A Critique of IndiaToday’s Post-Poll Surveys of Tamil Nadu, Kerala and Assam

Publication of exit-polls is banned till May 10^th 5 pm by the Election Commission (EC). But the EC notification did not mention anything about post-polls. IndiaToday was the first media group to circumvent the EC’s exit-poll ban by publishing a post-poll survey last night.

So what’s the difference between all these polls?

Pre-poll or an Opinion Poll is a survey conducted before the actual voting designed to understand voter intentions. Their major limitations are that intentions might change during and/or because of the campaign; and secondly, all eligible voters do not come to vote.

Post-poll on the other hand is an estimate who they voted for. There are basically two kinds of such surveys - exit polls and post polls. Though both polls give an indication on the pattern of actual exercise of voter choice, the essential difference between them is that while exit polls are undertaken on polling day, usually immediately after voters have voted, post poll are undertaken 1-2 days later. From a sampling perspective, a post poll is considered more accurate than an exit poll because while the later is self-selecting, the former, like pre-poll, we have better control over the sampling.

 

The IndiaToday post-poll surveys have a margin error of +3%, which means if the same procedure is used a large number of times, 95% of the time the true population average will be within the 95% confidence interval of the sample estimate plus or minus 3%. The margin of error can be reduced by using a larger sample i.e. if margin error is to be reduced to  + 1% a sample of at least 10,000 people is required whereas IndiaToday’s sample base was around 6,000 or less, depending on the three states it covered viz. Tamil Nadu; Assam and Kerala.

However sample size by itself is inadequate to get exit or post polls right; we need to get the sample profiles right too!

TAMIL NADU

 

IndiaToday’s post-poll indicates almost a dead heat, though the DMK alliance commands a marginal edge. The fact that the DMK alliance increased its vote share viz-a-viz both the 2006 Assembly and 2009 Lok Sabha does appear to suggest it commanded a pro-incumbency vote. However, if the survey’s + 3% margin error is factored, all outcome scenarios are still possible between these following two extremes and it would be too premature on the part of DMK supporter’s to open the champagne bottles:

 

As compared to 2006 Assembly performances , both alliances gain in vote share at the cost of ‘Others’, minor parties many of whom in 2011 have decided to align themselves to one of the two major alliances in the state. BJP’s hopes to open its account in the state according to IndiaToday’s projection appear to be dashed.

If IndiaToday’s pre-poll and post-poll projections are compared, it appears that just the shifting of 3% of votes accounts as the only difference between their two polls, though its impact on seat projections appear more pronounced as a difference.

So how does IndiaToday explain this difference?

- During the course of the campaign, rural voters decisively swung in the favour of the DMK and according to IndiaToday, this may turn out as the most decisive factor.  “During the Opinion Poll, 45 % amongst the rural voters intended voting for the DMK+. But after the elections, 50% have indicated they have voted for the DMK +, and 48 % for the AIADMK +. For years large subaltern support base formed the DMK's backbone. Its loss was seen as the party's Achilles heel. But a 5% shift in trend in rural Tamil Nadu in the last three weeks could turn this election on its head.”

- The DMK grabbed most of the young and Dalit voters, who AIADMK was expected to attract.  It follows that Vijaykant’s DMDK failed to deliver the youth vote to the AIADMK alliance and the DMK succeed in neutralizing the Vijaykant factor through deploying the Vadivelu, the popular film comedian. Through the dalit party, VCK and the possible sympathy factor arising from the arrest of ex-Telecom Minister Raja, a Dalit leader, DMK managed to carve out a base within the community, which it never had substantially.

- Though the vote margin is hardly 1%, the disproportionate seat distribution is hinted by IndiaToday as attributed to wasted votes by AIADMK, winning seats by large votes in urban areas, instead of their vote share being more evenly distributed within the state, where their surplus votes could convert many seats lost into winning ones.

- DMK’s campaign and poll machinery was more organized. The campaign itself was felt being livelier judging from the crowd response it provoked. In contrast, except for Jayalaithaa’s personal campaign, other members of the alliance proved a damp squib. 

In conclusion, the IndiaToday’s post-poll indicates it is too-close-to-call verdict. This is climb down for the media group after projecting a runaway victory for the AIADMK.

But looking at all these data, I asked myself what could be my hunch as the final outcome? All things considered, I am inclined to believe that a DMK sweep is the most probable outcome, a view which I am prepared to revise once exit poll results from other media groups are made available on May 10th. 

  

KERALA

The IndiaToday post-poll re-confirmed their pre-poll prediction of an UDF victory in Kerala on a margin of 4%.  If true, the UDF wins despite a - 1.38% swing against it viz-a-viz 2009 Lok Sabha elections, when it led in 98 assembly segments with the vote share of the Left remaining stagnant.

However, the appeal of UDF among younger voters is dropping. Of significance is the Muslim vote consolidating strongly around the UDF. This is a blow to the Muslim extreme outfit Jamaat-e-Islami who extended its support to the LDF in 124 constituencies though it supported the UDF in 15 constituencies. This when taken together with contraction of the ‘Others’ category from 21% in 2006 to just 11% projected this year, suggests that the chances of BJP opening its account in the state does not look bright. This category of ‘Others’ include minor parties and independents and also the BJP. Of particular heartburn to the Left is that they failed to cultivate the numerical strong Ezhava backward caste in the state, despite Achuthanandan, their Chief Minister belonging to this community.

So how does the IndiaToday post-poll compare with its own pre-poll and those carried out by Asianet?

We find that IndiaToday post-poll has slightly scaled back both the vote share and seat share of the UDF and more in line with the second pre-poll conducted by Asianet. This is consistent with the popular view that the Left conducted a very strong rear guard campaign with Achuthanandan ending up  the star campaigner of the entire election  drawing  crowds much larger than even Sonia and Rahul Gandhi. However, UDF supporters are less likely to attend rallies than LDF supporters though they vote as enthusiastically as their Left counterparts. 

Accepting the IndiaToday’s post-poll projections as it is, even so, a 4% vote lead should translate into at least 100 seats for the UDF. 

ASSAM

The IndiaToday post-poll projections for Assam indicate that the state is heading for a fractured verdict, with the Congress retaining its status as the single largest party. While the vote share for the Congress and the BJP remain unchanged from 2006, those of the AGP, the principal opposition party in the state and AUDF, a Muslim minority party are poised to increase by 3% each. Their gain is offset by a 6% swing away from the ‘Others’ category.

IndiaToday observed that the Congress remain favourites to form the government but have to choose between their  existing ally Bodoland People's Front (BPF) and pro-minorities outfit - AIUDF to hang on to power. And yet, the BPF finds no separate mention in their tabulation, being clubbed under ‘Others” . It also hints that with the AGP and BJP having a tacit poll understanding, this  could queer the pitch for the Congress to forming the government.

But IndiaToday fails to mention that the Congress and the BPF also had a tacit understanding and in fact allies in the outgoing government. If the vote share of AGP-BJP adds up to 35%; then those of the Congress-BPF add up to 43%. Similarly the combined seat share of AGP-BJP is projected by India-Today to be 46-53 seats, those of the Congress-BPF would be between 54-62 seats.  The fact is the BJP, AUDF and BFP command a similar vote share size viz around 12% and expected to each get around 12-14 seats.

So how does IndiaToday’s post-poll compare with their pre-poll survey? The main difference lies that the post poll scales back the vote shares of the Congress, AGP and AUDF as projected in their pre-poll survey and increased those of the ‘Others’ as much as 6%! Whether IndiaToday has botched up either or both their surveys, we would know on April 13^th.

For Assam, this blog remains confident that a Congress-BPF coalition government will take reign of the state.