Sample surveys, as the late ad guru David Ogilvy might have advised, should never be used as a drunk uses a lamp-post—for support rather than illumination. What’s being cast in light by such a survey could well be illusory too, since it’s filtered through a lens that picks up a scatter of data dots and offers only a hazy view of reality at best. Wispy markets such as satta bazaars for bets on electoral results tend to reveal even less, and Prime Minister Narendra Modi’s glee last Friday was hard to miss as he recounted how his Bharatiya Janata Party (BJP) routed their forecasts by achieving a clear Lok Sabha majority five years ago. This year, bookies were reported to be offering even odds—no-gain-no-loss—on a BJP tally of around 245 seats, and the average count of exit polls released yesterday now seems to confirm a clear victory for BJP and its allies—set to bag 300+ seats. As the spymaster rule goes, it’s silly to brush off the same signal from two separate sources. That in itself does not make it reliable, though.
Gambling dens are best dismissed as flawed markets: their trade is too sparse and clubby to get a wide grasp of a country as diverse as ours. But our “statistical" polls have been way off the mark, too. The last time round, almost all of them underestimated BJP support, only a touch better than their failure to spot Congress gains five years earlier. That said, drastic miscalls have fallen since their glaring credibility crash of 2004, when a dramatic Lok Sabha shift left upper-crust India stunned. Yet, the classic problems of polling persist. Truly representative random samples, with each voter equally likely to be picked, remain a distant dream in a country so vast. There exist varied clusters of think-alike subgroups, and since their ratios differ across constituencies, it’s hard to get clear vote shifts without feeding a costly field scan into a well-crafted formula. And then there are biases that Indian social conditions slip in. Some voters like to retain the secrecy of their vote even after leaving the booth, and silent supporters of one party or another could end up under-counted. Dominance of public spaces by a party in power or slice of society could also result in false but conformist responses. The conversion of vote shares into seats is a challenge that has become stiffer as multi-polar contests reduce the gap needed for a win. In 2014, the BJP, whose support is less scattered than that of its rival Congress, scored a sharp gain in seats on a relatively moderate swing. The exit pollster closest to bullseye that year was Chanakya, which may or may not have adopted a Bayesian technique based on likelihoods drawn from prior patterns. So while flukes are always possible, no matter how smart pollsters’ lenses get, blurry snapshots are all we can hope for.
Straw polls may be flawed, but even our democracy can’t always claim to represent the popular will much better. In a first-past-the-post system, after all, less than a third of the electorate can usher in a majority government, leaving two-thirds of it powerless, so to speak, over critical calls of governance. Various reforms have been proposed to address this asymmetry. Whether or not they are adopted, an awareness of the oddity should be incumbent upon the eventual victor.