There's one particular line in Osborne's speech yesterday that irritated me. It's this:

On the Treasury's current forecasts, £12 billion of further welfare cuts are needed in the first two years of next Parliament.

Now, there is one fact about the Treasury's forecasts that is more important than any other.

It's that they are wrong. The OBR has estimated that the standard error in PSNB forecasts three years out is 3.2 percentage points of GDP (table 2.2 of this pdf). One interpretation of this is that, given a forecast for PSNB of 2.7% of GDP in 2016-17, there's a roughly one-in-five chance of the actually government running a surplus then. The point about fiscal fan charts is that the fans are wide.

One reasonable principle of decision-making is that policies should not be based heavily upon a forecast with such huge error margins. The investor who bases his asset allocation upon an economic forecast, or the CEO who makes a big investment on the basis of one, are both being prats.

Yes, monetary policy is (sometimes) set on the basis of a forecast. But QE and interest rate changes are easily reversible if the forecasts turn out wrong. The same cannot be said of the benefit cuts Osborne has in mind.

Let's put this another way. In what circumstances is it likely that government borrowing will be high in 2017? To see them, remember what Mr Osborne doesn't seem to know - that the nation's finances are not the same as the government's. A government deficit, by definition, means that other sectors are running financial surpluses. This means some combination of three things: households are net savers, say because they want to repair their balance sheets; the corporate sector is investing less than its retained profits; and/or the rest of the world wants to be net savers - which implies weak overseas demand.

These circumstances will be ones in which the economy is weak and jobs scarce - which means that people will, through no fault of their own, find it difficult to move off benefits. And this is exactly the time in which benefits should not be cut. If anything, the time to cut benefits is precisely when cuts are not "needed" - when the public finances are healthy because the economy is booming.

Now, in saying this, I'm not arguing against benefit cuts per se. If you must argue for them, do so on grounds of claims about existing facts - such their possible disincentive effects - and not on the basis of the fictions which people call forecasts.