Afrobarometer voter preference data and actual election results
| Monday October 6, 2014 15:55
What might these findings tell us about the upcoming elections? To answer this question, I asked myself: how closely did previous Afrobarometer results correspond to actual election results? When the Afrobarometer team asked respondents how they would vote “if elections were held tomorrow” in 2008, 69% said that they would vote for the BDP, 13% for the BNF, 10% for either the BCP or BAM, and 1% for the BPP. In 2008, a similar portion of respondents (5%) said that they would not vote, but only 2% refused to answer the question and 1% said that they did not know (http://www.afrobarometer.org/files/documents/summary_results/bot_r4_sor.pdf). Elections took place roughly a year later. The BDP received 53.3% of the vote, the BNF got 21.9%, the BCP and BAM received 21.5%, the BPP 1.4%, and 1.9% voted for independent candidates. The comparisons are not ideal, for the previous Afrobarometer survey was conducted in September/October 2008, a full year ahead of the elections in October 2009.
The 2005 Afrobarometer survey was conducted not quite a year after the 2004 elections, in May - June. In response to the “if elections were held tomorrow” question, 53% of respondents said that they would vote for the BDP, 24% for the BNF, 9% for the BCP, and 1% for the BIP (http://www.afrobarometer.org/files/documents/summary_results/bot_r3_sor.pdf). The percentage reporting that they would not vote was 4%. Only 2% refused to answer, but 6% said that they did not know. By comparison, in the 2004 elections, the BDP received 51.7% of the vote, the BNF got 26.1%, the BCP 16.6%, BAM 2.8%, BPP 1.9%, and others got 0.8%.
These previous Afrobarometer surveys have over-reported support for the BDP and under-reported support for opposition parties. The 2005 data, collected a year after an election, were pretty close for the BDP and BNF, but further from the mark for the still relatively new BCP. The 2008 data, collected roughly a year before an election, were seriously distorted in favor of the BDP. That survey was conducted only a few months after Ian Khama ascended to the presidency and may have captured a “honeymoon” effect. It is also possible that vote preferences fluctuate more in the run-up to an election in response to sometimes dramatic developments during the campaign than they do immediately after an election, when things may calm down. One likely source of bias towards the BDP is the fact that more than half of the respondents believe that the survey was being conducted on behalf of either the government or the President’s office (67% in 2005 and 71% in 2008). The preliminary results for 2014 do not include data on who respondents thought had commissioned the survey.
In the sub-sample of Afrobarometer data that I’ve considered, the proportion of respondents who say that they do not know how they would vote ranges from 6% in 2005 to 1% in 2008. In 2005, the percentage who told the Afrobarometer team that they would support the BCP or that they did not know how they would vote (9% + 6% = 15%) approaches the support actually received by the BCP in 2004 (16.6%). Does this reflect the “softness” of support for a then still relatively new party, perhaps with some second-guessing about the BCP’s prospects after it elected only one MP? Or does it reflect a shyness to admit support for such a party to a stranger?
In 2014, the share of respondents that told the Afrobarometer team that they did not know how they would vote was not lower than in 2005 (3%). On the other hand, the proportion refusing to answer the question jumped to 6% from just 2% in both 2008 and 2005. While the spike in refusals may reflect a general decline in trust captured in this round (http://www.afrobarometer.org/files/documents/press_release/bot_r6_pr_trust_25Sep14.pdf), respondents who lean toward the opposition would seem more likely to conceal their vote preferences than would BDP supporters.
Since the 2014 Afrobarometer is only a few months from an election rather than a full year, one might expect the results to be relatively close to what we will see on Election Day. It probably overestimated support for the BDP and almost certainly underestimated support for the various opposition parties. The comparison of the voter preferences measured in the 2008 Afrobarometer with the outcome of the 2009 elections underscores the potential for volatility in voter preferences in Botswana. Although the 2014 survey was conducted only a few months prior to the elections, these last few months represent the most intense period of the campaign. High profile, emotionally charged events have occurred that are likely to influence voter behavior. The Afrobarometer survey is a well-executed survey and its findings should be taken seriously, but it was conducted too far in advance of the elections to offer a very reliable gauge of what to expect on October 24th.
*Professor Amy Poteete is a political scientist with extensive knowledge of Botswana politics. She is based in Canada.