An analysis of UP exit polls indicate that results could be diametrically opposite

Friday, March 10, 2017



Ask anybody. Last evening whatever little faith people had on pollsters, mostly evaporated almost totally after watching the slew of exit polls published yesterday. Even Rajdeep Sardesai who anchored one such programme on IndiaToday and Shivam Vij, Dy Editor of HuffPo appeared to have given up on the absurdity viewers were subjected to last night in the name of exit polls. 

Most of these polls bandied around seat shares, while only a handful were prepared to give details of their vote shares or survey internals.  While it is difficult to undertake a critique of these polls in all 5 states whose assemblies went for election, due to constraints of time, we limit our critique to UP as found below. Time permitting we would try to critique exit polls of other states later today..


To note CVoter is an Exit Poll while both CSDS & Today’s Chankaya (TC) are post polls. As observed from the table, the commonality among the three polling agencies projections is that the main contestants are between SP-Congress combine and BJP, with BSP coming in third position. 

The second commonality is the vote share estimation of the SP-Congress Alliance by all the three polling agencies is 32-33%. So a consensus exist that the Alliance is up 2-3% from 2014 and in doing so, demonstrating positive momentum with them. 

BSP in 2014 registered a tad less than 20% Accordingly, all three polling agencies acknowledge that BSP too is a net gainer by vote share, with CVoter & CSDS projecting the magnitude around +5% while according to TC, net gains are marginal viz around 0.5%. 

The real difference among the polling agencies comes to projections of BJP’s vote share. CVoter & CSDS projects this in the region of 32-33% - this reflects a 10% negative swing from 2014 or in other words, the BJP is experiencing strong negative momentum. The 1.1% difference between CSDS & CVoter is not really statistically significant as it comes within both polls Margin of Error (MoE). Whereas TC projects BJP at 42% - practically no change from 2014 Lok Sabha polls where NDA registered 42.3% So the difference between TC and CDSS-CVoter is basically accounted by TC squeezing BSP & ‘Others’

The TC projections are simply flabbergasting for two major reasons as elaborated below:


1. To reach 240-250 seats, the BJP has to maintain a strike rate of 60%+ for every phase, including the first two phases in Western UP where a very perceptible anti-BJP wave among the Jats was sensed. Further ground reports from different journalists suggest that the constituency level fights were very intense that such strike rates as TC projects would immediately raise suspicions of EVM rigging that would cause huge public uproar over the fairness of the polls. 

2. The internals of the TC survey strikes immediately as totally contradictory to the numbers TC projected and have led to widespread suspicion expressed in social media that their numbers had been crudely manipulated.

3.  By and large both pre-polls and exit/post polls have tended to over-estimate the BJP and under-estimate the opposition particularly the BSP. Congress spokesperson Sanjay Jha insists that the Alliance do not even see BJP crossing 30% vote threshold.  Accordingly, a conservative statistical adjustment for treatment of such biases could see the Alliance cruise comfortably pass the finish line with 190+ with BJP trailing far behind at 120 seat levels. See earlier table columns marked in yellow.

Shivam Vij in a totally captivating article today in HuffPo observed “In such an election, the BSP is the X factor. It is the BSP's performance that is the most unpredictable, and the most fortuitous. The more votes the BSP took away, the more it hurt the BJP, though in many places it seemed to be hurting the SP-Congress too.” Read full article HERE
This blog totally share this view of BSP being the X factor except that we feel the more BSP does well; it tends to more hurt the BJP rather than the Alliance.




You Might Also Like

0 comments

Like us on Facebook

Follow us on Twitter