Zambia’s doggy opinion polls

Zambia’s doggy opinion polls

How NOT to rig an election: Zambia’s dodgy opinion poll

By Nic Cheeseman

Opinion polls play an important role in elections. They don’t just tell us who might win – they may also create “bandwagoning” effects, where voters who perceive that their favourite candidate cannot win switch their allegiance to the favourite in order to be part of the winning team … and secure greater access to resources as a result. At the same time, if the media and election observers believe that a certain party is destined to win they are more likely to accept the outcome of the elections, even if there is some evidence of irregularities.
Because of this, elections in Africa often see multiple opinion polls being produced, with some balanced and credible polls being conducted by internationally respected organisations such as the Afrobarometer, and a raft of much less credible polls being funded by political parties to create the impression that they are ahead. These fake polls can be very damaging, confusing journalists and observers, creating false expectations among rival sets of supporters, and increasing the prospects for a controversial and contested outcome. The Zambian election – scheduled for August 12 – is the latest to follow this trend, but with a difference.
Instead of having a credible organisation do an opinion poll, leading to a survey that looks credible in all methodological respects, the government’s allies have completely bungled the process, leaving little doubt that the poll is dodgy. In fact, the poll is so remarkably bad that, like the Ugandan electoral commission sharing an “approved” election results sheet that clearly revealed fraud, it will serve as an “emperor’s new clothes” moment, spotlighting attempted electoral manipulation.

The challenge

Aware that the latest Afrobarometer survey conducted at the start of this year makes for grim reading for the government, the ruling party’s allies set about putting together a “blockbuster” poll to blow it out of the water.
The Afrobarometer poll was extremely challenging for the Patriotic Front for four reasons.
First, the Afrobarometer is a well respected international survey organisation based in Africa and so its findings are generally trusted and influential.
Second, it reveals widespread disappointment and disapproval of the current government, with large majorities stating that the country is moving in the wrong direction.
Third, the survey found that only 15% of respondents felt close to the Patriotic Front, while only 24.9% planned to vote for it. I cannot remember many elections in Africa where the incumbent has had such low support heading into the campaign.
Fourth, when you take out the people who said that they would not vote or did not answer the question (so that you mirror the electoral process and just look at the proportion of the vote candidates would actually receive), the results of the Afrobarometer survey places the opposition UPND ahead on 50.4% of the vote – enough to secure an outright victory in the first round.

It is true that poll results don’t necessarily translate into victory, and it is always important to interpret results with caution. Having more supporters does not mean you win if you cannot get them to the polls. But any survey showing the government behind is going to cause great consternation within the ruling party.
It is also true that a remarkably large proportion of people – 39.6% – refused to answer the question about who they would vote for. But this suggests that the survey actually underestimates support for the opposition. On the one hand, the most obvious reason not to answer this question is – as has been documented in past elections in repressive contexts such as Zimbabwe – fear of reprisals for supporting the opposition. There is little doubt that on average opposition supporters have more to fear by revealing their sympathies – especially as 41% of respondents mistakenly thought that the government was behind the research.
On the other hand, regression analysis by researchers from the University of Cape Town – which you can download here – suggests that on average the respondents who refused to answer this question have more in common with the typical UPND voter than the typical PF voter. In other words, when these people cast their ballots it is likely that a higher proportion will do so for the opposition.
Taken together, these two points suggests that if all respondents had felt comfortable in answering the question, the survey would have reported a bigger lead for the UPND than it in fact did.

Lies, damn lies and statistics

To overcome the Afrobarometer survey in the bandwagoning battle, government allies decided to run their own poll. According to this new survey, conducted, we are told, by the newly formed Political Science Association of Zambia, “44.5% of Zambians will likely vote for President Edgar Lungu while 30.3% will vote for Opposition leader, Hakainde Hichilema”. Moreover, “61.1% respondents [sic] stated they would vote for Mr. Edgar Lungu because of his policies such as preservation of jobs, massive infrastructure development such as housing units for service personnel, roads hospitals and schools.”
The problem is that the poll they put together is so bad only the most committed government supporter could think it is genuine.
The way the poll was written up is the first clue – as the second quote above demonstrates, it reads more like one of the President’s campaign speeches than a sober piece of academic research. Beyond this, there are four reasons that the survey is clearly unreliable.
First, misleading claims are made about the identities and experience of those who supposedly ran and endorsed the poll. According to the press release and the adverts circulated to drum up an audience for the launch of the poll on social media, “The Poll attracted experts such as Prof. Richard Elsen from the United Kingdom and Dr Masauso Chirwa of Zambia”.
This sounds great when you first read it. Only Dr Masauso Chirwa – the Principal Investigator – has no track record of running election surveys, and Richard Elsen is not a Professor. Not only that, he is not even an academic. Rather, Elsen runs a company called Farraline Public Relations.
By this point, I don’t think you will be too surprised to learn that Farreline specialise in helping out people with bad reputations. Elsen would therefore make a perfect advisor for President Lungu, but not a figure with any credibility to endorse an academic research project.
Second, Elsen doesn’t actually seem to know anything about the survey himself. When called up and asked about the survey, he actually says that he has not been involved with it and that “is not a pollster”, He adds – somewhat incredibly – that he “didn’t see the findings until after the evert”, and that the only thing he knew ahead of the launch was the sample size, which he thought looked “pretty good”.
Third, it is funny that Elsen says that he thinks the sample size is ok, because it is the sample that makes it clear the poll is a sham. According to the official launch of the poll, the sample size was 59,628. This would make it not only the biggest survey ever conducted in Zambia, but one of the biggest election surveys ever conducted in Africa. For comparison, most nationally representative surveys have around 1,200 or 2,400 respondents. (You can watch the official launch here).
This would be great if it was real – but sadly it is unfeasible. To put it into perspective, the Zambia Statistics Agency, which has enumerators throughout the country, takes about 6-7 months to reach around 13,000 respondents. Yet the Political Science Association of Zambia poll, we are told, managed to reach almost 60,000 people in just one month. Even working weekends, that works out at around 1,980 people a day. Given the need for face-to-face surveys using a random sampling technique, there is no survey organisation in Zambia capable of this feat.
The sample is also strange in that it doesn’t follow international best practice. You might think with such a vast number of people being interviewed the survey would have been conducted in all parts of the country. But apparently this was not so. Although Zambia has 10 provinces, the survey was only conducted in five, and not only that but it appears to have been conducted in just five constituencies in each province. More worrying still, no clear rationale has been given for this decision.
If this is correct, it means that the survey was conducted in just 25 constituencies out of 156 – i.e. just 16% of the total.
Finally, even if we leave aside the decision to only look at a tiny number of constituencies, the weighting of the poll in different parts of the country makes no sense. As Tobias Caesar Michelo has pointed out (see his table bellow), comparing the distribution of the sample in the poll to the distribution of the Zambian population “there was an underrepresentation on the CB [Copperbelt] and Western Province, and an over representation of respondents in Luapula Province.” This skews things in favour of President Lungu, as Luapula is one of the PF’s “heartland” areas.
So where does that leave us? The Political Studies Association of Zambia poll isn’t credible enough to tell us anything about who will win the 2021 elections. But it does tell us a lot about the election race. It suggests that the government and its allies know that they face an uphill battle to win in August; that they are planning to use a variety of misinformation strategies as part of their campaign; and, that they are working with foreign PR companies – and have co-opted a number of academics – to achieve this.
The way that the launch and PR around the event included three academics from the University of Zambia, with external white “experts” such as Elsen also suggests that the government understands the value of blending domestic and international “authority figures”, and will continue to deploy this combination around the polls. The danger of this strategy is that these individuals appear to be the kinds of figures that election observers would usually approach for briefings and advice – but are not simply neutral commentators.
In response, election observers, researchers and journalists will need to work harder than ever to sort fact from fiction, critically evaluating every claim and press release, in order to protect Zambian democracy from those in power.
Nic Cheeseman (@fromagehomme) is the Professor of Democracy at the University of Birmingham. He is a member of the International Advisory Council of the Afrobarometer but he is writing here in a purely individual capacity and his views do not necessarily reflect those of the organization.

Share this post