Tuesday, May 10, 2011

On why the Asianet-C Fore got it wrong with opinion polls

Asianet's post-poll survey vindicates my critique of their second opinion poll, made in an earlier post in this blog.

This is about what went wrong with the Asianet-C Fore opinion polls telecast on 31st March 2011. Asianet-C Fore's second opinion poll had stated the following on vote shares:
  • UDF: 46 per cent
  • LDF: 41 per cent
  • BJP: 9 per cent
  • Others: 4 per cent
Based on this, they predicted that:
  • UDF will win 80-90 seats
  • LDF will win 50-60 seats
  • BJP will win up to 2 seats
Now, I had argued that these results are flawed, because there probably was a major bias in sampling. This bias was evident when we looked at community-wise vote shares of 2011 and compared them with 2004 and 2009 post-poll surveys conducted as part of the National Election Study (NES) of Lokniti in New Delhi. In comparison with the NES surveys, the community-wise shift in votes in the Asianet-C Fore survey were extremely surprising.

Now, the post-poll results of the same Asianet-C Fore group (telecast today) show very different trends. Here are some quick numbers that are avaialable now. I will keep filling up the gaps tonight.

(Click to enlarge)
See the comparisons shaded in orange. Compared to the opinion polls of Asianet-C Fore, the vote share among upper caste Hindus of  the UDF has fallen from 60 per cent to 46 per cent. Among Muslims, the UDF vote share has fallen from 70 per cent to 50 per cent. Among Christians, the vote share of UDF has been stagnant.

In other words, the 2nd opinion poll of Asianet-C Fore had, indeed, a bad sample; they had probably oversampled from upper caste Hindu voters, who traditionally favour the UDF. My point in the earlier post was the following:
"How did such a large share of Hindu upper caste voters decide to vote against both UDF and LDF, and in favour of Others, just over a period of one month? Was there a deliberately induced bias in the second survey?

All these unexplainable trends show that serious questions can be raised regarding the randomness of the sample used in the Asianet-C Fore survey."
It appears now that Asianet-C Fore have now accepted that it was a terrible sample they had for the opinion polls. Thus, Asianet-C Fore have now, in their post-poll survey results, revised down the UDF tally from 80-90 seats to 72-82 seats. The LDF tally has risen from 50-60 seats to 58-68 seats. The reason for this is that due to a probably better sample this time, the vote share of LDF has risen from 41 per cent to 43 per cent. The vote share of UDF has fallen from 46 per cent to 44 per cent.

Now, let us move into the trends of the CSDS surveys. The final CSDS results are yet to come in, but we have some inidcators. These, a comparison of 2009 and 2011, are in green shade. First, a good movement of Ezhava votes into the LDF side; and secondly, a small fall in the UDF vote shares among Muslims and Christians.

So, if there is a wave in CSDS surveys, it is a small one towards the LDF. Is it enough in terms of seats to carry it beyond 71? They say, probably Yes. However, they have chosen to call it a too-close-to-call verdict, and given the LDF 46 per cent votes and 69-77 seats. UDF gets 45 per cent votes and 63-71 seats. There are no error rates given, but I trust Rajeeva Karandikar here. It is indeed too close to call. 46/45 vote share is indeed a dreadful sight for a statistician.

So, if there is a wave towards the LDF, what is its objective basis? The CNN-Week-CSDS poll argues that it is the development-record of the LDF government. They say:
"The edge that LDF seems to enjoy is because of the good work it has done in the field of electricity supply, drinking water, quality of education in government schools and medical facilities in government hospitals. While an overwhelming 83 per cent of voters are satisfied with the Achuthanandan government on the issue of electricity; drinking water (74 per cent satisfied), education in government schools (73 per cent satisfied), facilities in government hospitals (72 per cent satisfied) give the Front some breathing space vis a vis the UDF.

Even of the law and order front the government has done a decent job with 70 per cent of voters of giving their thumbs up despite the rise of fundamentalist groups in the last few years."

See their detailed results below.

2011 Kerala Assembly Elections Exit Poll Results
2011 Kerala Assembly Elections Exit Poll Results
2011 Kerala Assembly Elections Exit Poll Results

All figures from http://ibnlive.in.com/news/kerala-surprise-ldf-ahead-by-a-nose/151892-37-64.html