It's going to be close, and only you know how close

Forget polls. The election will be decided by ambiguous, confused, even ashamed voters

I would have noticed: the Cabinet minister who preceded me out of the restaurant had not slipped a tenner to the waiter. But the waiter swivelled as the Tory left and hissed, "Good luck to you on Thursday - whichever Thursday it is". Over lunch, the minister had been explaining why he thought the Tories could well win the election. I had listened politely, credulity suspended by the forkload.

But in the taxi back, the waiter's offhand comment reverberated. Conventional wisdom says that the polling gap is too large and too long for the Tories to really recover. Comfortable, liberal-minded Britain knows that the banal argument, ``time for a change'', is now irrefutable. In much of the City and the media, people have already made the mental adjustment: Tony Blair and Peter Mandelson are almost in power.

The problem, of course, is that conventional wisdom is so often foolish; that what liberal-minded Britain considers irrefutable has been so often refuted, and that the City and my profession are such bad futurologists. Perhaps there is too much hidden superstition in my argument. But there it is. The feeling lurks.

Can it be justified at all? Evidence of Labour invincibility is based, of course, on the polls, which have long shown a huge and consistent Opposition lead. This is the scientific age. Polling is practised with scientific seriousness. Pollsters gather in conferences, exchange information and discuss techniques. Often, they have been remarkably accurate. Their authority comes from their record, which was dented in 1992 but not demolished.

Yet polling is not science. It is asking questions and hoping for truthful answers. In the past, some of these answers haven't been - that is more or less common ground.

We believe that an unknown proportion of Tory voters lied to pollsters about their intentions. We strongly suspect that among those who refused to answer there were more Conservatives. We know that some recent polls overstate the anti-Tory vote: one highly-publicised one included another question about how people voted in 1992, which demonstrated pretty conclusively that Neil Kinnock won and has therefore presumably been in power for years. (It's just that nobody's noticed yet.)

My point is only that if people both dislike the Conservatives and also suspect that they may vote for them, noses pinched, to preserve a modest but long-established recovery, that behaviour would be peculiarly difficult for pollsters to tease out. This is not a frank country. These voters will feel ambiguous, perhaps confused, even ashamed. And if they haven't quite confessed to the bathroom mirror, they are unlikely to tell a pollster.

Though he didn't discuss the polls, this is essentially what my lunch guest thought was happening. He summed up the general mood as general fed-upness with his party, combined with a sneaking, half-resentful acknowledgement that things were getting better, and the Government might be partly responsible. I suggest that quite a few people feel that way.

Then there is ``agenda slide'', a new term which describes the difference between what the political classes are interested in - sleaze, constitutional questions, Brussels, intra-party divisions, hairstyles - and what the rest of the country cares most about - prosperity, safety, the environment and so on. The political agenda connects with the popular one via the media. At times they can seem virtually identical: sleaze dominates the headlines; Tory popularity sinks further. Ergo - the whole country thinks alike.

Except - er - it doesn't. However angry voters are about those headlines, they may regard them as only semi-serious - Westminster as farce - and actually vote on the basis of a subtly different agenda. If so, the more froth and turmoil in the papers about politics, the harder pollsters will have to search for slowly hardening convictions based on another, almost subliminal agenda.

These are, granted, impressions, not hard facts. But for me, they add up to a conviction that the election remains a closer call than many assume. It is the economy, stupid. But it is also politics, stupid. The choice is big. The game is wide open, and the rougher, hungrier set of politicians will win.

For the Tories, that means rediscovering discipline, and hiding their deep splits, while grabbing as much credit as possible for the recovery and painting Labour, in traditional fashion, as profligate and anti-patriotic federalists. If middle England, in particular, is feeling a touch more secure and prosperous, then even Blair could be made to seem a risk. These are genuinely conservative folk.

For Labour, it also, unavoidably, means the politics of fear - attacking the Conservatives as a party that now wants, in its heart, to leave Europe altogether (note how the word ``renegotiate'' has become a suddenly fashionable battlecry for the right).

It also means a far more aggressive approach to the social agenda and political reform than we have heard yet; Blair needs some raw roughhouse politics to help him escape from the Cassandra-style bitching and trivia of recent weeks. If it isn't time for a real change, it won't be time to vote the Government out.

I still think Labour's job the easier one and in the end, if they fight hard, that they are likelier to win. But this is a secretive, private nation. And in the past few weeks, there seems to me to have been a change in the mood, a sort of tremor in the air before the weather changes. In politics as elsewhere, it looks like a long and changeable winter.