The research techniques involved in opinion polling are no different from those in any other field. Here too, one has to design a good questionnaire, draw up a suitable sampling plan, do correct statistical analysis and interpret the data carefully. Opinion pollsters surface and get active during every election. How does it work and how do these polls throw up the results they do?
Actually, opinion polling is very much a part of market research, so all MR professionals would be reasonably familiar with this subject.
But isn’t it very different from researching a soap or toothpaste?
Not really, and let me explain that before I can be misunderstood. For any market research study to be successful, the researcher must understand both research techniques well and the domain being researched. So, if the study is on soaps, the researcher must understand the category of soaps as well as research techniques
The research techniques involved in political research (of which opinion polling is a big part) are no different from those in any other field. Here too, we have to design a good questionnaire, draw up a suitable sampling plan, do correct statistical analysis, and interpret the data carefully.
In addition, the researcher must understand the political framework and climate well. For instance, he / she should be familiar with the LS and Assembly election process, the various parties in the fray, the candidates who are significant leaders in the constituency, how to calculate “swing”…and other relevant parameters.
So, well, it is really just another application of market research
Swing? What is that?
Simply put, it is the extent of change in people’s voting preference since the last election. It is one of the key calculations that we do when we do seat prediction.
Political research and opinion polling is all about seat prediction. Is seat prediction easy to do?
No, it is not just about seat prediction. Seat prediction is what you read about in magazine articles and on TV.
But there are quite a few leaders who use research to understand how voters perceive them vis-à-vis their competitors, what they need to do to improve their chances of winning the next election, and – perhaps most importantly – what are the key issues in the voters’ minds which could affect their choices. And these leaders use the feedback from such periodic studies to progressively strengthen their base.
Seat prediction is interesting no doubt, but it is only a small part of the whole picture. In fact, seat prediction in India is a difficult thing to do.
Yes, that is something you wanted to ask about. We often read a poll’s results in a newspaper or magazine…and when the election is held, the poll is proven quite wrong. Why does that happen? Surely the results should be more reliable if good market research firms are doing such studies.
Off the mark
Opinion polls can go wrong for two or three reasons, and one of them, in fact, can affect any kind of study, not just opinion polls.
The first reason is the simple fact of statistical confidence intervals. When a study finds that a brand has 30 per cent awareness, what we really mean is that the brand’s awareness level is somewhere between (say) 27 per cent and 33 per cent. (depending on the sample size). This range is perfectly acceptable for a brand research.
However, the same range could mean the difference between winning and losing the seat in an election. If the poll says that candidate A will get 30 per cent, and candidate B will get 28 per cent (with the balance being distributed across others), then prima facie it looks like A will win the seat. But the statistical range factor could well mean that B ends up getting 29 per cent votes and A gets 27 per cent votes….making B the eventual winner.
The second reason for polls to go wrong is the ambiguity over whether voters vote for a party or for a candidate or for an influential leader at a state / national level. So, asking voters for their party loyalty may not serve the purpose if their preferred candidate or “leader” shifts parties or announces outside support for one or the other candidate. This problem gets compounded in the case of star candidates (for example, “I didn’t plan to vote for the BJP but Modi himself is standing from my constituency.”).
Thirdly, people do change their minds in the time between a poll and the actual election. So a poll is likely to be more accurate if it is held close to the election. Exit polls, which are held after the person casts the vote, don’t of course suffer from this. Finally, so many of the people sampled in a seat prediction poll don’t eventually vote, so their responses are as good as useless. This problem does not affect exit polls, but the other two reasons, especially the first one, do.
To read more from the FundaMental section, click here .