Why exit polls get it wrong
Published: 23rd December 2017 04:00 AM |
Another election, another exit poll that has gone horribly wrong. That sums up what has now become a regular feature of every election in India, perhaps even abroad. Hours after voting in Gujarat ended on December 14, TV channels began running exit polls. Political leaders, psephologists and analysts engaged in animated debates, as if the final word had been said and the winner identified. Almost every channel gave the BJP a comfortable win, some even projected a better performance than 2012. But in the end, the ruling party could not even notch up a century; and the Congress put up its best performance in decades.
The Gujarat results were not a one-off case. The same was the case with Uttar Pradesh, Punjab and Assam, where elections were held earlier this year. While no channel or polling agency foresaw the BJP wave in UP, they also hedged their bets on Punjab and Assam, where the Congress and the BJP won comfortably.
So barring exceptions, why do exit polls get it wrong on a regular basis? Although all surveys have a margin of error, the cause of their errors lies in the sampling. Most surveys in India have a sample size that is big enough to represent the total electorate. But where they perhaps go wrong is their urban and semi-urban bias. Polling agencies very rarely include the hinterland in their surveys, ignoring the fact that almost half the voters reside in the rural belt. They perhaps give the rural voters the miss to keep the cost of the survey low. But such a survey can only give a skewed result.
An analysis of the Gujarat result bears out the ignorance of the rural belt. Almost half the seats the BJP won came only from four cities in the state: Ahmedabad, Vadodara, Surat and Rajkot. And what had the exit polls projected? A comfortable majority for the party. While a direct correlation between the BJP’s city-centric voter base and the sampling of surveys would be erroneous, it does provide a clue as to why exit polls are off the mark so often.