Broken Okun

Robert Waldmann

is honestly puzzled by this post by Brad DeLong. Brad notes that firms don’t seem to be hoarding labor and goes on to predict a jobless recovery.

We Are Live at The Week with “The Jobless Recovery Has Begun”..

“Okun’s Law.” Here is the gist: if GDP (production and incomes, that is) rises or falls two percent due to the business cycle, the unemployment rate will rise or fall by one percent. The magnitude of swings in unemployment will always be half or nearly half the magnitude of swings in GDP.


Four reasons: (a) businesses will tend to “hoard labor” in recessions, keeping useful workers around and on the payroll even when there is temporarily nothing for them to do;


Okun’s Law is broken—especially with regard to the retention of workers in a downturn.


Manufacturing firms used to think that their most important asset was skilled workers. Hence they hung onto them, “hoarding labor” in recessions. And they especially did not want to let go of their prime productive asset when the recovery began. Skilled workers were the franchise. Now, by contrast, it looks as though firms think that their workers are much more disposable—that it’s their brands or their machines or their procedures and organizations that are key assets.

now out of order

“get ready for another jobless recovery.”

Huh ? Okun’s law was symmetric. Why isn’t the new broken-Okun law symmetric ? If the Okun’s law coefficient has shifted from 0.5 to over 1, wouldn’t that suggest that there will be a huge increase in employment when GNP recovers ?

Semi theory after the jump.

To get a bit almost theoretical, if firms don’t have hoards of hoarded workers, they will have to hire new workers in order to incrase production.

I think that Brad’s model, in which the change is caused by a reduction in the value of trained workers perceived by managers, should imply a prediction of a recovery with unusually an rapid increase in employment.

My guess is that Brad is not allowing his theoretical speculation blind him to the fact that the last two recoveries were jobless. Based on atheoretical empiricisms (which is as Paul Krugman notes always accidental theorizing) I too predict a jobless recovery, but I think that Brad’s informal model implies the opposite prediction.

I can reformulate Brad’s model. The key variable is managers perception of the persistence of shifts in demand for their products. If they think that demand for their products will recover soon, they hoard labor. If they think that it will stay low, they lay off workers until they can just meet current demand.

this means that a recession accompanied by a sectoral shift will cause a larger decline in employment. During the recovery, firms in the expanding sector will increase employment slowly as they can only train so many new workers (workers recalled from temporary layoff don’t have to be re-trained). The story is that asectoral shift causes high unemployment, a drop in demand causes high unemployment and the combination causes a sharp decline in employment followed by a jobless recovery.

Oddly, however, Brad and I have co-authored a paper with a third explanation.

Our claim is that in the USA the Okun’s law coefficient is an increasing function of the unemployment rate. The story is simple, if there is high unemployment, firms can lay workers off and then rehire them later, because the workers won’t have found new jobs and will just have to wait to get their old jobs off. This suggests that the modified Okun’s law gives a nonlinear and in particular concave relationship between unemployment and GNP minus trend (or GNP minus peak or GNP minus last years GNP). This means that the drop in GNP wich corresponds to an increase of unemployment form say 5 to 9 % is less than twice the drop in GNP which corresponds to an increase from 5% to 7%. A model in a published paper predicts well out of sample. So why is Brad psychoanalyzing businessmen ?

Of course the accidental theorist might be attracted to the accidental theory which says recoveries are jobless when the President is named George Bush. Fits the data and implies a normal recovery.

Comments (0) | |