Poll says Museveni wouldn’t win election

Importance of opinion polls

Opinion polls have always been enlisted mixed reaction with parties that are polled to be doing badly often dismissing the findings as merely being perceptions while those that find the poll findings quite favourable singing praises. The opinion polls are regularly conducted on voting intentions and leadership preferences but can be undertaken on any public policy issue, which the polling groups or commissioners of such polls deem important.

After the release of the poll, Kyagulanyi also known as Bobi Wine, took to his facebook page and wrote: “Whenever we approach elections, he (Museveni) sponsors fake organisations to release fake opinion polls giving him an upper hand so as to pave way for his rigging machine… We must remain ALERT. The signs are all over that the man is on the way out.”

According to some analysts, Museveni’s low numbers this time might mean that many voters are still undecided about the leading likely presidential candidates including Museveni, Bobi Wine, and Besigye. Some voters might even be undecided on whether to vote at all.

This is partly because many voters want their vote to matter. If they sense that it will not matter, they may stay away, especially in a two horse race elections as has happened in the past. According to some research, there is evidence that when the public is told that a candidate, in this case Museveni,  is extremely likely to win, some people may be less likely to vote.

But the accuracy of a poll in predicting an election outcome may also be down to technical issues.  Political polls only perform well when done well.

The manner in which a polling organisation conducts its survey has consequences for data quality and accuracy of the poll. Some methodologies lead to bias because they do not account for people who do not respond to the pollster, people who are targeted but are not sampled, and errors by interviewers, respondents, data processors, and other survey personnel. So the margin of errors tends to be higher than the typical 3 percentage points. It is more like 6 percentage points or higher which renders it virtually meaningless.

That is because good polls use a statistical adjustment tool called “weighting” to make sure that their samples align with the broader population on key characteristics. They adjust the datasets using a core set of demographics to correct imbalances between the survey sample and the population.  But using just a few variables like age, gender, setting (rural-urban), region, and level of education, might be insufficient for getting accurate results.

Challenges of the poll

Wakida’s team, in fact note additional challenges that might have affected the reliability and objectivity of their results. The first challenge is the regulatory and political environment in which surveys have to be conducted, including every researcher to be vetted by the Office of the President.

Uganda’s geographical diversity, with at least 15 distinct regions that are culturally, linguistically, and economically distinct indices also poses a major challenge. Wakida says most previous polls did not pay attention to these diversities and differences.

“By generalising perceptions at the national level, it is possible to suppress the differences in opinions driven by these cultural, linguistic and other differences based on geography,” the report authors say.

In this report, Wakida’s team set out to do face-to-face interviews with 2400 people in 15 regions over 10 in March 2020. But they failed to get 79 or 3% of their target sample in spite of trying very hard.  They visited selected homes in the evening when they hoped to find the head at home. When they failed they returned two more times. If that failed, they traced them to their work place.

In spite of that effort, in Ankole region the researchers interviewed only 187 respondents out of a target 229. They failed to get 40 respondents or 18% of their target. Perhaps to compensate for that the researchers ended up interviewing more people than targeted in other places. In Sebei, a small population, they planned to interview 24 people only but they ended up interviewing 67. In Tooro the target was 86 interviews but they did 125. These additions appear small but in statistical terms, they are quite significant.

Pollsters spend a lot of money to get as large a number of respondents as possible. That is because the margin of error, which indicates the level of confidence one should have that a poll result would reflect the result of a survey of the entire population, depends on it.

The sample of 2321 that RWI used, gave it +/-2% margin of error and at a 95% confidence level. The sample, without the inflated figures from Sebei, Tooro, Karamoja, Bukedi, and Acholi would have been lower. This means the margin of error of the survey would have been higher. That means the results would have a lower confidence level than 95%. In other words, Besigye would be right. That what a poll by RWI says is completely nothing compared to the reality on the ground.