Don Anderson | A pollster’s perspective on polling
I like to hear people say, which they often do, “I have never been interviewed by any political pollster,” because it gives me the opportunity to explain why.
Sample size considerations
Our organisation, Market Research Services Limited, has relied on the design of a nationally representative sample of persons throughout our 45 years in the business as the basis for making accurate and reliable projections to the entire population. Whether we design one with 1,000 persons or 3,000 persons is not the most important factor; rather, it is to design a sample that precisely mirrors the national profile.
In other words, it is one that is representative of the population from which it is chosen. This allows for the derived data to be accurately projected to the population. This means that in a 1,000 sample, which is most often used, 2,799,000 do not get interviewed. In the USA, the best pollsters in the world hardly ever use a sample of more than 1,000 to project the outcome of the presidential elections. Population? 350 million.
Importance of trend data
Pollsters are best when they have trend data from a successive number of polls to work with. The trend is what enhances the ability to project the outcome of an election. For example, in the 2007 national election, we were able to predict that the Jamaica Labour Party (JLP) would win 32-28 by virtue of the benefit of doing five or six polls between January and the September election.
The result of that election: JLP 32, People’s National Party (PNP) 28.
The trend for the 2020 election, though over a longer period, makes interesting reading.
Jan 2016 Feb 2016 Feb 2018 Mar 2019 Feb 2020 July/Aug 2020
% % % % % %
PNP 26 34 19 18 22 20
JLP 23 31 26 29 30 36
/DK 31 35 55 53 48 44
Source: Don Anderson National Polls conducted for the RJRGLEANER group
Note that the 34 per cent/31 per cent (dead heat with a margin of error of plus or minus three per cent) final poll reading for the 2016 election was exactly how the election went, with the 50 per cent/50 per cent eventual voter support for the respective parties in the election, hence spot on.
Who uses polls?
The polls are important to a number of different segments.
• Political parties use it to assess their standing and to plan strategy, to test candidate strength, and to develop appropriate messaging for the campaign.
• The general public use it to help them decide how to vote and for sheer interest.
• Sponsors use it to determine which party to back more.
Polls can influence voters
When the contest is really close, poll results tend to have the effect of galvanising the respective supporters in order to gain the competitive advantage. When polls suggest that one party is significantly ahead of the other (10 per cent or more), the effect is potentially twofold. The party that is significantly ahead is usually energised, and other persons, who like to back a winner, join forces in support – the bandwagon effect. On the other hand, the party that is behind quite often finds supporters losing heart and retreating to not voting.
Parties are inclined to diss poll results when they do not favour them. This is natural, and I have experienced enough of this not to take them seriously when they do, although I never ignore them. Fact of the matter, those who are most vocal in dissing polls conducted by me are generally the first ones who call me the morning after to ask if I have any other information to share. The strategy in discrediting poll results is designed to convince voters not to accept the results as if this will by itself boost their numbers. If the poll numbers look good for you, you are the best. If not, diss them; it makes you look better.
This has been used so often that the voting public understands that there is quite often no merit to the public denouncement.
Recently, polls have been increasingly questioned because on occasion, the ‘predictions’ have not been consistent with the actual outcome. One reason for this is that a number of extraneous factors, such as voter influencing through inducement of tangible personal benefits either to vote or not to vote, intimidation, breakdown in on-the-day transportation strategy, and simply voter apathy on the day, have distorted the scientific process. These are not factors that can be easily measured or calculated for. This is the reality.
In the final analysis, political polls are designed to plot voter intention leading up to an election. The last poll should be done as close as possible before election date to capture the impact of campaigning.
What really matters most, however, is how the machinery works on election day. Elections are, after all, decided on election day; polls only provide a guide as to how this could go.
- Don Anderson is executive chairman of Market Research Services Limited and senior adjunct lecturer in research methods at The University of the West Indies, Mona. Send feedback to email@example.com