In all such exercises, you’re looking for the “signature” associated with one or another story; and the signature here is clearly the one you’d expect with a general fall of demand. Keynes roolz.I'm getting a picture of Krugman jumping around his living room in his underwear, like Tom Cruise in Risky Business. Seems to me that sloppy thinking roolz.
1. Krugman's numbers are consistent with sectoral reallocation. Krugman's data consists of observations on aggregate hiring and separations at the establishment level. Separations will in general include workers who leave an establishment to work in another establishment, either in the same sector or in another sector of the economy. In sectors where the employment share is increasing, hiring will of course tend to be high, as this is necessary to expand employment, but separations will tend to be high as well, since this is in part how workers move up the ladder. Workers with human capital specific to the sector move to higher paying jobs that are not available in their own establishment, possibly by forming their own firms. In declining sectors, hiring is low, but separations are low as well. Workers find it difficult to move up the ladder in their own sector, but moving across sectors is insufficiently attractive, because their human capital is specific to the sector in which they are currently working. What do we observe in the aggregate? The key idea is that sectoral reallocation frictions imply that, when there is unusually large dispersion in cross-sectoral factors - technological change or demand factors - aggregate hiring and separations will both be low. First, establishments in sectors growing relatively quickly will not be hiring as much as they would if workers did not have to acquire new skills or move across geographical regions to work in a different sector. Second, in sectors growing relatively slowly, separations are lower than they would otherwise be because moving across sectors is costly, due to skill acquisition and relocation.
2. "Deficient demand," is a cop out. In the mind of an Old Keynesian, if real GDP goes down, this is always due to deficient demand. By virtue of national income accounting, we know that some component of GDP decreased, and the Old Keynesian can then say that we have experienced an "autonomous" decrease in aggregate demand. Of course, this doesn't explain anything. We would like to know what caused some economic agents to be spending less. In New Keynesian models, at least the ones that do not degenerate into Old Keynesian language, one actually has to be explicit about what aspects of preferences, endowments, technology, or policy, are acting to cause aggregate fluctuations. In a fully-articulated New Keynesian model, just as in any fully-articulated general equilibrium model, "aggregate demand" does not have any meaning. It is possible to tie Keynesian ideas to coordination failures and multiple equilibria, as was popular in the past (see this), but that is not what Krugman and company are up to.