The World Of Opinion Polls

Written by Subhash Agrawal | Updated: May 6 2004, 05:30am hrs
In a remarkable display of the influence of opinion polls, almost Rs 50,000 crore were erased from the market capitalisation of leading Indian companies in a single day after exit polls from the second phase were released which showed a possible hung Parliament. Since then, there is a nervousness that borders on panic.

But the amazing thing is that, in the roller coaster of Indian politics, far too many opinion polls and pundits have been wrong in the last decade.

Even the exit polls during the 1999 elections, a post-facto capture of reality that should have been immune to swings in public mood, got almost every major state wrong, and the only reason some of them got it right overall was because their errors cancelled each other out. An amazing record, and reflecting more luck than science.

At one level, polling has come a long way since the infamous Harry Truman-Thomas Dewey presidential contest in the US in 1948 when top pollsters, including Gallup, predicted a Dewey win, and when the Chicago Tribune came out with a premature banner headline Dewey Defeats Truman. Since then, polls have become far more accurate. But they have also failed occasionally, sometimes miserably.

In the 1993 general elections in Australia, almost every opinion poll till the very end had the Liberal party pegged as the winner (in fact, so unanimous were the polls that it was called the unlosable election by the Australian press) but Labour went on win. In 2002, most polls got it wrong on the mid-term elections in the US; Democrats were supposed to retain control of the Senate but in fact they lost widely. Moreover, in key elections in many countries, opinion polls have shown as much as a 15-point margin of blunder.

Psephology is prone to errors and biases so regard opinion polls with healthy distrust. But we do need them because they make democracy work better
Clearly, India is not alone. Bad polls are common everywhere. But why do we have them consistently bad in India The answer to this probably lies in our unique and evolving sociology. I suspect there are many who give false, misleading or incoherent answers to poll questions, perhaps out of suspicion of city-slickers or perhaps because they respond differently face-to-face than in the anonymity of the voting booth.

Unlike market research, opinion research in politics relies far too much on close-ended questions, like Do you think Sonia Gandhi will make a good PM This is practical and quick, but it also imposes artificial choices, ignores the background context in which decisions are formed and then refined, and overlooks finer nuances. Changes in question wording can result in wildly differing estimates of support.

For instance in the US, any number of polls show that the proportion of support for Abortion on demand is much smaller than the proportion of support obtained with other wordings of the same question, such as: Do you think women should have the right to an abortion if they want one

Another problem in India is that, unlike countries where there is compulsory voting, the total population does not equate into total voters. Because of this, the polling business gets more complex because it is difficult to determine who is going to turn out to vote. The people who stay home end up distorting pre-election surveys as much as those who actually turn out to vote but change their minds during the last days.

But do opinion polls serve a useful social purpose or do they distort results Evidence on this score is incomplete. While pollsters swear that polls hardly impact on the final outcome, there is some empirical evidence to suggest there can, in fact, be both self-defeating and self-fulfilling prophecies. The most notable example of the former was the unexpected triumph of Edward Heath in the 1970 British elections when widespread predictions of Tory defeat led many Labour voters to stay at home.

But, in their defence, opinion polls also make democracy work better by making politicians more aware of public mood, no matter how rough the yardstick. In fact, what we perhaps need is more, not less, polls, otherwise we would have a public opinion that effectively existed only at election times. The bottom line is that psephology falls somewhat between social science and journalism, and is prone to errors, subtle biases and incompleteness. For these reasons, and because this field has become so sexy overnight, there ought to be healthy distrust. But we need them nonetheless.

The author is an analyst of Indian political and business trends and the editor of India Focus, a political risk report for international investors