Some Evidence on the Diagnosis

There is also some evidence that raises questions about the diagnosis. To understand this evidence we need to recall the definition of permanence in economic statistics.

Permanent

In a series of observations of an economic measure over time, the measure is said to be relatively highly permanent if the measure for one period is relatively close to the measure for the previous period, with a high probability.

Transitory

In a series of observations of an economic measure over time, the measure is said to be relatively highly transitory if the measure for one period is independent of the measure for the previous period, but close to a stable average value with a high probability.

Observations of unemployment show that unemployment is highly permanent, rather than transitory. This is a problem for the New Classical interpretation of unemployment. Here is the reason.

Recall, according to the New Classical interpretation, potential employees come to the labor market without knowing exactly what their labor is worth in wages. If they overestimate, then they will be unemployed for a long time as 1) they don't get offers up to their standards, and 2) learning from this, they gradually revise their standards downward. However, because they are rational -- so their estimates of what their labor is worth are "rational expectations" -- the mistakes they make are unbiased. They are as likely to underestimate as overestimate. What happens when most people underestimate the value of their labor? The answer is that most of them get offers right away for their reservation wage, so they take the jobs and are off the market, so unemployment drops below its long-term average. Since overestimates of the wage are as likely as underestimates, unemployment numbers below average are as likely as above.

This leads to the expectation that unemployment would be relatively transitory. It would fluctuate around an average level, and would be about as likely to be below its average in any year as above. But that is not what we observe. Instead, we see that when unemployment is above its average in one year, it is much more likely to be above its average in the following year than not. If it is below its average in one year, it is much more likely to be below its average in the following year than not. This is what it means to say that unemployment is highly permanent, and that's why the evidence that unemployment is permanent disagrees with the New Classical expectations.

Many New Classical economists have acknowledged that the permanence of unemployment is a problem for their theory, but problems can often be solved, and a number of proposals have been made to modify the New Classical theory in small ways, without giving up the basic idea, and reconcile it with the fact that unemployment is highly permanent. In the terms of the Reasonable Dialog, the evidence that unemployment is permanent undercuts the New Classical interpretation. So does the evidence mentioned on the previous page that suggests that changes in unemployment have little relation to the opportunity cost of waiting for a better job offer. But, as so often is the case, they don't undercut the New Classical theory fatally. As I write, there are New Classical economists "searching" for a modification of their theory that will undercut the undercutting evidence and restore the New Classical theory to its full credibility. Perhaps soon they will succeed. Until then, we can only say that the New Classical theory has some problems with the evidence. And yet ... all the same ...

All the Same ...