[Reader-list] Why opinion polls are often wide of the mark

[Asit Das]

Why opinion polls are often wide of the markAtul
Thakur<http://timesofindia.indiatimes.com/toireporter/author-Atul-Thakur.cms>
,TNN | Apr 13, 2014, 02.43 AM IST

NEW DELHI: Going by the media coverage and the results of various opinion
polls, it might appear that Narendra Modi's ascent to power in Delhi is
almost certain. But an analysis of past five general elections shows that
opinion polls tend to overestimate BJP, underestimate Congress and at times
go very wrong in assessing regional and smaller parties.

For instance, in 2009, pollsters predicted a near neck-and-neck fight
between Congress-led UPA and BJP-headed NDA. It was a general consensus
among various poll pundits that although UPA had an edge over NDA, the
latter could not be written off. If we take the minimum and the maximum
seat forecast of opinion polls, conducted just before the elections, it
works out that UPA was supposed to get between 201 and 235 seats while the
corresponding numbers for NDA was between 165 and 186 seats. The individual
forecast for India's two largest parties ranged between 146 and 155 for
Congress and from 137 to 147 for BJP. The actual results surprised everyone
— pollsters, media and parties. UPA got 262 seats while NDA was confi ned
to just 157.

Party-level predictions also went completely wrong — Congress went ahead to
win 206 seats while BJP could manage to bag just 116 Lok Sabha seats. For
BJP, the 2009 results was perhaps a scaled-down replica of the 2004
debacle. After the good showing in assembly elections in four states, the
NDA government decided to go for early polls.


Initial opinion polls suggested that NDA was going to sweep the elections.
According to poll results of India Today-ORG Marg, published in February
2004, NDA was supposed to get 330-340 seats while Congress and allies were
to be confined to 105-115. Later the predictions were revised to 260-286
for NDA and 154-180 for Congress and allies. Then PM Atal Bihari Vajpayee
became so confident by what the opinion polls had to say that in an
April 17rally at Nagpur (three days before the elections), he
expressed his
discomfort with the coalition government and said, "My worry now is, if we
are again saddled with a 22-party coalition... such a situation is better
avoided." Perhaps he was indicating that it would be easier to govern the
country if his party secured a majority. The results, however, shocked BJP
as NDA was reduced to 185 seats, while Congress and allies won in 217
constituencies.

In the 1999 elections as well, the initial poll predictions suggested that
NDA would emerge a clear winner. An opinion poll conducted by DRS for TOI
for the August 5-9 period — about a month before the elections — suggested
that NDA would get 332 seats while Congress and allies would manage only
138. India Today-Insight opinion poll also confirmed TOI's prediction,
giving 322-336 seats to NDA and 132-146 to Congress and allies. Both
opinion polls suggested that BJP and Congress would increase their
individual tally by cutting into the votes of regional parties.

The newspapers were full of stories about a "Vajpayee wave", which was the
focus of BJP's election campaign. A news report based on TNS MODE survey
even suggested that Sushma Swaraj was going to beat Sonia Gandhi in
Bellary. The estimates were later revised to 279-336 for NDA and 132-162
for Congress and allies — much closer to the actual results. Swaraj,
however, lost from Bellary and state parties won 162 seats.

Opinion polls for 1998 and 1996 elections were, however, closer to the
actual results. Why do opinion polls tend to favour a particular coalition
while underplaying others? First, unlike Western countries, India's
population is not homogeneous — caste, religion and region play an
important factor in elections. Also, people, particularly from the weaker
section of society, are reluctant to reveal which party they are going to
vote. BJP might be getting overplayed because the main support base of the
party is the middle class — a section which is not afraid of any
post-election harassment and hence they are vocal in sharing their views.
On the contrary, key voters of other parties are from vulnerable sections
of society and hence they are afraid to share their opinion.

Experts also argue that opinion polls work much better in bipolar
elections. In India, where elections are multi-cornered, it is possible
that pollsters fail to estimate the strength of regional parties.