71 Exit Polls Analysed : What did we find?
We considered 71 Exit polls from 2014-2017 covering 12 large States – Assam, Bihar, Delhi, Gujarat, Haryana, Kerala, Maharashtra, Punjab, TN, UP, Bengal and Jharkhand. What did we learn about Seat Forecasts?
Average of Polls are Unreliable (16% of the time accurate)
– Never calculate an average of all polls – They miss the actual result by an average of 17%
– Only in 2 States has the average missed the actual result by less than 10%
The maximum is more reliable but only sometimes (33% of the time accurate)
– The Poll giving the highest seats for any party is more reliable than the averages, they miss the actual result on an average by 7%. But….
– Out of the 12 elections, only 4 times did the max come close to the winning result
Axis most reliable but not always (50% of the time accurate)
– Axis is the most reliable poll and leads all pollsters in getting the result, but….
– Out of the 8 elections that they have predicted, they have gotten it right 50% of the times
Hyped elections often result in poor average performance
– Highly covered elections like Delhi, Bihar, Gujarat and UP have a much higher average miss than less hyped elections(more than 20%)
Only Good News
One Pollster always gets it right. If one takes all the agencies that got the election right, the miss was just 1.7%. The only problem is, we don’t know which one gets it right until after the results
In sum, exit polls are good entertainment, they probably give you the direction but as far as Seats are concerned, wait for the results