The problem with exit polls: Why can't Indian pollsters achieve accuracy, like BBC does in UK?

One of the highlights from Britain, which announced results to its general elections on Friday morning, was the relative accuracy of the exit polls prediction. Only one agency, Ipsos MORI, conducted the exit poll for BBC and Sky News, and it declared results exactly at 10 pm on Thursday, after the voting process came to an end.

Even in India, exit poll predictions are broadcast soon after the last votes are cast. But hardly ever has any exit poll agency had the degree of accuracy and consistency that the British pollsters have registered in successive elections.

Here are the facts: At 10 pm on Thursday, BBC forecast that out of the total 650 seats, the ruling Conservative Party would get 314 (down by 17 seats compared to its 2015 tally), and the main opposition Labour Party would gain over 30 seats compared to its 2015 performance to notch up 266 seats. The poll also predicted 34 seats for the Scottish National Party (SNP) and 14 seats for the Liberal Democrats.

The BBC had also forewarned that there would be a tight race between the Conservatives and Labour in over a dozen constituencies and that the end result might vary slightly from exit poll predictions.

Representational image. AP

Representational image. AP

As the first election results trickled in after a couple of hours of the exit polls' declaration, BBC made two amendments to its predictions: It increased the Tories' tally by four seats and predicted that it would win 318 seats; it also consequently decreased Labour's tally from 266 seats to 262 seats.

When the final results were declared by afternoon, the seat tally was as following: 318 for Conservative, 262 for Labour, 35 for the Scottish National Party, and 12 for Liberal Democrats.

As it turned out, the BBC exit poll missed the exact count of the two smaller parties: It gave SNP one additional seat, and Liberal Democrats two more seats than what they actually managed to get.

But the polls' predictions about the two major political parties were spot on.

And it's not as if exit polls in the UK hit the bull's eye only this time; by and large, most of the BBC exit polls in general elections held in last two decades have correctly predicted the outcome, with minimal margin of errors.

There was one prominent exception: The result of the 2015 elections. As in 2017, the 2015 exit poll too predicted a virtually hung Parliament, with the Conservatives expected to get 316 seats. But the ruling party actually went on to win 14 additional seats, taking its tally to 330. Similarly, the Labour Party was expected to get 239 seats, but actually managed just 232, seven less. The Liberal Democrats got two seats less than what was predicted (eight instead of 10). Other smaller parties were expected to win 85 seats, but they won five less to finish with 80. In 2015, the margin of error was slightly more than the other elections.

In 2010, the BBC had more or less correctly predicted the election result, just like it did in 2017. Seven years ago, the exit poll said Conservatives would win 307 seats, which was exactly the number they finished with. Labour saw a slightly different result; it ended up with 258 seats, three more than what exit polls predicted. Liberal Democrats got two less (57 instead of 59), and others one less (28 in place of 29). It appeared the Labour Party won the three seats that other minor parties lost.

Five years before that, in 2005, there was again a marginal degree of error. The Tories were then expected to win 209 seats, but actually won 198. But even then, the predicted tally for Labour was spot on — it won 356 seats, exactly the tally BBC predicted it would. The Liberal Democrats were expected to win 53 seats, but they carved out victories in 62 seats. Other parties won 30 seats whereas the prediction for them was 28 seats, meaning the 11 seats that the Conservative Party was to get as per exit poll predictions actually went to the smaller parties.

This tells us that the exit polls' record in Britain has been fairly accurate; the gap between what a party was predicted to win and what it ended up with has never been over 20 seats, and quite often, it was under 10.

Compare this with the margin of errors in case of exit poll predictions in India. Back home, the exit polls usually predict the winner of the political race correctly, but there is a big mismatch between seats predicted and actual seats won in most cases. The problem in India is compounded by the presence of multiple survey agencies that carry out exit polls for different media houses.

In the 2014 general elections, for instance, there were four major agencies conducting exit polls, all of whom predicted an NDA victory. The alliance was projected to win between 249 to 340 seats. These polls gave UPA, the main rival political formation, between 70 and 148 seats, while other smaller parties were expected to win between 133 and 165 seats.

The actual seat count in the election was as follows: NDA won 336 seats, UPA won 59, and the other smaller parties put together clinched 145 seats.

The agency that came closest to this was the exit poll announced by News24-Today's Chanakya group, which predicted 340 seats for NDA, 70 for UPA and 133 for the other parties.

The other big election that took place in India recently was the Uttar Pradesh Assembly earlier this year. Six major exit polls gave their predictions, and all gave BJP the edge, although the quantum varied widely — as wide as 155 seats by one count (India TV-CVoter) to 285 seats by another (News 24-Today's Chanakya). Predictions for how many seats the SP-Congress alliance would win also ranged widely — from 88 to 169 seats. The BSP, the third player in Uttar Pradesh, the prediction hovered between 27 and 90 seats.

The actual election result was startling for all, even for the exit poll agencies. BJP, as expected, was the winner, but by winning 312 seats, its tally exceeded the most conservative prediction by 157 seats, and even exceeded the most exuberant prediction by 17 seats. News24-Today's Chankya again came the closest to making the correct prediction.

Should the News 24-Today's Chanakya agency be considered the standard bearer for exit polls in India just as BBC-Ipos MORI is for UK?

Well, the agency's habitual propensity to give a massive edge, sometimes contrarian, to the BJP and its allies has earned kudos when it approximated the actual result, but the organisation has had egg on its face in the past, when it predicted a landslide victory for the BJP-led NDA in the Bihar Assembly elections in November 2015. Some exit poll predictions (by NewsX-CNX, Times Now-CVoter, News Nation) had then pointed to a victory of the JD(U)+RJD+Congress alliance. ABP News-Nielsen and India-Today-Cicero had given only a marginal edge to the BJP+ alliance. But the Today's Chanakya was the outlier prediction: 155 seats to the NDA, and 83 seats to the JD(U)+RJD+Congress alliance. As it turned out, NDA managed to win just 58 seats whereas the Nitish Kumar led alliance registered a spectacular victory with 178 seats. Today's Chanakya showed over-enthusiasm but it stood deflated in that election.

Clearly, in India, we do not have as yet any professional organisation which can match the degree of accuracy and consistency that the BBC-sponsored exit polls have achieved routinely in predicting the outcome of the general elections in UK in recent years.

That should be the food for thought for our poll predictors.


Published Date: Jun 10, 2017 03:48 pm | Updated Date: Jun 10, 2017 03:48 pm


Also See