Even honest pollsters may get it wrong
By Swapan Dasgupta
My friend Mani Shankar Aiyar has been in a self-congratulatory mood (at least on TV) ever since a post-poll survey suggested last week that the DMK-led alliance would be a nose ahead of the AIADMK-led front in Tamil Nadu. So resounding was his faith in a survey that went against the grain of Madras chatterati wisdom that he belligerently accused me of assaulting Tamil asmita by daring to suggest that a booster dose of money power had marked the final days of an already extravagant campaign.
Since Aiyar's take on politics is decidedly eccentric—he feels the CPI(M) is losing in West Bengal because it abandoned Marxism and embraced Manmohan Singh-ism—it is often difficult to separate his conviction from posturing. Yet, if, by the afternoon of Friday the 13th, the Congress MP who was also a founding member of the Trinamool Congress, looks silly, he will be just another statistic in the long list of politicians who equated exit polls with the final results and lived to regret it.
This is not to endorse the smug complacency of Communist apparatchiks in Alimuddin Street who sincerely believe that an eighth Left Front Government in Kolkata is pre-ordained. The suggestions of a pro-Mamata tsunami, they have repeatedly asserted, is a media conspiracy. Last Tuesday evening after a clutch of exit polls also indicated that the Left Front was certain to lose power quite decisively—the only debate centres on the Left Front's ability to touch 100 seats in the 294-member Assembly—the CPI(M) reaffirmed its condescending disbelief in the results. In a pro-Left Bengali channel, a CPI(M) functionary recounted in graphic detail how the polls got it all wrong in 2001 and would again get it wrong in 2011.
The Left genuinely believes it is winning in both West Bengal and Kerala. This optimism which, quite remarkably, has also been internalised by its cadres, is based on field reports that are held to be a more reliable guide to public opinion than surveys conducted by market research organisations and pollsters.
The faith in internal assessments isn't confined to the Left alone. During the 2009 general election, a lot of importance was attached to internal reports by full-timers suggesting the BJP was in a winning position in some 32 of the 80 Lok Sabha constituencies in Uttar Pradesh. This was reinforced by some self-serving polls conducted by in-house pollsters. When the votes were counted, the BJP won just 10 seats in the largest state.
Political functionaries don't invariably exaggerate a party's popularity. After his decisive re-election last year, Bihar Chief Minister Nitish Kumar told me that his bid to galvanise his political workers prior to the polls had not been a grand success. At the state political convention of the Janata Dal (United) shortly before the campaign kicked off, an open session on the performance of the state government became the occasion for venomous outpourings of disgust. An outside observers would have concluded that the JD(U)-BJP alliance would suffer from the indifference and hostility of its own activists.
In 2007, I witnessed a similar alienation of Gujarat Chief Minister Narendra Modi from his own party functionaries. Many apparatchiks told me "in strict confidence" that Modi would be lucky to win with a narrow majority. The results indicated otherwise.
Both Nitish and Modi reposed greater faith in independent pre-election surveys than in internal feedback. Realising their potential vulnerability on the organisational front, both mounted spirited presidential-style campaigns aimed at galvanising voters over the heads of sullen activists. Confronted by a sharply divided party, Kerala's indefatigable CPI(M) warhorse V.S. Achyutanandan appears to have adopted a similar approach in this election. The results of his late surge are awaited.
Unlike the West where political parties employ experienced professionals to dissect poll data, Indian politicians tend to be rather casual in their approach. Some of the wariness is understandable. Although the methodology for political polls has been fine-tuned since Prannoy Roy and Ashok Lahiri first employed the Index of Opposition Unity to gauge the quantum of Congress vulnerability in 1982, there is no infallible mathematical model to forecast the conversion of popular votes into seats in India. The multiplicity of parties, some with very narrow territorial presence, the continuing relevance of caste, class and religion and the understandable wariness of people to reveal their voting preferences to total strangers have posed problems for pollsters.
Bob Worcester, a pioneer of the MORI polls in post-War Britain, used to say that opinion polls are 95 per cent skill and five per cent luck. In the case of Indian elections, the luck factor is at least 25 per cent. In last year's Bihar Assembly election, for example, a NDA victory was the common conclusion of all the major pollsters. However, there was a wide difference in estimating the quantum of victory.
If Bihar's unequivocal verdict could produce such vastly different estimates, it is not difficult to gauge why today's counting will be marked by the uncertainty of verdicts in Tamil Nadu, Kerala and Assam. West Bengal is the only state where there is no dispute about an unequivocal mandate for Mamata Banerjee. Even in the case of Kerala, a state where there is a stable two-party system in operation, there is no unanimity over the likely winner.
Assam poses a typically Indian challenge to the pollsters. On the face of it, the Congress is facing a badly splintered opposition and should be the biggest beneficiary of the first-past-the-post system. Yet, the fragmented opposition space is not as fragmented as it appears since the sub-regional pattern suggests two or at best three cornered contests. The Congress is the common factor throughout the state but its challenger changes from region to region and even shifts according to community. A pan-Assam extrapolation of data could therefore produce results markedly different from a conclusion centred on ethnically-defined sub-regions.
But there is an associated complication. Experience suggests that pollsters are more likely to forecast a correct result for the entire state than they are for a single Assembly segment. The larger the area the greater the scope for accuracy since errors cancel out each other. The implication is obvious: the projected results for single Assembly seats should be viewed with the scepticism reserved for both the mythical "BBC predictions" that resonated in North Indian small towns till the mid-1990s and the Intelligence Bureau assessments that formed the staple diet of 'informed' media speculation.
Finally, there is the issue of the credibility of the polling agency. Most of the reputed ones attempt an honest exercise, even if some of their sampling is suspect. However, over the past decade the market has witnessed the arrival of quite dubious pollsters and pseudo-psephologists. The cleverest of the dodgy brigade base their conclusions on random tea-stall conversations with villagers on the highway and blend these with assessments culled from local journalists and police. The more brazen ones speak to a few politicians and then arrive at their conclusions. Occasionally, they even get their conclusions right.
The most unscrupulous ones undertake 'paid polls' where the results are pre-determined and aimed at countering the negative impact of other, well-intentioned polls. The bespoke pollsters are most in demand in northern India where the ferocity of competitive politics has nurtured a local media dedicated to extortion and blackmail. The criminalisation of psephology is a uniquely Indian invention.
This is why it pays to still wait for the votes to be counted.