Exit polls are supposed to be much more accurate than Opinion polls. Then how is it that the various exit polls are predicting such differing results in yesterday’s elections? Aajtak is giving 93/140 to the NDA while NDTV-Indian Express guys are giving 75/140. What is happening?

I think what is happening is that they are getting the vote percentages reasonably correct, but
a) they use different models to translate percentages into seats and/or
b) an error in vote percentage gets magnified when it gets translated into seats and/or
c) because of the nature of the elections, it is possible that those errors don’t get cancelled out in other states.

For example Aaj Tak found a 3% swing in favour of the BJP in Gujarat. NDTV found a 1% swing away from the BJP. Assuming that they are calculating the swing from the same base, a 4% difference is not a big deal statistically, but it gets translated to 23 vs. 18 seats for the BJP a difference of 27% (or 21% if you take 23 as the base) That’s what I meant by b). Now the problem is that 4% means a lot in Gujarat, but it means nothing in Orissa, because there the margin of victory for the BJP/BJD alliance was huge. So the problem is the error you made in Gujarat doesn’t get cancelled out in Orissa. If you have been unfortunate enough to make a mistake in 2-3 sensitive states, you are gone.

Sorry, but I just had an attack of geekiness.