By Dr Parkie Mbozi
I HAVE asked for my views on the recently published opinion polls around the August 2021 polls, specifically on the controversial and contradictory results between the two major polls by the Political Science Association of Zambia (PSAZ) and University of Cape Town’s Afrobarometer.
The PSAZ poll claims that the ruling Patriotic Front and its presidential candidate Edgar Lungu will carry the day with a far wider margin in 2021 compared to 2016 and 2015. The poll by Afrobarometer, to the contrary, predicts a marginal win (of about 5 points) for the opposition United Party for National Development (UPND) and its presidential candidate Hakainde Hichilema.
The questions are: 1. why the contradictory results; 2. How credible are the two polls against the benchmarks or cannons of conducting a credible and universally acceptable survey? In this article I share my two pennyworth on what constitutes a credible poll generally and how these principles resonate with the two polls.
To be on the same page, an ‘opinion poll or survey is simply defined a ““a human research survey of public opinion from a particular sample. Opinion polls are usually designed to represent the opinions of a population by conducting a series of questions and then extrapolating generalities in ratio or within confidence intervals.”
Almost every post 1991 election in Zambia has attracted opinion polls. The credibility of previous polls has been in question, unfortunately, due to wrong predictions. In 2011, for instance, a poll conducted Neo Simutanyi and team wrongly predicted a Rupiah Banda and MMD victory. This year two polls have hit the news headlines over the last month, with ‘wildly different predictions for Zambia’s election’ (according to the Mail & Guardian newspaper of South Africa). The poll by the newly formed PSAZ was the first and biggest, in terms of sample size. It was reportedly conducted by PSAZ in in collaboration with two “international think-tanks”, an entity calling itself Faraline of the United Kingdom and another, Media Theory of USA. It should be pointed out that the profiles of the two institutions and the claimed ‘experts’ has been questioned.
Locally the PSAZ study was conducted by a three-some team of local researchers comprising Dr. Masauso Chirwa as the Principal Investigator and Mr. Joe Nabwa and Mr. Aaron Siwale. Based on media reports (as its reports is not available of its website), the purpose of the poll was to ascertain which presidential candidate was likely to win this year’s election. The poll was conducted in five provinces – Copperbelt, Luapula, Eastern, Southern, and Western provinces. It involved a sample size of 59,628 respondents.
It is reported that according to this particular poll, no candidate would get an outright 50 + 1 % of the vote but Lungu is favoured to get a ‘solid’ 44.5%, followed by his arch-rival Hichilema who trails at 30.33%. In distant third is Socialist Part President Fred M’membe (2.61%), then MMD President Nevers Mumba (2.13%,), and Democrratic Party candidate Harry Kalaba coming in at 1.93%. The rest of the candidates follow with insignificant numbers. The study further asserts that Lungu scored highest in Eastern, Luapula and Copperbelt while Hichilema scored highly in Southern and Western Provinces. A whopping 15% were reported to have either declined to respond or remain undecided for various reasons.
In contrast, the survey by Afrobarometer, tells a very different story about who is likely to win the polls and which party has gained or lost in popularity since 2017. The Mail & Guardian reports that, “Afrobarometer conducts public attitude surveys in more than 30 African countries, and has been doing so since 1999. Its data is used by the likes of the World Bank and the Mo Ibrahim Foundation, and it has partnered with some of Africa’s most respected research institutions including South Africa’s Institute for Justice and Reconciliation; Benin’s the Institute for Empirical Research in Political Economy, Ghana’s Center for Democratic Development; and Kenya’s Institute for Development Studies.”
In December 2020, Afrobarometer interviewed a representative sample of 1,200 Zambians of voting age drawn from all of the 10 provinces. It asked them a wide range of questions from corruption and climate change to taxation and traditional leaders. Afrobarometer also asked about current voting preferences. In response to the question ‘If general elections were held tomorrow, which party’s candidate would you vote for?’, only 22.9% of respondents said that they would vote for the PF (down from 44.8% in Afrobarometer’s 2017 survey). 25.2% said that they would vote for the opposition, the United Party for National Development. Close to half (45.6%) of respondents either refused to answer or said they did not know.
In an analysis of the data, Jeremy Seekings and Hangala Siachiwena — researchers with the University of Cape Town’s Institute for Democracy, Citizenship and Public Policy in Africa wrote that “This erosion of (PF) support seems to be widespread, among both urban and rural voters. A large minority of voters — much larger than in previous surveys — declined to declare who they would vote for if elections were held. Most of these ‘undeclared’ voters are dissatisfied with the country’s economic performance under the present government.”
My first view about the credibility of the two polls is of a general nature. It is anchored on the validity or acceptability of any survey based on the following fundamentals: resonance of the results to known realities (immediate past and present); methodology adopted, in particular sample size, sampling method vis representativeness to the population; reliability of the data (internal validity), data collection and analysis methods, and the setting (or environment in which the study was conducted).
Starting with sampling, the principle of sampling is that the selected individuals must look like the population from which they are drawn. Therefore where and how the sample was drawn is a significant factor.
If we take the survey by (PSAZ) for starters, in terms of resonance of the results to ‘known knowledge’ or ’logic’, I am prompted to ask, “what has changed positively and so significantly in favour of the PF that Lungu can open a 15-point lead to UPND’s Hichilema over the last five years?’ Remember, in 2016 Lungu beat Hichilema by 100,530 votes or 50.35% (ECL) to 47.67%. (HH).. He narrowly survived a re-run by 0.35% (having polled 50.35%). In sharp contrast, the Afrobarometer study of December 2020 in fact reported a decline in PF popularity across the board (rural and urban). It reports that the PF popularity (or people who support it) declined to 22.9% in December 2020 from 44.8% in the 2017 survey conducted by the same organisation.
Similarly, an online poll conducted by News Diggers newspaper on readers rating of Lungu’s performance on corruption and fight against Covid 19 rated him horrendously low. For instance, on the fight against corruption, from a sample of 1,648, 81% said pathetic, only 6% rated him as ‘good’ or ‘excellent’. On the fight against Covid 19, 75% rated the government performance as ‘worrisome’ or ‘pathetic’ and only 13% said ‘good’ or ‘excellent’. In addition, as individuals who live with the population, we are subjected to a lot of anecdotal evidence of the waning support for the PF government from among the general population, owing to a number of factors. So, simply on account of the results vis-à-vis logic, the credibility of the study becomes highly questionable.
In terms of representativeness of the sample in the PSAZ study, the first question is, why was the sample, of close to 60,000, drawn from five instead of all the 10 provinces? Worse, why three known PF strongholds and only two UPND strongholds? Second, how was the sample drawn? Was it stratified to accommodate the different population demographics (.e. by age, gender, etc)? Was it systematically and randomly picked so that every household and person had an equal chance to be included, as with the Afrobarometer study?
To be acceptable, the sample should have been drawn randomly (‘blindly’) from all the 10 provinces and randomly sampled constituencies and wards across. This is what it takes for a study to attain what we call external validity or generalizability. Sampling from five, not all 10 provinces, renders the sample purposive. In this case the results do not qualify for generalization across the country but rather should be confirmed to the provinces from which it was sampled. From this perspective, the Afrobarometer study is more credible, having been drawn from all the 10 provinces and 103 constituencies. The smallness of sample size for this study, however, also gives its critics some ammunition to punch holes in it.
The next considerations are the source of the data and setting of the study or environment in which it was conducted. Regarding setting, with the extent of polarization and intolerance for the opposition in the country, most people are too terrified to show open support for the opposition. Hence, even durinf a campaign period like now opposition supporters are too terrified to wear their party regalia.
Without an elaboration on under what conditions the interviews were conducted, as in the Afrobarometer study, it is hard to take the results at face value. A reader calling himself Razar drives the point, “Hahaha! Who can say the truth in opinion polls in Africa unless they want to be followed by shushushu. Just wait for August bane that’s when the true opinions of Zambians will be seen.”
Finally, in communication we use the term ‘source credibility’ to refer mainly to whether the source of a report (as in this case) can be said to possess the values that can qualify it as trusted and believable. Mail & Guardian sums up the concerns about the PSAZ study, “As Cheeseman observes, the so-called international think-tanks involved in producing the survey — London-based Farraline Public Relations and Washington, DC-based Media Theory — are actually both public relations firms that specialise in reputation management. And the Political Science Association of Zambia is a newly formed organisation with no track record of conducting surveys on this scale.”
Given the fact that the two firms that supported the poll are PR (‘image building’) firms, those who are claiming that the figures they churned were cooked to make the PF look popular, for the sake of what is known as ’momentum politics’ (chalola umwela) can be excused. After all, have not such fake polls been done before? Only time will tell. After all, we are less than 25 days to go. Last week I talked about the shortest verse in Zambian politics: Banda wept!
Political Science Association of Zambia
The Political Science Association of Zambia is the learned, scholarly society which exists to develop and promote the study of politics. It is the leading professional organisation in its field in Zambia, with a wide membership including academics in political science and current affairs, theorists and practitioners, policy makers, researchers and students in higher education.
The author is a researcher and scholar with the Institute of Economic and Social Research, University of Zambia. He is reachable on [email protected]