Paul Krugman is being accused of hypocrisy for calling for an extension of unemployment benefits when one of his textbooks says "Generous unemployment benefits can increase both structural and frictional unemployment." I think he can be rescued from this charge, if we recognize that economics is not like (some conceptions of) the natural sciences, in that its theories are not universally applicable but rather of only local and temporal validity.

What I mean is that "textbook Krugman" is right in normal times when aggregate demand is highish. In such circumstances, giving people an incentive to find work through lower unemployment benefits can reduce frictional unemployment (the coexistence of vacancies and joblessness) and so increase output and reduce inflation.

But these might well not be normal times.It could well be be that demand for labour is unusually weak; low wage inflation and employment-population ratios suggest as much. In this world, the priority is not so much to reduce frictional unemployment as to reduce "Keynesian unemployment". And increased unemployment benefits - insofar as they are a fiscal expansion - might do this. When "columnist Krugman" says that "enhanced [unemployment insurance] actually creates jobs when the economy is depressed", the emphasis must be upon the last five words.

Indeed, incentivizing people to find work when it is not (so much) available might be worse than pointless. Cutting unemployment benefits might incentivize people to turn to crime rather than legitimate work.

So, it could be that "columnist Krugman" and "textbook Krugman" are both right, but they are describing different states of the world - and different facts require different models; I fear that Krugman's excessively combatative style is ill-judged in this context.

The question to ask of any model or theory isn't merely: is it internally coherent? It's: does it apply here and now? Here are three other examples of this:

- You can write some sort of model in which QE is deflationary. The question is: does that model apply to the economy today?

- Brad De Long has suggested that the "liquidationist" (pdf) theory of recessions - the idea that unprofiatable capital must be destroyed - was a terrible idea during the Great Depression of the 30s, but valid for some previous recessions. Different times require different theories.

- The idea of expansionary fiscal contraction might have applied in some times and places, but it hasn't applied to the UK recently.

My point here generalizes, I think. In the social sciences, very many hypotheses, models or theories are true of some times, some places and some people. One job of the social scientist is to judge which theory works when. We don't need theories of everything. We need theories of here and now.

Another thing: economic efficiency isn't the only relevant value here. There's also the principle of luck egalitarianism, which says we shouldn't penalize people for things thery can't control. This argues more more generous unemployment pay in hard times, as it's more likely in such circumstances that people are out of work through no fault of their own.