The Tamasha of Exit Polls

Post the last voting day on 12th of May, there has been a splash of exit polls across news channels each coming up with their party wise seat forecasts. All of them have tied up with one research agency or another and claim to be the biggest and most accurate. One news channel is already in soup as the research agency claimed it did not carry out any survey for the said channel.

Most of the news channels predict a Modi led govt. in the center, some project outright majority and some close to majority.

Party / Alliance Times Now News 24 CNN-IBN C-Voter ABP News
NDA 249 340 277 289 281
UPA 148 70 97 101 97
Others 146 133 169 153 165

While the end result is the same, what is interesting though is that most of them have very different state wise projections. This is bizarre, sample of similar people in each constituency gives different state wise results but more or less same end result. Lets see below.

Uttar Pradesh: All polls point to a BJP+ / NDA sweep. Today’s Chanakya perhaps is an outlier here and if its nos. are true then you have to give credit to Modi for converting a caste based vote bank state vote on issue of development. Range of SP is 4-15 and BSP is 3-13 big ranges.

Madhya Pradesh: All polls point to a BJP+ / NDA sweep. Times Now perhaps is an outlier here and gives 9 seats to Congress. What was Arnab’s team smoking when they did the survey?

Maharashtra: All polls point to a BJP+ / NDA sweep. Today’s Chanakya perhaps is an outlier here and if its nos. are true then Congress / NCP is decimated here. Range for Congress is 9-19.

Karnataka: This is the comedy of errors state. 4 / 6 channels say NDA will get more seats, 2 / 6 say UPA will get more. Times Now and Today’s Chanakya are the complete outliers here. Range for BJP is 8-20 and Congress is 7-17. Huh!

Andhra Pradesh (Seemandhra + Telangana): This is another comedy of errors state. 3 / 5 channels say BJP / TDP will win, while 2 / 5 say Jagan’s YSR Congress will win. Utter confusion. Again here Times Now / H. Today are outliers. Range for NDA is 12-18 and YSR is 8-18. Can’t make anything out of it.

Gujarat: The state is clearly Modi-fied.

W. Bengal: All polls point to a Mamata TMC sweep. Range for TMC is 20-28 and Left is 4-16.

Rajasthan: All polls point to a BJP sweep. Except for our favorite Arnab who says Congress will get more seats than BJP. Laughable.

Bihar: All polls point to a BJP lead. ABP-News shows a slender lead. Again Times Now is a blooper here. 10 seats for JD(U) / Nitish. Range for Congress 2-10, JDU 1-10.

Odisha: The comedy of errors state returns. 4 / 5 pollsters say Naveen’s Biju Janata Dal will lead while 1 / 5 says BJP will get the most seats. Range for BJP 2-10.

Tamil Nadu: All polls point to a AIADMK lead. But again differ on seats. Range for AIADMK is 22-30, DMK 5-12 and NDA 0-7. What to make out of it?

Kerala: All polls point to a UPA lead. But again differ on seats. Range for UPA is 11-17 and Left is 3-9.

Punjab: The comedy of errors state returns. 3 / 5 pollsters say NDA lead while 1 / 5 says UPA and 1 / 5 says NDA & AAP will get 5 seats each. Ranges are bizarre – NDA 5-8, UPA 3-7 and AAP 0-5. Uff!

Assam: 2 /3 polls say BJP will get maximum seats while 1 says BJP and Congress will split honours.

4 states are clear comedy of errors, no clue who is right, who is wrong. In 6 states there are severe range issues. Truth will come out on 16th so lets wait.

The pollsters take sample from the same set of people who voted in a constituency, views of a small sample is extrapolated for the entire voting universe. And then if all of them have different results, then it raises questions about accuracy and brings in the element of margin of error. And all of them barring News 24 have more or less same results. Lets wait for actual results on 16th.

All these surveys have a stated margin of error of 3%. In 2009, 113 / 543 (20%) seats were decided on a victory margin of <=3%. This time due to multi cornered contests in many states plus high fever campaign plus Modi wave (real / created) the seats were margin of victory will be <3% will minimum double for sure. This time for many seats victory margin will be <15,000 votes (76 seats in 2009).