As this country’s national elections coming “any time before next September,” the frequency of political opinion polls will multiply. In every country of the world, especially in democratic ones like ours, public opinion polls are quite common - an industry. Opinion polls have become a useful management tool. Between elections, governments are interested in knowing “how people think” about certain issues.
People like to know what others think, the foundation of public relations. But such curiosity makes them vulnerable to persuasion and even propaganda. In a world of growing contradictions and the human desire for economy of thought, information about a coming election is in great demand. In a climate where every party usually claims some superiority, real or imagined, the poll appears as Andrew Liver Salts. It tends to settle things a bit. The human mind hungers for closure, from putting dragons in unknown seas to visiting the obeah man, the psychic.
The Public Opinion Quarterly has repeatedly argued, political opinion polls must be done as scientifically as possible because of the implications of the results. I am of the growing view that no political opinion poll results should be published on the day before or even three weeks before the election. Backed by substantial research, the bandwagon effect is alive and well, especially here. And further, such pre-election publication could very well diminish the natural chances of third or minor parties. But there is free speech.
Since 1935, there has been an intense debate over the concept of “public opinion” in terms of its volatility and usual vagueness. The results of a political opinion poll are as important as the methods used. Strongly complaining against editors and politicians who treat all polls as equal. Pollster George Gallup way back in 1966 noted that there are “at least 30 kinds of polls,” All polls are not equal. As a former practitioner myself, I can tell you a fair and proper “scientific” poll could be rather costly and time-consuming.
There are three reasons: one, the sampling frame must be quite representative of the population targeted, and finding such a sample demands expertise; two, interviewers have to be paid and these days, they don’t come cheap; three, getting the exact persons chosen for the sample can be quite tedious and often not found (sample replacement helps). When I hear affected politicians quarrelling with one another over “what the poll says,” I often wonder if they really know what the poll really says and how exactly did the pollster arrive at the percentages.
Now a lot of pollsters boast that their poll “predicted” the election result. But when the prediction fails, they use an escape clause. They say the poll was just “a snapshot of political opinion” and subject to opinion change. When a pollster uses a sample spread across the country and gets, say, a 56 percent vs 44 percent difference between two parties A and B, there is the prediction that party A will “win” the election. Not necessarily. What wins a national election is not the overall national percentage.
It is a constituency by constituency race. And a party can win a national election gaining more seats than the opposing party which has a higher number of the national votes. But even if the poll is done on a constituency by constituency basis, the stratified demographics, especially in certain areas, must be taken into account. The pollster can apply random selection but he must have the full population list from which to choose so that every person will have an equal chance of being selected. Then there are the non-responders.
But it is more. Since different ethnic groups, social class or gender may not be of equal number in the population targeted, random stratified sampling will be “more scientific.” Then there is this “margin of error” which seems carelessly used. All results from samples are population estimates, the basis of statistics. A “margin of error” means, for example, the extent to which the sample result is far from the true population score. Random sampling makes it more scientific. So there are polls and polls. For such reasons, publishing the sampling details help the more discerning, even when dissenting. And for fair elections.