Monday marked the final day of voting in India's enormous national elections, a truly gargantuan project that saw hundreds of millions of people cast votes over more than a month.
Although the final results will not be out until Friday, exit poll results are already being released. So what is everyone predicting? That's a surprisingly tricky question to answer.
While most polls have been predicting a victory for Bharatiya Janata Party (BJP) and Narendra Modi, they are far from uniform. As Niharika Mandhana of the Wall Street Journal points out, the first polls are "all over the place," with significantly divergent predictions — some forecasting that the BJP will win an outright majority, and some saying that the BJP would be forced to form a coalition government. The governing Congress party, widely predicted to lose a large number of seats, announced that it will not take part in exit poll discussions.
It may sound confusing. But for India, skepticism and distrust of exit polling has become part of the culture. And not without reason, as the two charts below show.
Political analyst Praveen Rai looked at poll predictions in the past few years for the Mumbai-based Economics and Political Weekly. Rai found that although the forecasts for the 1998 and 1999 elections were broadly accurate, the predictions for 2004 and 2009 left a lot to be desired. For example, in 2004's election, Rai noted that most polls predicted that the National Democratic Alliance (NDA), led by BJP, would win a substantial number of seats and hold on to power. In fact, the United Progressive Alliance (UPA), led by Congress, won more seats than the NDA.
And then, while 2009's predictions were far more accurate, they were not able to predict the scale of the UPA victory:
There's a multitude of reasons why polling in India is a difficult task, some of which are simple things like scale (India is really big and diverse) and technology (many people don't have phones or Internet access still). As Jonah Force Hill wrote in the Diplomat in February, few polling companies actually have the resources to do so many face-to-face interviews on a national scale and thus are forced to extrapolate data for their forecasts.
The practical difficulties feed into another issue – a widespread belief, perhaps justified, that political surveys in India are biased. Almost all polls are conducted by Indian media companies, many of which are thought to have an implicit or even explicit bias: One TV news sting operation from earlier this year reportedly found that 11 polling companies were willing to alter their results for a fee.
None of this means that Modi isn't on course for a victory, of course – for one thing, the remarkable increase in cellphones over the past decade should make polling easier (at least in theory). But it does help explain why Congress hasn't given up hope quite yet.