The results of the opinion poll by Research World International (RWI) have predictably provoked debate (and in some cases anger and outrage) not only about the accuracy of the results, but also the timing of their release.
The Uganda Governance Pulse survey, which was bankrolled by civil society organisations led by the Great Lakes Institute for Strategic Studies (GLISS), was conducted in March. The poll has a margin of error of +/-2% at the 95% confidence level.
Let’s first get the science out of the way. The margin of error describes how the results from the sample are likely to vary from those of the actual voting population. The reported margin of error of the Governance Pulse survey means that if RWI conducted the same survey 100 times, the results would fall within 2 percentage points of the wider population 95 of those times. So the results on each survey question would be 2 percentage points higher or lower than what one would get if they had done a census of the whole voting population.
The results from the 2,321 respondents are generalised to the whole voting population because the sample was selected randomly, meaning each and every voter had an equal chance of being selected to participate.
Of course so many things can go wrong, including asking biased, leading, loaded, or confusing questions; polling assistants not sticking to the sampling plan; selected respondents refusing to answer; or fearing to reveal their true intentions; or pollsters fiddling with the results.
Professional pollsters work to avoid all these problems or to have mitigations e.g. for non-response.
There are challenges but when conducted by credible organisations polls are the closest things we have to getting close to the true pulse of the public.
Back to the RWI poll, the results, released more than four months later, show that if presidential elections had been held in March, 47% of Ugandans would have voted for incumbent Yoweri Museveni, while 22% would have gone for new comer Robert Kyagulanyi aka Bobi Wine and 17% for four-time contestant Kizza Besigye.
Even after factoring in the margin of error, this means that President Museveni would not have received the 50% + 1 vote required to avoid a run-off. But it also means that Bobi Wine was still far from matching Museveni’s popularity and that Besigye, who has not indicated he will contest for the fifth time, had fallen behind in support.
Predictably each camp has dismissed the poll, with apparatchiks saying their candidates have far higher support. None of them has provided an alternative measure of this support.
I always feel for Dr Patrick Wakida, the RWI chief executive, as he appears on one radio and TV show after another explaining the legitimacy of his polls and the results.
As an academic I will always give him the benefit of the doubt, until somebody produces credible evidence to suggest his polls are questionable.
In any case, the sponsors of the polls, if they so wished, can always allay public fears about the quality of the surveys by giving access to selected organisations to conduct independent audits.
A number of people have rightly questioned the timing of the release of the results. RWI and their sponsors argue that they couldn’t have released the results earlier as the country was in the middle of a response to an unprecedented crisis, the Covid-19 pandemic; the opinion poll would have been an unnecessary distraction.
I think we should all agree that the results released this week tell us very little about the current opinion of Ugandans on who they will vote for in 2021. Polls measure public opinion at a particular point in time; they are a snapshot of opinion. And public opinion is fluid; it continues to shift in response to real world events and stimuli.
In our case the government’s Covid-19 response, including the lockdown and the gradual re-opening of the country, must have had an effect on the thinking of voters. Also, Covid-19 has seen Museveni dominating the airwaves while his opponents have not been allowed to campaign (not that they have taken it lying down; they have figured out other ways to get their message across).
For those who think the government’s response has saved Uganda, Museveni would probably get their nod, and therefore if an opinion poll were conducted today, he would conceivably be more popular than he was in March. But for the millions who have suffered the indignity of not being able to feed themselves or their families or those who have lost loved ones or whose rights have been wantonly violated because of the senseless enforcement of the lockdown, Bobi Wine or Besigye could be the answer.
There are obviously other intervening developments that would all have an effect on the public pulse today.
Admittedly, some of the results may not be significantly affected by the passage of time (remember that the poll asked several other questions beyond voting preferences) but we need to have more opinion polls by RWI and other pollsters to gauge current opinion.
Finally, I don’t agree that the results of the RWI poll conducted in March are useless. As long as all the necessary quality control measures were deployed, they provide an important baseline that we can use to measure the growth in support (or lack of it) for different candidates as well as changing public perceptions about the state of our politics.
Some tips for journalists:
- Who sponsored the poll?
- Always ask for the methodology
- What kind and size of sample was used?
- Ask for the response rate
- Beware of non-response bias
- Ask for the margin of error
- Ask for the confidence level
- When was the poll conducted?
- What specific questions were asked?
- What quality control measures were in place?
- How do results compare to previous polls?
- Remember that differences (e.g. between candidates) that are not well above the margin of error are not ‘statistically significant’.
- The fact that you or your colleagues or relatives have never been interviewed for a national opinion poll doesn’t make polling dubious.
Dr Mwesige is co-founder and executive director of the African Centre for Media Excellence (ACME) Email: firstname.lastname@example.org; Twitter: @pmwesige