Sunday , February 25 2018
Home / ARTICLES 2008-2015 / Good opinion poll culture

Good opinion poll culture

By Patrick Wakida (PhD)

Research World International, The Independent and Daily Monitor Polls show high reliability

Since the publication of the latest opinion poll on the Forum for Democratic Change (FDC) presidential race, which was conducted by Research World International (RWI) between Sept. 25 and Oct.3, a number of criticisms have been made of the poll.

In the poll candidate Mugisha Muntu garnered 49% support from the delegates and Nandala Mafabi 25%, Geoffrey Ekanya 3% and 23% undecided.  The Independent Magazine and Daily/ Sunday Monitor have both published independent polls of their own, all showing a similar trend in which General Muntu is ahead of Nandala.

Whereas, it is understandable for the Nandala and his supporters to be shocked by the poll findings, their criticism of the RWI poll are of a propaganda nature. They have no basis in the science of opinion poll research, which has been proven globally since the mid-20th century to accurately reflect the voting intentions or opinions of an electorate, if the exercise is conducted professionally.

Indeed, most major corporations in Uganda and the rest of the world use similar market research to improve their understanding of market needs, competitors, market gaps, product and brand weaknesses, etc. The strength of political brands is measured in the same way.

Anyone who followed the American presidential election must have noted that opinion polls carried out by media houses or political stakeholders form the basis of how the relative strengths of President Obama and Governor Mitt Romney were being measured. From America’s population of over 250 million citizens and dozens of millions of voters, opinion polls have, over many decades, been repeatedly able to accurately predict voter intentions nationally by sampling the opinions of as few as 700 voters. Most poll samples in Africa have 1,000-2,000 respondents, representing the views of millions of voters.

How the Science of Polling Works

The logic of quantitative research  is that you first study the demographic make-up of the electorate you want to poll and structure your sample with similar characteristics. For example: how many delegates does FDC have from each region of the country? How many are men and women? The sample of voters to be polled must have the same percentages of gender and regional breakdown for it to be representative of the entire list of FDC delegates.

Within that representative framework, the researcher selects the sample respondents randomly to render the sample free of any bias. Data is collected via field interviews or telephone interviews, with respondents answering questions from a questionnaire. This data is tabulated using computer software, analysed for core findings and results reported.

As a quality control measure, researchers may have 5% accompaniment, meaning 5% of respondents are interviewed in the presence of a supervisor while up to 5-10% of questionnaires are subjected to a back-check, meaning the respondents are contacted independently by supervisors and asked if the answers recorded as theirs accurately reflect  the answers they gave the interviewer. Supervisors also check the accuracy of computer data entry and analysis before confirming reports as final. Once scientific methods have been applied, a sample of 700-,2000 respondents is adequate to tell you the voter preferences of even 100 million people, i.e. less than 1% of the population is sufficient to form a sample as long as its demographic make-up is representative of the voting population. Where the voting population is very small, such as in the FDC delegates’ conference, which has fewer than 900 voters, a pollster can increase the sample to say 10-15% to reduce the margin of error in respect of representativeness.

This is how professional research is done by global pollsters like Gallup in America, and others in Uganda such as Synovate and Afro-Barometer. We at RWI follow the same standards.

When a candidate is unsatisfied about the results, they either audit the pollsters work to cross-check its quality, or to professionally carry out and publish a parallel poll of their own, which should itself be open to independent audit. This is what RWI has been telling the Nandala campaign.

Unscientific complaints

Nandala Mafabi and his aides; Rubaramira Ruranga, Sarah Eperu, Ms Margaret Wokuri and Shawn Mubiru have repeatedly alleged bias on the part of RWI but presented no evidence of any professional problem. One such unscientific attack on RWI’s poll came in the Daily Monitor of Tuesday Oct. 30, in the story titled “FDC party denounces poll findings” in which Nandala’s supporter, Toterebuka Bamwenda, the FDC deputy spokesperson rejected RWI poll findings saying the poll did not interview all the delegates! Clearly, Toterebuka does not know the difference between a polling sample response and a census which interviews the entire population. This is not methodologically necessary and is unprecedented in sample-based polling.

Partisan disagreements are okay but they should not be made using ridiculous arguments drawn from ignorance about professional methods. I would, therefore, advise the Nandala campaign to get a professional pollster to advise them on how to respond to the evident shock this poll has given them. When a professional research report says you are losing, you do not shoot the messenger or throw mud at your opponents. Instead, you should strategise for a more persuasive appeal to the voters.

In this case, the FDC Electoral Commission availed a full list of all members of the party’s Delegates Conference, which list is also available to the presidential candidates. So sampling could suffice. We interviewed a scientific randomly representative sample of 445 delegates, representing approximately 60% of the total number of delegates, much more than the minimum 10-15% that is necessary.

Our sample size of 60% is almost double that of the Daily Monitor poll done several weeks ago, which had only 240 delegates sampled, and more than double triple  that of the recent Independent Magazine poll which interviewed 180 delegates.

So why has Nandala’s campaign not attacked the polls done by Daily/ Sunday Monitor publication and The Independent magazine, which have broadly similar results to ours? Why does Major Ruranga not publish the poll he says the Nandala campaign has carried out, to prove others wrong? Could it be that their poll confirms RWI? Or do they not have a poll at all?

Independent Audit

As the Chief Executive of RWI, I personally invited and hosted Maj. Rubaramira Ruranga, head of Nandala’s campaign, at our offices, and took him through the poll results. I offered him the opportunity to take away photo copies of all our poll questionnaires. He could then get a research professional to back-check our work by calling either all the 445 delegates we interviewed or a decent sample of them, and confirm if the answers on our questionnaire are those they truly gave to our telephone interviewers. Maj. Ruranga and Nandala’s team are happy to bash our poll every day but they ran away from a transparent offer to audit it, given to them with utmost good faith. Why?

Since Nandala’s team is afraid to audit our work and confront the truth of our findings, we hereby offer this same opportunity to audit our poll to Synovate Research, Afro-Barometer pollsters, Daily Monitor, New Vision, The Observer and The Independent magazine.

RWI is confident that any comparative poll by any professional research organisation will produce results similar to ours.

Assuming that any person reading this article now understands the science of polling, the basis of our confidence in our poll report and the baseless nature of the Nandala team’s attacks on it, I can now share our detailed findings below, especially the regional breakdown of the results, which have not been given much attention in the media.

What is on the ground?

First let me start with the poll conducted by Sunday Monitor newspaper between Aug. 10-25 in which a sample of 240 delegates was interviewed. In this poll, Nandala Mafabi was scoring 20.9, Muntu Mugisha was at 36.6, Ekanya had 10% and 32% were undecided by then. In this poll, Muntu beat Nandala in all the four regions of Uganda.

Over the subsequent fortnight, as campaigns progressed, RWI commissioned a telephonic poll with a sample of 335 quite bigger than that of Monitor. In this poll, Nandala Mafabi had grown his support from 20.9 to 25.7, while Gen Muntu had grown from 36.6% to 47% and Geoffrey Ekanya had dropped to 0.6%.

A month later, RWI commissioned our latest poll, published in October, again using the delegates list from the refined/final FDC Electoral Commission.  We interviewed 445 delegates of the 500 we attempted to contact. The sample was proportionately distributed across the regions and a region with a larger number of delegates, had a bigger sample.  (See Table one above)

In this poll, not only did we ask who they would vote for if elections were held, but we also asked why they would vote for that candidate. Mugisha Muntu scored 49%,  Nandala 25%, Ekanya 3% and the number of undecided had dropped to 23%.

We also split these regions to further dissect the findings and in it we were able to see that out of the 13 sub regions, Nandala was leading in 2 sub regions: Elgon and Teso, while Muntu beat him in the 11 other sub-regions

We advise political actors to get professional help and understand how to respond to polls so that, like neighbouring Kenya, for example, we can grow a culture of polling in Uganda.

I was happy to learn that in one of the Management Committee meeting of the FDC chaired by Dr. Kizza Besigye, the party leadership took a very critical stand against the attacks on the RWI poll and the mudslinging tactics by some candidates. Besigye is reported to have tasked all to avoid dishonourable campaign tactics against colleagues in a primary contest between comrades. I concur.

Dr Patrick Wakida is the Chief Executive of Research World International

Leave a Reply

Your email address will not be published. Required fields are marked *