The latest employment numbers suggest that the tide has turned.
The BLS reported on Friday that the number of Americans employed on nonfarm payrolls fell by a seasonally adjusted 63,000 workers between January and February. As Menzie observed in a very nice picture yesterday, the BLS on Friday also revised down its earlier estimates for December and January, making the confirmation of a downward move in total employment much clearer now than it had been for either of the previous two months.
There’s still a lot of noise in these monthly estimates and potential for further revisions, up or down. But a longer average such as the 12-month growth rate should be more robust with respect to such statistical noise. I was struck by Calculated Risk’s plot showing that the recent behavior of year-on-year growth rates looks like something we’d only expect to see in an economic recession.
I was curious to check whether CR’s graph is indeed an indicator that can be trusted given the revisions in the data– what looks like a clear signal in the revised data may have been more ambiguous in the data as it was originally released. Fortunately there are some very nice databases available today that make answering this kind of question much easier than it used to be, such as ALFRED (ArchivaL Federal Reserve Economic Data), which is maintained by the Federal Reserve Bank of St. Louis. ALFRED allows you to look up the values of thousands of different variables as they were actually reported in real time as of any specified historical month. I used that database to calculate the year-on-year percentage change in the monthly seasonally adjusted nonfarm payrolls number as it would have been reported in each historical month. That exercise confirms that CR is on to something here– there is no example of either the real-time or the revised year-on-year growth series ever getting as low as it currently is unless the economy is entering or recovering from a recession.
Another feature worth remarking on is that for once the various indicators of the number of people working all seem to be in agreement about the latest direction. Automatic Data Processing’s estimate based on the 24 million workers whose payrolls it processes directly was that national private-sector employment fell a seasonally adjusted 23,000 jobs in February, while the BLS’s separate survey of households reckoned we lost 255,000 jobs in February.
It will still be many months before we would expect to see an “official” declaration that a recession has indeed begun from the Business Cycle Dating Committee of the National Bureau of Economic Research. Granted, the latest data look recessionary. But the Committee would be pondering the following: suppose these data are revised up or next month’s numbers start to improve. Would what has happened so far be enough to characterize as a recession? The answer is pretty clearly no, and that is why no declaration from NBER will be forthcoming any time soon. For the 4 most recent recessions, the NBER did not announce that a recession had begun until 6-9 months after it had actually started. So, if the recent downturn indeed began in December, the earliest we’d expect to hear a confirmation of that from NBER would be May.
Here at Econbrowser, we offer a mechanical GDP-based summary (described here and in more detail here) that like NBER is quite conservative before making a call. We will wait until our index rises above 67% before declaring that a recession has begun.
In the mean time, though, if you want to claim that the recession has begun, that now strikes me as quite a reasonable working hypothesis.