|Competing exit polls will storm the small screen Saturday evening, soon after voting ends in the seven-phase assembly elections in Uttar Pradesh, where pollsters have often forecast the broad trend but always failed to hit the bullseye. Will the Samajwadi Party (SP) return to power as many are suggesting?
The exit polls, involving a survey of voters immediately after they cast their franchise, promise to provide the answer before the results come in on March 6. But as the Mayawati-led Bahujan Samaj Party (BSP) strives to stave off the challenge from its arch rival, with a resurgent Congress led by Rahul Gandhi and the Bharatiya Janata Party (BJP) also battling to recover lost ground, pollsters will be hoping to avoid a repeat of their previous forecasts.
For instance, while the polls managed to predict the BSP would emerge as the biggest party in 2007, its margin of victory eluded them by a long mile. The party ended up with 206 seats, comprehensively beating forecasts that ranged between 117 and 168 seats. "Exit polls should make news only when they go wrong," says Devendra Kumar, founder of Delhi-based Research & Development Initiative.
"But the plurality of Indian society and polity makes exit polls unsuitable in our context," he says, adding that though exit polls are a science, in India they have to be a mix of science and art because data needs to be interpreted in the local context. This brings in an element of subjectivity, which can lead to varying interpretations of the data. The sheer diversity of India's most populous state, with complex four-cornered electoral contests that depend on varying local factors, including castes and shifting political alliances, make forecasting a pollster's nightmare. Ahigher turnout and a sizeable chunk of new voters this time around will only add to the complexity.
"Exit polls in India are indicative of the results, but not accurate," says CV Madhukar, founding director of Delhi-based think tank PRS Legislative Research. "This has to do with the evolution of the science in India," he says. "The challenge for pollsters is to stratify the data and get their samples right. Another issue is how many respondents tell the truth because of various pressures." Psephologist GVL Narasimha Rao, who is associated with the BJP, explains the extent of the challenge.
The conventional wisdom, Rao says, is the BSP starts with about 20% voteshare because of its solid Dalit support base while the SP has about 10% Yadav votes to begin with. Again, while the BSP can win about 25-30 seats with 22% votes, it can go up to 100-odd seats with just about 25% votes because of the manner in which the Dalit voters are spread across the state.
The BJP, on the other hand, can win up to 80 seats with just 20% of the total votes because its voters are not so scattered. Moreover, unlike Madhya Pradesh or Rajasthan, which have a bipolar polity, he says, most caste groups in Uttar Pradesh often change their party affiliations, making the job even more difficult for the pollsters.
While Muslims traditionally vote for the SP and the upper castes for the BJP, for instance, sections within both communities backed the BSP in 2007 to dislodge Mulayam Singh Yadav. This time, the Congress worked hard to woo the Muslim electorate, in particular, throwing up new possibilities. The sample size depends on the forecasting required, says Rao.
For a state-wide forecast, for example, a sample of 2,000 respondents might suffice, but for a phase-wise forecast such sample is required for each phase. This not just calls for more data, but also more reports from the ground, so that data can be matched with the perceptible mood in various constituencies, he says. Given the varying concerns across the state, and the possibility of a substantial difference in the number of seats with relatively small swings in vote share, forecasts can still easily go wide of the mark. "I tune into exit polls only as a matter of curiosity," says Madhukar.