Much has been written about August’s negative payroll print, and about whether it does or does not mean that a recession is imminent. Treasury Secretary Henry Paulson’s commentary was typical: “It’s not the kind of number I’d like to see. Data does not always move in a straight line, so occasionally you will find some surprises. The economy will continue to grow in the second half of the year.” Also widely cited was the statement by Bespoke Investment Group that “a negative non-farm payrolls report has predicted twelve of the last four recessions.”
As a natural skeptic, I decided to dig into the data a bit further. What I found was somewhat different from the rosy outlooks implied above.
I pulled the data for monthly payroll figures back to the post-WWII period. To avoid duplicate signals, I scanned for months in which the BLS reported a negative payroll number which had been preceded by at least 11 positive payroll numbers. To avoid random sampling errors in the BLS data, I further required those signals to be confirmed by another negative payroll print within one month (ie the month following the first negative, or the next one).
There were exactly eight signals produced: 7/31/56, 9/30/69, 8/31/74, 4/30/80, 8/31/81, 7/31/90, 6/30/00 and 8/31/07. For ALL EIGHT of these signals, the economy experienced a negative GDP print either in the same quarter, or in the next quarter!
These results re-emphasize in my mind the importance of employment for GDP, and also explain why the Federal Reserve seems to focus so much on employment.
Just to highlight that this signal is not simply capturing negative GDP “blips” which do not correspond to extended economic weakness, consider that the average GDP for the TWO YEARS following the signals was as follows:
Further, if you define a recession as a rolling one-year average GDP below 1%, then this signal was not only a sufficient condition, but a necessary condition for every recession back to 1960 (the signal missed the late-1960 recession), as you see on the chart below:
It seems an inescapable conclusion that the odds of a recession are quite high, contrary to the prognostications of those who would be out of a job if they were to say so.
Of course, how the stock market behaves if a recession does indeed occur is not quite as obvious. For three of the seven signals above (8/13/74, 8/31/81 and 7/31/90), the following two years were better than average for the stock market, while the other four preceded worse than average returns.