THE Tamil Nadu election results have taken everyone by surprise. This state has once again proved to be electorally volatile and swung decisively in favour of the BJP-AIADMK alliance. The extra 25 seats or so seats that this alliance gained from the DMK-TMC combine, compared to our seat forecast published in Outlook, has made all the difference in our seat prediction at the national level, which otherwise would have been fairly accurate.
One must remember that the fieldwork for this poll was conducted between January 24 and February 1, 1998, when the new BJP alliances in West Bengal, Orissa, Karnataka and Tamil Nadu were just taking shape. In fact, in all these states our poll, carried out nearly three weeks before the elections, failed to pick up accurately the impact these alliances would make in the voting pattern. The sweep by the BJP-AIADMK alliance in Tamil Nadu, in particular, eluded every opinion as well as exit poll.
The Outlook poll conducted in end-January did indicate a swing towards the alliance of the order of 7 per cent, but given the poor performance of the BJP and AIADMK in 1996, the swing was much less than the 35-plus per cent required to account for the final results. Our poll in January was a snapshot of how people felt in Tamil Nadu at that time, but obviously the swing gathered momentum over the following three weeks.
The problem every pollster (who conducts polls so far in advance of the election date) faces is how much to project the swing forward into voting day. In the past, we have done this quite successfully, as there was an issue which we could justify as the basis of the forward swing—for instance, the anti-Jayalalitha wave in 1996. But for Tamil Nadu in 1998 there did not seem any particular issue to justify a forward projection.
The author is MD, AC Nielsen
