With the elections looming next year we have scores of
pundits pontificating on polls on primetime. So I thought, why don’t I give it
a shot. After all given their past records in forecasting, I can hardly do
worse J
Anyway, I am not a statistician and given this handicap (or
is it a strength?), I am not entirely convinced that a sample size of a 4-5,000
(a dozen or so respondents per constituency) will given prediction with any
credible degree of accuracy. Imagine the same pool of a few thousand broken up
into different constituency, different communities, different age groups,
different age groups, different income groups etc etc. So how accurate can a
statistical prediction be? I frankly don’t know. To my discredit of course.
What I have attempted to do is a seat-by-seat prediction
across states. The calculation can be described as making some educated (and
some uneducated) guesstimates of factors such as incumbency, increase in voting
percentage, the Namo swing factor etc etc. In fact, the genesis of this
calculation is the assumption that 2014 polls will be the most presidential
style elections in India
Let me first spell out the seat predictions, state-wise:
Uttar Pradesh 52
out of 80
Madhya Pradesh 25
out of 29Karnataka 22 out of 28
Jharkhand 12 out of 14
Rajasthan 11 out of 25
Chattisgarh 10 out of 11
Haryana 04 out of 10
Himachal Pradesh 04 out of 04
UTs+ NCR 03 out of 14
Orissa/ Odisha 03 out of 21
Uttarakhand 02 out of 05
Arunachal Pradesh 01 out of 02
J&K 01 out of 06
Kerala 01 out of 20
Total 215 out of 454
IBTL prediction 163
out of 543
Times Now Prediction 162
out of 543(based on IBTL tweet on 16th Oct 13)
Andhra Pradesh (42 seats), Tamil nadu (39 seats) and north
eastern states will be extras. I have not estimated these seats given the lower
chances of BJP here and paucity of time for me. Anything here will be a bonus.
I have assumed a 5% Namo vote share swing in Uttar Pradesh, but if this was to
be 6% the seats would rise to 54 and if it was 10% the seats would be 65. But I
have left it at 52 for the sake of being conservative. Notably, I assume BJP enters no alliances in UP pre poll.
I assume BSY will merge with BJP in Karnataka. The key point
to be noted is that given the ‘quant’ nature of the estimates, these estimates
make more sense when taken overall. For example, there are a lot of people who
believe BJP will sweep the NCR in the Lok Sabha polls. However, my quant model
does not give such results, hence I have given only 1 seat to BJP here. Similarly
my estimates of MP and Karnataka may seem on the higher side compensated by
Rajasthan and Gujarat that seem lower. But I
have to apply the same logic uniformly to be sure I don’t unfairly penalise or
give benefit to BJP.
While I do have some interesting observations, I do not have
the time to detail them out now. One statement though: BJP has in the last 10
years given away a lot of easy seats to allies and even in a few places it was
present it seemed comatose. Orissa for example where BJP used to win in several
seats with a massive 30-50% vote share is now a distant dream for the party. West Bengal where BJP contestants get about 2-6% share of
votes in virtually every seat has been ignored, as has the southern states. In
fact, it seemed that the BJP leadership wanted to convert it into a JDU style
regional party. However, it is my assumption and hope and prayer that the
advent of Namo will change this comatose and lazy style of functioning and
recharge the cadre to a national level.
As you can see under the table presented (above), my
prediction is a more decisive than other polls (IBTL and Times Now) that appear
to mirror each other. Are they too afraid to stray from consensus or am I plain
wrong…only time will tell.
No comments:
Post a Comment