No leader of the opposition has been as unpopular as Ed Miliband and gone on to win, I was told last week. No politician can survive in the public eye for longer than seven years (and David Cameron has now been leader of the Conservatives for eight), I was told in another conversation in the House of Commons.
The seven-year theory is rehearsed by George Eaton in the New Statesman this week. It was given credibility by Gordon Brown, who fretted in 2004 that he had already been chancellor for seven years and it was getting too late for him to be prime minister. "Once you've had seven years, the public start getting sick of you," Damian McBride reported him as saying. "After that, you're on the down slope."
This may be true, but the trouble is that it is no guide to the winning or losing of elections. Tony Blair had been prime minister for seven years by then as well – in fact, he had been on the public stage as leader of the Labour Party for 10 years – and he went on to win the 2005 election comfortably.
Nor is the counter-theory much use. Miliband is unpopular all right, but we are in an unusual period when the Leader of the Opposition and the Prime Minister are unpopular at the same time. Usually, when one is down, the other is up. At other times, if they are both a bit unpopular, the leader of the Liberal Democrats is up, but today all three of them are very unpopular. Which leaves Nigel Farage as the most favourably regarded party leader.
Anyone can play the game of the iron laws of politics. You might start with: "The Tories must lose: no governing party since the war has increased its share of the vote." Then someone says: "What about 1964-66 and 1974?" Those don't count, you conclude, and add to the rule: "... after a full parliamentary term". But then someone else will point out that the Conservatives increased their share of the vote between 1951 and 1955.
At this point the sensible person might conclude that the search for iron laws is a foolish one, and recall the golden words of John Habgood, Archbishop of York, who responded to a line of questioning by Jonathan Dimbleby in a BBC On the Record interview by asking: "Has it ever occurred to you that the lust for certainty is a sin?"
We should abandon theories which say either that Cameron has no chance at the next election or that Miliband has no chance. But, as a nonplussed Dimbleby did not say to the bishop, where would be the fun in that?
Indeed, last week this column departed from its usual Habgoodian reticence and declared, on the evidence of the second-quarter growth figures, that the election was over. Actually, in the small print, I hedged this prediction by saying that if wages rise faster than prices next year, I thought the Tories were in with a much better chance than the current opinion polls suggest.
One of my colleagues helpfully pointed out that this appeared to be at variance with something I had written the month before, which looked at data compiled by the excellent Electoral Calculus website. This collected opinion polls in the two years before each of the last seven elections, and observed an erratic pattern of opinion tending to move back towards the governing party. However, it also observed that opinion would have to move further towards the Tories in the current cycle than in the average of previous elections, for Cameron to win a majority.
So it was not an absolute rule. In any case, to stay on as PM, Cameron needs the Tories only to remain the largest party, which requires the Tory share of the vote to be four points ahead of Labour. That is not unreachable.
Last month, Stephen Fisher of Trinity College, Oxford, carried out a new analysis of the data examined by Electoral Calculus. In addition to the tendency of polls to move in favour of the government before elections, he took account of two other factors: the tendency of polls to overstate Labour support, and the tendency of party support to "regress to the mean" – that is, to move back towards its long-run average level of support.
His model suggested an 88 per cent chance that the Conservatives would win most seats at the next election, and only a 12 per cent chance that Labour would be the largest party. I suspect that this is tilted too much in the Tories' favour, because the model does not take into account the effect of the UK Independence Party. That cannot be predicted from past elections, because Ukip and its predecessor the Referendum Party has never won more than 3 per cent of the vote, but it must be likely that it will take more votes from the Tories than Labour.
Fisher's model is very different from the probabilities suggested by the bookmakers' odds, which are that Labour has a 58 per cent chance of winning the most seats and the Tories 41 per cent.
I don't want to risk Lord Habgood's disapproval: suffice it to say that I think that the chance of David Cameron winning a second term is better than either the betting odds or the opinion polls suggest.