Don't count those chickens, 16 May could surprise us

When every exit poll seems to be pointing broadly in the same direction, one would presume that they have got it more or less right. Well…The Lokniti-CSDS-CNN-IBN post-poll gives the NDA 270-282 seats in the 16^th Lok Sabha. The Aajtak-Cicero poll gives the NDA 261-283 seats, and the India TV-CVoter poll 289. Even the maverick Times Now-ORG poll, reportedly carried out with the largest sample size of over five lakh voters over 180 Lok Sabha constituencies, gave the NDA 249. ABP News-Nielsen also gives the NDA a clear majority of 281 without a need for post-poll allies.From the broad trajectory of projections, it would seem that we can take a vacation on 16 May since we already know what will happen. We can read about the details when we come back, just to learn about the odd things that may not have gone as projected.Don’t count on that. Do cancel your ticket to Hawaii or Angkor Wat.16 May promises to be even more exciting and full of surprises. The problem with all the exit polls is that though they all seem to be pointing in the same direction, they seem to be hugely divergent on how they got their numbers in various states.[caption id="attachment_1520729" align="alignleft" width="380"]

Indian election officials seal an electronic voting machine after the closing of a polling center in Kunwarpur village. AP[/caption]Among the polls, Times Now-ORG provides surprises unlimited. A sampler: it shows the JD(U) doing far better than the RJD-Congress in Bihar, with 10 seats to two. The remaining 28 go to BJP-LJP. In Odisha, the BJP gets to No 2 in vote share, but gets one seat out of 21. The Congress with fewer votes gets more seats. In Karnataka, the BJP does better than the Congress, but in Rajasthan the Congress does better than the BJP, and in Madhya Pradesh and Maharashtra, it does not do as well as expected.The Lokniti-CSDS-CNN-IBN post-poll has significantly different results in all these states. The other polls also show big divergences when it comes to the state level.The worry is that when the broad national numbers are in one direction, but state-level numbers are wonky and divergent, there is scope for big surprises on 16 May. Remember 2004 was overturned in two states - Andhra Pradesh and Tamil Nadu.As we noted earlier, while the broad trend may still be more or less there, the state-level projections could be significantly different. Major reverse for the NDA in one or two states – say in Maharashtra, or Andhra Pradesh or Uttar Pradesh - could significantly dent the outcome.Secondly, even though exit polls tend to be more accurate than opinion polls, the chances are they may still fail to capture the complete trend. This is because of the bullwhip effect. When a state is trending one way, the seat projections may show a higher figure for the party making the gains, but, in the end, it could still be understating the final sweep. This last-minute acceleration was what brought the Samajwadi Party to power in Uttar Pradesh despite having just a four percent vote difference with the BSP. (For more on why polls get it wrong, read this).Suffice it to say, the game is not over. 16 May is what counts, not 12 May.