Analysis & Predictions for TN LS 2014 Polls

I am a firm believer that sampling vote shares at state level and then extrapolating it to wins for each party – does not work well in the “first past post” Indian system. This post is my attempt to gauge constituency wise mood to arrive at seat predictions for Tamilnadu in the upcoming LS polls 2014; where people get to cast their vote on 24^th April 2014.

Several leading regional media houses have published constituency wise survey results. I am going to triangulate these sources, plus apply my own observation from several field reports to arrive at the predictions. Warning: Such triangulation by heuristics have systemic biases which can affect reliability of results; however this is the best one can do, when he/she doesn’t have access to the primary data to do a formal statistical Meta-Analysis of the results. The sources that I referred to are:-  

• Kumudam surveyed a record 66,100 respondents from all 39 constituencies of Tamilnadu. The demographic breakup was 62% male – 38% female; which is having a larger male skew. Their demographic sample included 11% first time voters, 13% school dropouts, 32.7% graduates+. Again they have a higher skew of graduates. They partnered with Statistics professor from Madras University T. R. Gopalakrishnan.  
• ThanthiTV surveyed 20,000 respondents from all the constituencies. They have not shared the demographic breakup, however have stated that all the standard precautions in stratification of sample have been taken care. Their partner was Kris info media.
• Thuglaq weekly surveyed 10,000 respondents from all 39 constituencies. They did not follow a statistical process, but they use a qualitative methodology to interpret results based on findings from the field. Their results were fairly accurate in the 2011 assembly polls giving it some credence.
• Nakkeran magazine also did its predictions – but I summarily discount them as a source – given their political lineation and previous track record of complete failure.
• Vikatan has done a survey of 27,000 respondents and given projected state level vote share. However since constituency wise break-up & demographic break-up is not available, I am constrained to not use this as a source. However I am using the field reports filed in Junior Vikatan.

  

Summary

Region	Party	> 70% Probability	50 to 70% Probability	Low Margin  Close Call	Total
Chennai (4)	ADMK		1		1
	DMK	1		2	3
	NDA				0
North TN (10)	ADMK	2	2	1	5
	DMK		3	1	4
	NDA			1	1
Delta (6)	ADMK	1	2	1	4
	DMK	1	1		2
	NDA				0
Kongu (9)	ADMK		1	3	4
	DMK		1		1
	NDA		1	3	4
South TN (10)	ADMK	1	2	3	6
	DMK			1	1
	NDA	2	1		3
State (39)	ADMK	4	8	8	20
	DMK	2	5	4	11
	NDA	2	2	4	8

So the predictions are ADMK - 20 , DMK - 11 and NDA - 8 

The above table takes into account our stab on who would win each of the 16 constituencies that were classified as close calls. If our predictions on closed calls go the other way, then there are upside and downside possibilities for each of the alliances. Accordingly we provide the forecasts as range of seats below. These can be treated as safe bets of the likely electoral outcome.

ADMK - 12 to 28

DMK – 7 to 16

NDA – 4 to 13

For the complete analysis including seat-wise mood meter & predictions, please read the full article hosted on CRI website here