given assurances of anonymity and confidentiality and provided informed consent before proceeding with an interview. 1 --------------------------------------- 2 We make four points. First, using a standard survey question, we show that the preferences of the electorate are evenly split between the two main political parties. Second, we justify this result in terms of the relative accuracy of our survey methodology. Third, we move beyond mere description of voting intentions to test explanations, including the margin of terror. Finally, recognizing that many survey respondents refuse to answer, we estimate how these reticent voters might actually vote, thus allowing us to speculatively revise observed survey results by taking political fear into account. Voting Intentions, July 2012 Figure 1 shows how Zimbabweans said they would vote in late July 2012 if a presidential election were held tomAccororrdoiwng. to these overt responses, the two major parties are in a statistical dead heat: ZANU-PF would garner 32 percent of the vote and MDC-T would receive 31 percent. A survey with 2400 cases contains a margin of sampling error of plus or minus 2 percentage points. Therefore, actual voting intentions lie somewhere within a range of 30-34 percent for ZANU-PF and 29-33 percent for MDC-T. As a result, either party could have been slightly ahead and, if any election had been held in July 2012, the outcome would have been too 2 close to call. And a second-round run-off presidential election would have been necessary. 2 There are other reasons why caution should be used in using Afrobarometer results to predict the outcome of elections in Zimbabwe. The survey question refers to a hypothetical event ± an election held ³tomorrow± rather than an actual contest. The date of the next general election remains highly uncertain (but no later than November 2013) and much change can occur in public preferences between now and then. Finally, any election forecast should be based only on persons likely to vote (i.e. excluding ³abstainers who say they won¶t vote); this adjustment -PF: MDwouldC -Tch ratioang fe romt he ZANU 32:31 to 35:34. 2 Figure 1: Voting Intentions Zimbabwe, July 2012 50 40 3231 30 22 20 9 10 3 12 0 ZANU-PFMDC-TMDCOthersDon't KnowRefusedWon't vote Would vote for this party's presidential candidate Note: Actual percentage distribution of intended votes should exclude abstainers´ (won¶t vote) --------------------------------------- 3 Figure 1 also confirms that Zimbabwe possesses a two-party system. No minor political party, including MDC (Ncube), ZAPU or Mavambo-Kusile-Dawn, can boast more than one percent support from the electorate. But these parties are relevant to political outcomes in the event of an extremely close election, when they might hold a balance of power. Perhaps the most important result in Figure 1 is that nearly a quarter of all citizens refused to answer the voting intentions question. They often correctly claimed that their vote is a private matter, a secret that does not have to be revealed. The partisan preferences of this reticent group are therefore unknown. Yet such voters are sufficiently numerous (22 percent) that their support could potentially swing an election decisively toward one party or another, even to the point of negating the need for a second-round runoff. The unrevealed preferences of reticent voters ± especially if engendered by fear of intimidation or violence ± are therefore critical to understanding the state of play in partisan politics in Zimbabwe. A Closing Partisan Gap? The distribution of voting intentions in July 2012 marks a radical shift from patterns observed in earlier surveys. Figure 2 shows results for the same standard question on voting intentions as asked in four Afrobarometer surveys between 2005 and 2012. The trends reveal a recent resurgence in overt support for ZANU-PF and concomitant erosion in citizen willingness to openly identify with MDC-T. Shortly after the formation of Zimbabwes Inclusive Government (IG) in February 2009, MDC-T apparently enjoyed a massive edge in expressed popular preferences over ZANU-PF (57 percent versus 10 percent). Since that time the partisan gap seems to have closed, at least in terms of what Zimbabweans are willing to confide to a survey research team. 3 Figure 2: Voting Intentions by Political Party Zimbabwe, 2005-2012 60 57 50 39 40 36 32 3231 29 30 2224 22 20 18 1010 0 OctoberMayOctoberJuly 20052006200720082009201020112012 ZANU-PFMDC-TRefused to answer Percentage who indicate that they would vote for a candidate of this party if a presidential election were held tomorrow´ Source: Afrobarometer Rounds 3, 4, 4.5 and 5. --------------------------------------- 4 Two caveats are worth considering about survey results from earlier periods. First, peak levels of expressed support for MDC-T were probably inflated by a mood of euphoria (or at least relief) following the signing of a power-sharing agreement and the inauguration of a coalition government. The MDC-T faced high popular expectations in 2009 of what could be achieved by the IG, hopes that had certainly dissipated by mid-2012. Second, earlier samples did not always penetrate land resettlement schemes and large-scale commercial farms because, for security reasons, these areas were deemed pol-go ziotnesica. Bl noy the time of 2012 surveys, this sampling bias was corrected. Nevertheless, past surveys may have overestimated MDC-T support and underestimated ZANU-PF support, especially among persons who had benefited from fast-track land redistribution. We continue wonder, however, whether political fear is infecting results. In May 2009, Zimbabweans welcomed the return of relative peace in the aftermath of devastating election violence. But by October 2010, the power-sharing regime had reached a stalemate and the major political parties had fallen back into polarized political camps. ZANU-PF began to push hard for early elections. By July 2012, the party mobilized its apparatus of selective patronage and political terror, not only in the countryside, but also in urban townships. Under these circumstances it was risky for citizens to openly align with any particular party, a development that may well have had the effect of suppressing free expression. An Artifact of Method? Survey researchers can probably agree that the MDC-T¶s fortunes havente rly weanec d. But to what extent? In July 2012, the Afrobarometer (AB) found substantially higher levels of overt support for the MDC-T presidential candidate (31 percent) than did Freedom House (FH) just one month earlier (19 percent). We note that almost all of this difference is attributable to the fact that far more people refused to answer the voting intentions question in the FH survey (36 percent) than in the AB survey (22 percent). Indeed the proportion of reticent respondents in the FH survey was larger than the proportions expressing an intention to vote for either major party! We contend that survey methods may account for some of these observed differences. Methods diverge across AB and FH surveys in at least three ways: 1. Sampling. The AB sample is consistent with ZimStat¶s official 2011 population projections for Zimbabwe. Relative to this standard, however, FH over-sampled Harare and Manicaland and under-sampled Bulawayo and Matabeleland. Given that the northeast of the country has been the epicenter of recent political violence, this bias likely contributes to an increase in fear-induced refusals. 2. Clustering. The FH sample (N=1198) is half the size of the AB sample (N=2400). Providing it is a random sample, however, this fact alone should not impugn its accuracy. But the FH sample is clustered more tightly (12 interviews in each of 100 sampling areas) than the AB sample (8 interviews in each of 300 sampling areas). Thus, if only a few of the areas sampled by FH happened to fall in a given party¶s stronghold, r esults could be distorted. 3. Questionnaire. Researchers know that the content and order of survey questions can affect respondent answers. Compared to the AB questionnaire, the FH questionnaire contained long strings of items on political violence before it posed the voting intentions question. Attention in the interview to this disturbing subject may well have gripped respondents with suspicion and fear and primed them to take refuge in cautious answers. 4 --------------------------------------- 5 4. Interviewers. An experienced field team from the Mass Public Opinion Institute (MPOI) was available to the Afrobarometer but not to Freedom House. Instead, a novice team was mounted for the FH survey who had never previously worked for MPOI or never before done a survey interview. MPOI supervisors suspected that fear was prevalent among the interviewers. To protect themselves, interviewers may have lacked confidence in correctly asking sensitive questions or provided respondents with safe passage to neutral responses. As artifacts of faulty methods, the results reported by Freedom House probably over-estimate ZANU-PF support and under-estimate MDC-T support. Because we place more confidence in Afrobarometer results, we use them in the analysis that follows. Explanatory Factors Research should not stop short at description. It is instructive to report descriptive statistics (like levels of party support) and to observe changes in these indicators over time as we have done above. But the real power of survey methods lies in statistical explanation. We want to know why political parties in Zimbabwe are neck-and-neck in expressed voting intentions and why the partisan gap has apparently closed over time. As a starting point, we pose two simple hypotheses to explain ZANU-PF¶s resurgence and MDC- T¶s decliThe fne. irst is the positive effect of improved government performance under the IG. The second is the negative effect of political fear induced by state-sponsored violence. 1. Government Performance. On balance, the overall public mood, while gloomier than in 2009, is still somewhat upbeat. Slightly more people think the country is moving in the right direction (48 percent) rather than the wrong direcrtlyi, mon (ore pe43 ople percent). Simila think that, over the previous year, the country¶s economic condition became better (35 percent) rather than worse (23 percent). And considerably more people expect the country¶s economic condition over the next year to improve (52 percent) rather than decline (20 percent). Turning to the perceived performance of the IG, the record is more mixed. For example, Zimbabweans are split right down the middle on whether the coalition government has managed the economy well (49 per barcentdly) o (50 per cent). Much depends on the policy sector. People give the government high marks for addressing educational needs (71 percent say they are handling this task well) and improving basic health services (69 percent). But they give it low marks on creating jobs (12 percent say well), providing a reliable supply of electricity (14 percent), and fighting corruption (15 percent). And, while a large minority (41 percent) credits the IG with resolving political violence, a larger majority (56 percent) thinks it has done ba dly on this score. The question is: which political party gets credit or blame for perceived government performance? Several proxy indicators are available. First, when asked how much they trust political parties after the Inclusive Government, 46 percent of the pu- blic say they trust ZANU PF somewhat or a lot, while exactl-Ty ( hal50 perf cetnrt)ust. Sec MDond,C when asked to evaluate the job performance of the two top political leaders, 58 percent approve of the work done over the previous year by President Mugabe compared to 66 percent for Prime Minister Tsvangirai. 5 --------------------------------------- 6 Thus, both parties enjoy a measure of institutional trust and both leaders are appraised positively for performance in office. These results in July 2012 stand in stark contrast to earlier Afrobarometer surveys in which Tsvangirai benefited from wide margins of approval over Mugabe (82 versus 24 percent in May 2009). In welcoming the IG¶s delivery of economic stability and relative peace (at least as compared to 2008), the general public now seems to grant almost as much credit to ZANU-PF as to MDC-T. We test this hypothesis below. Table 1 provides insight into the divergent moods of each party¶s supporters with regard to government performance. ZANU-PF loyalists are more likely to say the country is headed in the right direction (63 versus 40 percent) and that the economy is performing well (65 versus 38 percent). ZANU-PF supporters even take pleasure in improved education services. Thus ZANU- PF supporters express consistent satisfaction with the IG¶s economic and social policies and, accordingly, stand ready to reward their favored party at the polls. At the same time, MDC-T supporters ± possibly because they once harbored high expectations ± are disappointed with what the IG has achieved and seemingly blame their party for not having achieved more. 2. Political Fear. Zimbabweans remain deeply concerned about political violence. Fully 88 percent think that multiparty competition often or always leads to violent conflict. This figure represents an increase since 2009 (80 percent) and is far higher than any other country in the Afrobarometer, including even Nigeria (74 percent) and Kenya (76 percent). In addition, some 63 percent of Zimbabweans say that, during election campaigns, they personally fear becoming a victim of political intimidation or violence. They also worry about freedom of expression. Fully 89 percent asoftsener ort that a lpewayoplse ha³ve to be careful of what they say about politics. Again, this is the highest rate ever recorded by the AB in Zimbabwe or anywhere else in Africa. And some 61 percent consider that the government often or always silencioen pas opportiessi tor their supp orters. 6 acententions by Perceived Government PerformnTable 1: Voting I Zimbabwe July 2012 Would vote Would vote Refused to for for answer/ ZANU-PFMDC-TDont know Country is going in the right direction 63%40%44% Government is managing the economy 65%38%43% well Government is addressing educational 79%63%71% needs well Percentage of survey respondents who agree with the above statements --------------------------------------- 7 Furthermore, survey respondents may censor themselves in an interview situation. Despite MPOI¶s best efforts to introduce its interviewers as independent observers, almost half of the sample (44 percent) sees the survey as sponsored by a governmental agency associated with Zimbabwes pa-stratte, ay nd therefore potentially with ZANU-PF. Only slightly more respondents (47 percent) attribute sponsorship to a non-governmental or non-political entity. Thus, it is necessary to explore whether these (mis-) perceptions systematically shape expressed voting intentions. If fear is a factor, then voting intention results (see Figure 1) probably represent a comparison between overt (meaning openly declared) ZANU-PF support and core (meaning deeply committed) MDC-T support. These days, we suggest that only core MDC-T supporters remain willing to express overt support for their party. Instead, the incentives of political fear drive the party¶s more passive followerr s trute pro heferide encets, heat mi inimum by refusing to answer a sensitive voting question. At the same time, would-be voters have every incentive to display political fealty to ZANU-PF, either by expressing sincere support for the party or by falsely claiming a political allegiance that they do not genuinely feel. The effects of these incentives are reflected in Table 2, which shows an uneven distribution of political fear among the supporters of different political parties. More than three quarters (78 percent) of persons overtly aligned with MDC-T say they fear intimidation or violence during election campaigns compared to fewer than half (47 percent) of declared ZANU-PF supporters. It is also striking that more than twice as many MDC-T than ZANU-PF loyalists consider that opposition parties and their supporters are often or always silenced by government (81 versus 37 percent). We fully expect that this gulf in fear-induced political vulnerability will affect overt expressions of partisan support, a hypothesis that we now test. 7 eTable 2: Voting Intentions by Political Far Zimbabwe July 2012 Would vote Would vote Refused to for for answer/ ZANU-PFMDC-TDont know Fear intimidation or violence during 47%78%66% elections Think that opposition parties are silenced 37%81%66% by government Perceive survey as sponsored by a 58%38%38% government agency Percentage of survey respondents who express the above sentiments --------------------------------------- 8 Explaining Voting Intentions So, what determines an individual¶s stated intention to vote for a particular party? Following a political science convention, we examine intentions to vote for the party of the incumbent president, in this case ZANU-PF. Because the object of explanation is a binary choice (vote for Mugabe¶s party or not), the appropriate method is logist ic regression (see Table 3). In Table 3, six explanatory factors are clustered into two groups: government performance and political fear. All indicators are popular perceptions as measured in the Afrobarometer Round 5 survey of July 2012. It turns out that all the selected indicators are statistically significant forecasts of a vote for ZANU-PF. In other words, both government performance and political fear are at work in shaping voting intentions in Zimbabwe. The positive signs on the performance indicators strongly suggest that ZANU-PF is gaining electoral advantage from its association with the IG. Correctly or not, some voters apparently attribute the country¶s right direction, the economy¶s good management, and improved delivery of educational services to ZANU-PF. They do so even though the MDC-T has shared the leadership of the coalition government and headed the Ministries responsible for economic management and education. At minimum, this result suggests that the MDC-T has failed to get out a message, or to convince diehard supporters of the old regime, of its own contributions to better governance under the IG. 8 Table 3: Predicted Probability of a Vote for ZANU-PF Zimbabwe, July 2012 B Statistical Marginal Coefficient SignificanceEffects GOVERNMENT PERFORMANCE Country Right Direction.571.000***+12% Economy Well Managed.265.001**+17% Education Services Improved.213.007**+13% POLITICAL FEAR Fear of Election Violence-.447.000***-29% Opposition Seen as Silenced-.718.000***-46% Government Seen as Survey Sponsor.924.000***+20% Constant.525.153 The B column shows logistic regression coefficients. All predictor variables are statistically significant. The Marginal Effects column shows the predicted probability that survey respondents would vote for ZANU-PF --------------------------------------- 9 When it comes to political fear, we uncover a couple of anomalies. First, if ZANU-PF were successful in intimidating reluctant voters to express support, then we would expect a statistically positive relationship (a plus sign) between an individual¶s expressions of fear and his stated intentions to vote for the incumbent president¶s party. But we do not. The relationship between a persons expectofat viictons imization by violence during election campaigns is negatively (a minus sign) related to an intended vote for ZANU-PF. In other words, even though people admit that they are politically fearful, they still decline to tell an interviewer that they will support the former ruling party. In short, they feel fear but they are able (courageously) to overcome it. Second, the perception that the government silences opponents also yields a negative statistical effect on voting for ZANU-PF. In other words, the more that a voter perceives the suppression of political speech, the less likely she will wish to return ZANU-PF to power. The question becomes whether perceiving repression is the same as feeling political fear. To be sure, the silencing of opposition does seem to induce people to be careful when talking about politics (r =.312***). But, once again, it does not persuade them to change their stated voting intentions. The only way that political fear has the predicted positive sign is when people perceive government sponsorship of the survey. If they fear such government surveillance, they are more likely to express a voting intention for ZANU-PF. This result can be interpreted as a manifestation of protective self-censorship. Moreover, the last column of Table 3 indicates that this form of fear has larger effects on voting intentions than any single aspect of government performance. For example, a person who thinks the country is moving in the right direction is 12 percentage points more likely to come out openly for ZANU-PF. But a respondent who fears government surveillance is 20 percentage points more likely to do so. In other words, the fear factor is substantively large enough to seriously distort, not only survey results, but potentially an actual election. If we are interested in rooting out the effects of political fear ± in other words, the margin of terror ± then we need to control for (mis-)perceptions of government survey sponsorship. The next section attempts this task. How Will Reticents V ote? One of the persistent mysteries about public opinion in Zimbabwe is the political allegiance of persons who refuse to divulge their voting intentions. In a democracy, citizens are entirely at liberty to hold their vote secret. For one reason or another, including most probably fear, they do not wish to reveal their partisan preferences. Secretive voters are a distinct group and should not be confused with those who say they will not vote ("abstainers") and those who have not made up their minds who to vote for (the "undecided"). It would be very useful if we could estimate the latent partisanship of this reticent group. To do so, we propose a method to discern how much the overt vote is inflated by fearful perceptions of government sponsorship. We already know that perceived government sponsorship inclines voters to declare allegiance to ZANU-PF. Whereas 32 percent do so in the sample as a whole, 43 percent do so if they think the government is behind the survey (See Table 4). By contrast, if respondents think that a non-government agency is behind the survey, they tend to lean towards MDC-T (38 percent versus 31 percent in the sample as a whole). Let us assume that the "true" breakdown of partisan sentiments is unencumbered by fear of government sponsorship. If so, then the distribution of intended votes in the column headed "see 9 --------------------------------------- 10 non-government sponsor" in Table 4 is the one that comes closest to reality. In other words, in a perfect world where no citizen feared government sponsorship of surveys, 22 percent would opt for ZANU-PF, 38 percent for MDC-T, and 5 percent for other parties. But the denominator for these percentage calculations includes "abstainers" and the "undecided." We take their opinions at face value, that is, that they mean what they say (and that they are not retreating into wont vote and dont know as alter Fonrat the pue frorposems of of reticence). this analysis, however, we make the conservative decision to exclude these cases. This correction allows us to consider only those respondents who express an overt partisan preference (N=734). Using raw figures for the ratio of MDC-T:ZANU-PF:other parties we can see that 429:248:57 produces the following corrected partisan breakdown: 58 percent for MDC-T: 34 percent for ZANU-PF: 8 percent for other parties Applying this ratio to the 523 persons in the "reticent" group, we infer that it produces an additional 303 intended (but secretive) votes for MDC-T, 178 for ZANU-PF, and 42 for other parties. Adding these "unrevealed" votes to overt voting intentions we get the following: MDC-T 742 + 303 = 1045 49 percent ZANU-PF 774 + 178 = 952 45 percent Others 100 + 42 = 142 7 percent 10 arlerceve Government Survey Sponsor by Political PtyTabe 4: Pid Zimbabwe July 2012 See See Non-Total Government Government SponsorSponsor ZANU-PF447248775 (43%)(22%)(32%) MDC-T259429741 (25%)(38%)(31%) Other Parties3957100 (4%)(5%)(4%) Percentage of survey respondents who report the above perceptions. Column totals to not add to 100 percent because the perceptions of reticent, abstainer´ and undecided groups are not shown. Row totals include figurs´ (esnot for s hodonwn).t know --------------------------------------- 11 This simulated result (see Figure 3) is interesting in several ways: 1. MDC-T moves slightly ahead of ZANU-PF in terms of voting intentions in July 2012. 2. Partisan support is still quite evenly balanced but now beyond the survey's margin of sampling error. 3. By excluding "abstainers" (9 percent), the result correctly refers only to those who actually plan to vote. But it still excludes voters who might return from the diaspora to cast a ballot. 4. Figure 3 almost replicates the officially reported results of the first round of the last presidential election in March 2008 (Tsvangirai 48 percent, Mugabe 43 percent). 5. The analysis implies that, if voting intentions do not change, Zimbabwe can expect another close election in 2013. 11 Figure 3: Adjusted Voting Intentions of Intended Voters Including the Estimated Vote of the "Reticent" Zimbabwe, July 2012 49 4550 40 30 20 7 10 0 ZANU-PFMDC-TOther Parties Would vote for this party's presidential candidate 3HUFHQWDJHFDOFXODWLRQVH[FOXGHWKH³XQGHFLGHG´GRQ¶WNQRZDQG³DEVWDLQHUV´ZRQ¶WYRWH --------------------------------------- 12 Conclusions 1. In terms of overt (declared) voting intentions in July 2012, the two major parties in Zimbabwe are in a statistical dead heat. 2. Zimbabwean voters make voting choices on the basis of both the positive achievements of the Inclusive Government and the negative sanctions of intimidation and violence. 3. While ZANU-PF seems to derive more benefit from recent government performance than it probably deserves, MDC-T seems to derive less. One reason may be that the former has paid more attention to grassroots organization and mass communications than the latter. While ZANU-PF has invested resources to rebuild its party machine and mobilize its political base, MDC-T has relied too heavily on a strategy of expecting political credit for improved service delivery. 4. A large proportion of voters, however, are reticent to express a partisan preference. Political fear, while pervasive, may be less effective than its purveyors might wish since many would-be voters in Zimbabwe seem able to overcome it. 5. That being said, fear of government sponsorship of social surveys is a major reason why some people say they support ZANU-PF when they do not. Adding estimates for reticent voters increases the proportion of citizens intending to vote for MDC-T. 7. But this adjustment does not give any party a decisive edge. Any future election in Zimbabwe remains too close to call. No political party in Zimbabwe can afford to be complacent about an easy electoral victory. The Afrobarometer survey of popular voting intentions in July 2012 strongly suggests that, at present, neither ZANU-PF nor MDC-T could secure the presidency without a second-round run-off election. About the Afrobarometer The Afrobarometer is a collaborative survey research project conducted by a network of social scientists from more than 30 African countries. The Center for Democratic Development (CDD- Ghana) provides overall project direction. At the sub-regional level, the following Core Partners coordinate survey and other activities: the Institute for Democracy in South Africa (Idasa), the Institute for Empirical Research in Political Economy (IREEP) in Benin, and the Institute for Development Studies (IDS) at the University of Nairobi, Kenya. Michigan State University and the University of Cape Town provide analytic and technical support services. The Afrobarometer Network gratefully acknowledges generous contributions from the UK Department for International Development (DfID), the Swedish International Development Agency (SIDA), the United States Agency for International Development (USAID), the World Bank, and the Mo Ibrahim Foundation. Grants from these donors support research, capacity building and outreach activities in Afrobarometer Rounds 5 and 6, 2010-15. For more information, see: www.afrobarometer.org 12
Author(s) Bratton, Michael and Eldred Masunungure