Policing the polls

Opinion polls are even more hotly contested than the elections they are supposed to second guess. So much so that even the opinion poll cond...

Opinion polls are even more hotly contested than the elections they are supposed to second guess. So much so that even the opinion poll conducted by the Press Council to ban opinion and exit polls for the two weeks prior to voting is disputed.

Conducted by an unnamed agency a straight violation of the Election Commission guideline that says details of the organisation commissioned to do the poll should be mentioned, the poll sampled 2,594 people in Delhi, Mumbai and Ahmedabad and discovered that 70 per cent of them had read opinion polls and of them, 50 per cent had been 8220;influenced8221; by the verdict. No one can quite explain whether this influence will translate into voting behaviour but the Election Commission thinks so. In fact, instead of banning publication, publicity, and dissemination of the polls, the Election Commission should encourage it. Especially since these elections have been like no other.

It8217;s not just that since 1989, the traditional Index of Opposition Unity model doesn8217;t work. Inthat model it was clear who the voter was: either pro-Congress or anti-Congress. Now, says psephologist Surjit Singh Bhalla, 8220;in a multi-party system, you don8217;t know whether the voter is anti-Congress or anti-BJP8221;. That is why, he says, most forecasters met their deathknell in Andhra Pradesh in 1996. That8217;s also why, he adds, opinion polls rarely make a mistake in two-party States like Rajasthan and Delhi.

What makes the 1998 election particularly risky for the pollster is the alliances. Says N. Bhaskara Rao, Chairman of the Centre for Media Studies: 8220;This is not a typical election. If you add the Samajwadi Party to the Congress in your Maharashtra calculations, you can8217;t do the same for Uttar Pradesh because they are rivals. You add the BSP in Punjab but not in UP.8221;

Bhalla thinks the particularly nettlesome States this time are Andhra Pradesh, Uttar Pradesh, and Orissa. In Andhra Pradesh, it is difficult to say whether the new BJP/Lakshmi Parvathi voter is the traditional Congress voter or the TeluguDesam voter. In 1996, when practically every forecast was wrong about Andhra Pradesh, the Telugu Desam got 32.6 per cent of the votes and 16 seats and the Congress got 39.7 per cent of the vote and 22 seats. The BJP worked up only 5.7 per cent of the vote and NTR TDP won 10.7 per cent of the vote.

This year, in Uttar Pradesh, one doesn8217;t know to what extent the BSP will play the spoiler. Will it hurt the SP more than it hurts the BJP? Orissa is tricky because it has the largest number of don8217;t knows.

Even in the normal run of things, much of the methodology is needlessly wrapped in arcana. Clearly, there are five clear sources of error in any poll:

Story continues below this ad

The population might not be representative

The respondent may be lying

The constituency being studied might not behave like another in the same State

The translation of votes into seats is highly subjective

The voter can change his mind

The greatest source of error is clearly in the translation. Ask any pollster how hetranslates the all-important voting percentage into seats and watch him clam up. Says B.S. Chandrasekhar, Director, Audience Research, Doordarshan: 8220;This is where the researcher8217;s prejudice plays a big role. A party can fritter away its higher percentage of votes in the winning seats.8221; Or like in 1996, two parties can get the same percentage of votes and yet have different seat tallies. For instance, the BSP and SP got practically the same percentage of votes in Uttar Pradesh 20.6 and 20.8 per cent respectively. Yet the BSP got only six seats and the SP won 16.

N. Bhaskara Rao, Chairman of the Centre for Media Studies, the oldest pollster in the business, deigns to tell you some of the variables he uses in his statistical model: the homegeneity of the chosen constituency, the behaviour of the caste group, the perception of the voter influenced by the party affiliation, the possibility of tactical voting, and the preference for the candidate. Bhalla tells you only that his is a statistical model.

Butthe methodology of practically every poll is being viewed with some doubt. Even the Centre for Study of Developing Societies poll for India Today, quoted the most often, has been questioned. 8220;Their method of selecting a panel of 8,938 voters which includes 7,457 of those interviewed in 1996 is suspect. How have these voters been shortlisted,8221; asks Bhaskara Rao.

The Times of India/Development and Research Services poll, according to him, extends over too long a period. In an election where the situation is changing daily, an 11-day period makes nonsense of the base factors. As for the C-Voter poll for Pioneer, the Congress has evidenced some circumspection over the organisation8217;s BJP leanings, which they feel is manifested in the high proportion of seats 8212; 268.

Story continues below this ad

A new survey by the CMS indicates that 40 per cent of the voters are influenced by pre-poll surveys and Bhaskara Rao insists that greater exposure has made it a level playing field for voters. But did Manohar Singh Gill8217;s January 21 diktatprove effective in a scenario where opinion polls look like they miss more often than they hit? We8217;ll know in a fortnight.