PollsterGeist

8216;8216;One plus one is not always two,8217;8217; sighs Partha Rakshit, managing director, A C Nielsen Research Services, peering at a jigsaw of numbers on his laptop. All around him in the plush office in sleepy Chinchpokli, south Mumbai, 64 teams jam phone lines to call in sample data from exit polls as the day8217;s voting end.

When the telephones stop ringing, there will be 90 minutes left. To collate, check, re-check, analyse and air results on prime time television.

If all democracies are obsessed with forecasting election results, India is a late bloomer. In the year before the 1988 elections in France, 860 opinion polls were conducted. In Britain, before the 1992 elections, newspapers had 60 poll forecasts. Not just nationwide forecasts but even candidates8217; fortunes are tracked and reported on a daily basis.

In India, many more than just the familiar Prannoy Roy team are sticking their necks out, predicting figures and fortunes. The 2004 elections, though, are more remarkable for the impact they have had on 8212; not just on sleepless nights of the central figures, but on the stock-markets, the party campaigns and political journalism. Welcome to the world of super reporters 8212; a hybrid of number crunchers and political pundits.

With the stakes so high 8212; go wrong, and the humiliation is public 8212; many market research organisations stay away from the election business. The forecast industry is dominated by a handful who dream, think and eat elections. The game has never been more exciting than now, when market research conglomerates supply the data from large constituencies and larger sample sizes for the psephologists to mull over, chew and spit out in the form of election forecasts.

Take A C Nielsen. Election research makes up less than three per cent of their annual turnover; retail research, television audience measurement and customised research for companies take care of the rest. In fact, MD Rakshit8217;s empirical model of election forecasts was the trigger behind their first foray into election forecasting anywhere in the world.

Even then, they have provided only the forecast for seven states for The Asian Age. For The Indian Express-NDTV poll, they did only the fieldwork, while Roy8217;s team 8212; a sociologist, a political specialist and a statistician 8212; decided the questionnaire, the constituency size, the profile of interviewees and interpreted the data with their own methodology.

Entering a newly liberalised India in the 1990s, A C Nielsen acquired a comparatively small company called MRAS with a strong client base. Even as it was eyeing ORG-MARG 8212; known for its strength in retail auditing 8212; the company was sold to Dutch giant VNU.

Last year, ORG-MARG became part of the A C Nielsen network of market research companies in over 100 countries. But the merger is not yet complete; the companies operate out of different offices and have independent clients India Today for ORG-MARG and The Indian Express-NDTV for A C Nielsen for poll work. Their modus operandi, they claim, is as distinct as Coke and Pepsi. Even if the mother company for both is VNU.

8216;8216;We compete, we have two different wings,8217;8217; says Rakshit.

G V L Narasimha Rao 8212; formerly with ORG-MARG and now head of his own company Development and Research Services DRS 8212; though, pins his faith on 8216;8216;adaptation8217;8217; and 8216;8216;homegrown solutions8217;8217;, rather than imported projection software.

Besides an in-house team of investigators for data collection, Rao8217;s core team has an indigenous formula for interpretation. He has conducted opinion and exit polls for The Times Of India, exit polls for Doordarshan in 1999, 2000 and 2001 8212; said to be the only exit polls in those times 8212; forecasts for Zee News and opinion polls for Sahara News during the 2003 Assembly elections.

With 500 parliamentary constituencies, they claim they have the largest sample size among Indian pollsters. 8216;8216;A wider sample is better as each constituency behaves differently in India,8217;8217; says Rao, who is working on 20 surveys simultaneously.

The election fieldwork for DRS is managed by Parimal Kumar Singh. Subordinate to him are 500 field supervisors, who are supposed to backcheck the data collected by close to 2000 investigators. 8216;8216;Regular graduates from semi-urban backgrounds are the best for data-collection. They manipulate less,8217;8217; says Singh.

Despite checks like mobile-calls and signatures of polling agents in case the projections are sued to ensure the agents are actually going into the field, 8216;8216;there are slip-ups8217;8217;, admits Rao. 8216;8216;But we catch them. If there are aberrations, we call them for explanations. If they haven8217;t been on the field, they are not able to explain their results.8217;8217;

Once the data is verified, it is fed into a company-developed software and the results are analysed by DRS director Diwakar Srivastava. Rao, with his political acumen, takes the final call on three probable scenarios.

8216;8216;We began with an issue-based poll, which helps us understand why we are getting a particular data. For example, contrary to what I think, the Telgi issue is not important in Maharashtra. So it is in sync with our field data results, which show a strengthening of the NCP-Congress alliance,8217;8217; explains Rao.

C-Voter8217;s Yashwant Deshmukh, a former journalist, has been involved with poll predictions for close to a decade. He has a full-time staff of about 15 and takes on 2,000 temps during elections.

8216;8216;The industry has not matured in India,8217;8217; he says. 8216;8216;There8217;s no R038;D work. Electorates do not behave the same way in the Lok Sabha and assembly polls. So the lessons we learnt in, say, the Rajasthan assembly elections, can only be tested in the next assembly election.8217;8217;

While most organisations take vote share calculations in their stride, they agree that forecasting seats is nothing less than hazardous. Small swings can cause big changes in votes. The methodology used to convert vote shares into seats is where the real challenge lies, and each agency guards its formula jealously.

Roy, for instance, subscribes to the theory of Index of Opposition Unity IOU, which takes into account vote percentage of the dominant party and of the largest opposition party. The method has its critics, with some saying the formula doesn8217;t work since every state now has its own multi-party system. The debate on the best formula is endless.

There is debate, also, on the ethics of making exit polls public before the elections conclude. In the US, unlike in India, results are not revealed till the last vote is polled.

To ensure some standardisation and quality control, the big three in the Indian market 8212; A C Nielsen, ORG-MARG and DRS 8212; are affiliated with ESOMAR, the European Society for Opinion and Marketing Research, an international body representing market research specialists.

Though silent on exit polls, ESOMAR has strict guidelines on how and where to sample, and demands transparency on who commissioned the survey, the sample size and the margin of error.

Despite these guidelines, some say forecasting is too much of a game and stick to analysis. CMS, for instance, refused to participate in election-related polls until after May 14. 8216;8216;We don8217;t want to be used by any party 8230; As if we are conducting its internal survey,8217;8217; says its director Bhaskar Rao.

But CMS has a media lab, with about 15 people, who survey and study 8216;8216;in detail8217;8217; election-related and media-related information from sources like newspapers and other polls. This information is used for election analysis, not opinion or exit polls.

Poll predictions, therefore, are not for the faint-hearted. 8216;8216;Poll research is so much in the public face, the risks are very high if it boomerangs,8217;8217; says Rakshit. But it8217;s just a matter of days before election frenzy subsides. Soon soaps and brand wars will bounce back on the television set.