A lot of people are excited today not because the unemployment rate is low (it’s very high — 8.9 percent), nor because the economy is adding jobs (it lost another 539,000 last month, according to statistics just released by the BLS), but merely because it’s losing jobs less quickly. That is, the second derivative of the employment rate — the change in the rate of change — has improved. This is what the situation looks like:
The economy started losing jobs in January, 2008 and has continued to lose them ever since. The peak month for job losses — so far — was January 2009, in which 741,000 jobs were lost. The month at which the second derivative bottomed out — the time when the rate of job losses was increasing the fastest — came in November.
The $787 billion question, of course, is whether a decrease in the rate of job losses indeed portends a recovery, or whether such data is subject to false starts. Let’s take a somewhat high-level view of the progress of the employment situation over the previous five recessions.
First, the recession of 1973-1975. The red bars indicate, by the way, when we were “officially” in a recession, according to the NBER:
This was, at first, a “jobless recession”, the primary concern instead being the extremely high inflation rates triggered by (among other things) the oil crisis. The economy, however, began shedding jobs with a vengence in November 1974, with job losses peaking at 602,000 in December 1974 — a rate that would be equivalent to about 1 million job losses given today’s population. By January, 1975, however, it already looked like the worst was over, with job losses decreasing to 360,000, and indeed the economy had officially pulled out the recession by April and was adding jobs again shortly thereafter. Here, then, improvement in the second derivative does appear to have had some predictive powers. Next up, the recession of 1980:
This was a short-lived and relatively orderly recession; an improvement in the second derivative in June 1980 was followed by the end of the recession two months later. It was, however, followed not long afterward by a much worse recession…
Ah, the good ol’ recession of 1981-82, the one to which the present one is most frequently compared. And note the presence of not one, not two, but three false starts. In February 1982, job losses slowed all the way down to just 6,000 jobs, but things got worse — not better — in March and April. Then in May, job losses slowed to 45,000, before accelerating again throughout the summer. Finally in August, job losses slowed from 343,000 to 158,000 before increasing again in September and October. This is the sort of example that should make us very cautious.
Next, the early 1990s recession:
There is arguably a false start here in September 1990, when job losses slowed from 208,000 to 82,000, although the recession had barely even begun then. The job-loss situation also appeared to be getting a bit better in December, 1990, but the peak month for job losses in fact wouldn’t come until two months later in February.
Finally, the recession of 2001:
We have to include a lot of data here because employment didn’t bottom out until June 2003 — some 19 months after the recession had officially ended! Although the main period of job losses, from the summer of 2001 through the spring in 2002, proceeded in a relatively orderly fashion, there were still places where we could have been deceived. For instance, the economy began adding jobs in June 2002, but then lost them again for the next three consecutive months; a similar story took place in January 2003.
In sum, employment rate data is fairly stochastic, and if there is ample reason for optimism (and I believe there is), there is also ample reason for skepticism.