And to think Tony got the nod over Gordon because he was popular

This is awkward, isn't it? The whole point of Tony Blair is that he was the answer to Peter Mandelson's question: "Who will play best at the box office?" Blair's claim, which he put patiently and insistently to Gordon Brown in 1994, was that he might not have such a rapport with the party but that he was more popular with the wider electorate.

This is awkward, isn't it? The whole point of Tony Blair is that he was the answer to Peter Mandelson's question: "Who will play best at the box office?" Blair's claim, which he put patiently and insistently to Gordon Brown in 1994, was that he might not have such a rapport with the party but that he was more popular with the wider electorate.

Only now he isn't. The opinion polls over the past week have been astonishingly good for the Chancellor of the Exchequer. The impact of Tuesday's NOP poll in The Independent may have been muffled because it was difficult to make a direct comparison with standard opinion poll figures. Let me set out the data. When people were asked for which party they intended to vote, Labour stood at 38 per cent, the Conservatives at 30 per cent, the Liberal Democrats at 21 per cent and other parties at 12 per cent. When people were then asked how they would vote "if Gordon Brown were leader of the Labour Party", Labour shot up to 50 per cent, the Conservatives were virtually unchanged on 29 per cent, the Lib Dems were down to 16 per cent and other parties down to 6 per cent.

These are raw figures, because The Independent did not follow up the Brown question with the "squeeze" question which asks the don't knows and won't says which party they are "most inclined to support". Nor are the answers to the Brown question adjusted to take account of what people said about how likely they were to vote, because a Brown leadership might persuade people to turn out who might otherwise stay at home. When these adjustments are made to the figures with Blair as leader, to produce results comparable with other polls, the Conservatives are two percentage points higher on 32 per cent and the "other" parties lower.

But the "raw" figures for Blair and Brown are comparable with each other, so a Blair lead of eight points over the Tories becomes a Brown lead of 21 points. That Brown effect is remarkable. It is twice as big as anything picked up by the opinion polls so far. The last time NOP asked the Brown question, in September 2003, it added one point to the Labour lead. Another polling company, ICM, found a four-point Brown bonus in May last year and a six-point advantage in July. But to increase Labour's share of the vote by nearly a third is extraordinary.

There are two ways to try to dismiss this finding. One is simply to say, as my very good colleague, Mr Watkins, does (opposite), that one does not believe in opinion polls. To which the reply must be that, in between elections, they are the worst way of measuring public opinion, apart from any other way.

Nor is it as if the NOP poll looks like a rogue. A poll by ICM for the News of the World last weekend asked this question: "Putting aside your own party preferences, which of the following politicians do you think would be the best Prime Minister?" The results were: Brown, 27 per cent, Blair, 26 per cent, Charles Kennedy, 15 per cent, Michael Howard, 14 per cent, none, 8 per cent, and don't know, 11 per cent. That is the first time for more than a decade that someone other than Blair has been the most popular choice for the top job.

The other way to dismiss Brown's apparent popularity is to say that it would not be like that if he actually took over. Perhaps not. It may be that Brown is benefiting from the halo effect of simply not being Blair. It may be that if he became prime minister he would quickly drive away those centre-ground voters who are attracted to Blair precisely because he seems so non-Labour. It is likely that many Brown- voters are disaffected former Labour supporters in safe seats whose return to the fold would make no difference. But the Brown effect is so marked that it is impossible to wave it away.

This spells trouble for the Prime Minister. It was central to Blair's politics that the Labour Party must be prepared to do whatever it takes to attract every possible vote - or to avoid repelling any possible vote. If the Labour Party faces a choice between a small majority under Blair or a large majority under Brown, then surely Blair is obliged, at last, to step aside for the one who will "play best at the box office".

There is, therefore, only one possible explanation for Blair's failure to step down and solve the problem of letting his Connaught Square house by moving into it himself.

It is that he has decided that he is a better judge of what is in the national interest than both the voters and Gordon Brown. He has decided that fighting the next election on an "unremittingly New Labour" manifesto is more important than maximising votes. Never mind whether Brown said he cannot trust Blair, the only conclusion to be drawn is that Blair does not trust Brown to carry the New Labour torch. The transformation is complete. From being all things to all people, he is now a conviction politician who believes that his principles are more important than mere electoralism.

That explains why, beneath the familiar, conciliatory manner, his interview with David Frost last weekend was so aggressive. He said what being unremittingly New Labour meant: a manifesto that will "drive through the market-based reforms in the health service [and] independent state schools". Those policies are brave and right, but they are not popular.

Not only that, but the problem for Blair is that they have barely begun to be implemented. His argument has to be that these unpopular policies will deliver better services and that people will come to realise that. But time is running out. Most people regard market mechanisms in the NHS as a form of privatisation, and independent state schools as a meaningless oxymoron. I think both are necessary if broadly egalitarian public services are to survive in the long run. It is a pity that, after eight years, Blair and Alan Milburn are so far away from winning the New Labour argument. And an even greater pity that Brown is so successfully undermining them by playing to the other side of the gallery. His "market bad, public service ethos good" message is popular, but simplistic.

This election campaign is going to be most peculiar. It will be fought by a prime minister who is less popular than his rival, on policies that are less popular than those of his rival - policies, moreover, that are still in an early stage of development and testing. Yet, because his rival is in the same party and same government, the voters are largely shut out of the argument. Because the other message of last week's opinion polls is that the Blair-Brown split has not damaged Blair's chances of winning again and the official opposition is still nowhere.