Poll surveys are here to stay

For an exit poll the one and only criterion for evaluation is its accuracy. In the case of pre-poll surveys, the criteria would be multiple, including: (a) how well ahead of the poll was the likely overall outcome indicated, (b) how close are its predictions to the final outcome, (c) what was sample size of the constituencies and of voters for conducting the pre-poll survey and, (d) more importantly, how well the issues involved in the context of the polls were brought out and how detailed and desegregated the voting preferences were given.

If an exit poll does not go beyond the level of pre-poll survey in precision, obviously the very idea of an exit poll and the methodology followed deserve critical consideration.

Accuracy of a sample survey is expressed and measured in two different ways. One is in terms of "margin of error" and the other is in terms of "level of confidence". A "90 percent confidence" does not mean that a survey prediction could be plus or minus 10 percent of the projected figure as one chief of a news service had interpreted the results of an exit poll on the night of the counting day, November 28. Rather, it implies that such asurvey, with given methodology, would yield the same result 90 times if the survey is repeated 100 times. In an electoral context, however, such statistical confidence level is irrelevant, since such criteria apply more in controlled experiments/situations. In an electoral context what is more pertinent is the "margin of error" in the projections.

To be worthy, the margin of error expected and allowed in such surveys should be within four percent plus or minus. That is, if the projected number of seats of a party by an exit poll is 100, "it is ok" even if the final result is between 96 and 104. Here again, the larger the spread and more the regional distinctions, as in the case of MP, the margin of error could be within plus or minus five percent.

However, when an exit poll is preceded by pre-poll surveys, it is expected to have even smaller margin of error. Similarly, in the case of sequential pre-poll surveys, they are futile and misleading if they do not bring out the changing phenomena and theundercurrents over time during the campaign period.

Relevance of an exit poll is only when there is a time gap between the poll date and the counting date. Exit polls by their very nature add or intensify and even prolong certain speculation and guessing about a poll outcome. Anxiety of voters and parties would of course be there to know the outcome the same day of the poll.

All this apart, an exit poll has the potential of vitiating the voting pattern of "re-poll" in certain booths/constituency, particularly when the number of booths involved in the "re-poll" the next day or the following day are many. The "bandwagon effect" in such cases could be detrimental for the final outcome when the contests are too keen as in some constituencies of Delhi last month. Whereas pre-poll surveys, on the other, are relevant both to the parties and the voter equally. The parties could get the pulse of the voter and help select the right kind of candidates and take up the issues bothering the voter for the campaign. Thevoter too will feel comfortable that he or she had been enlightened about the dynamics of what others were feeling about.

The writer is chairman of Centre for Media Studies, New Delhi