In view of last week's jobs data, and with Bonddad's permission, I am reposting in its entirety an entry of mine from September 21 of this year, which was entitled, Getting it wrong about Initial Jobless Claims and Nonfarm Payrolls (with a few updates at the bottom) for reasons that I hope will seem obvious as you read along:
----------
Some time ago, Prof. Brad DeLong of Berkeley, thinking aloud with graph, drew a line across the 1991 and 2001 recessions and recoveries, making a "note to self" that it appeared that Initial Jobless Claims post those recessions had to decline to 400,000 or less before payroll jobs were added. Thus, mused Prof. DeLong, it must be so as well, post this "Great Recession." This "note to self" was subsequently repeated by Bill McBride at Calculated Risk, from which it has now been picked up and repeated at Prof. James Hamilton's site, Econbrowser. It is well on its way to becoming Holy Writ.
Let me say first of all that I have the highest respect for all 3 of the above gentlemen. Nevertheless...
IT IS WRONG.
The 1991 and 2001 recessions were very mild. Peak initial jobless claims in those recessions were 501,250 and 489,250, respectively. It would be nuts to think that jobs would be added to the economy anywhere near the 500,000 high water mark in jobless claims from those recessions.
The 1973-4 and 1981-2 recessions are much better comparisons. They were the two most severe post-WW2 recessions up until now, respectively featuring 9% and 10%+ unemployment. Furthermore, peak initial jobless claims in those recessions were 560,750 on February 1, 1975 and 674,250 on October 9, 1982, respectively; both peaks being much closer to our recession's peak initial claims number of 658,750 on April 4, 2009.
In the case of the recoveries from both of those recessions, payrolls started to grow as the ievel of initial jobless claims crossed 500,000, not 400,000.
Here it is in graph form, including initial jobless claims (in blue and green respectively) ending as of the week each crossed below 500,000, together with nonfarm payrolls (in red and orange) ending the same month:
It is easy to see that nonfarm payrolls troughed in the preceding month, and grew during that month that initial claims crossed below 500,000.
Here is the raw data from the St. Louis Fred site:
1975 Recovery
Initial jobless claims:
1975-02-01 560750 (peak)
1975-04-05 543250
1975-04-12 540500
1975-04-19 535750
1975-04-26 525250
1975-05-03 517250
1975-05-10 514000
1975-05-17 509750
1975-05-24 504750
1975-05-31 497250
and Nonfarm Payrolls:
1975-01-01 77297
1975-02-01 76919
1975-03-01 76649
1975-04-01 76463
1975-05-01 76623
1975-06-01 76519
1975-07-01 76768
1975-08-01 77154
1982-3 Recovery
Initial Jobless claims
1982-10-09 674250 (peak)
1982-12-04 586250
1982-12-11 569750
1982-12-18 554500
1982-12-25 523750
1983-01-01 518000
1983-01-08 512250
1983-01-15 503000
1983-01-22 500500
1983-01-29 492750
and Nonfarm Payrolls:
1982-10-01 88894
1982-11-01 88770
1982-12-01 88756
1983-01-01 88981
1983-02-01 88903
1983-03-01 89076
For good measure, in the 1980 Recession, in which unemployment peaked at 7.8%, jobs showed growth in August 1980 at the same time as Initial Claims averaged 546,000:
Initial Claims:
1980-06-07 629000 (peak)
1980-07-05 599250
1980-07-12 584500
1980-07-19 576500
1980-07-26 559250
1980-08-02 556750
1980-08-09 556750
1980-08-16 546250
1980-08-23 534750
1980-08-30 518000
Nonfarm Payrolls:
1980-05-01 90415
1980-06-01 90095
1980-07-01 89832
1980-08-01 90092
[NOTE: When I originally posted this, a commenter noted that the way I posted the above was somewhat confusing. As I responded at the time, all of the data above is positive, e.g., +536,000 new jobless claims. The bolded numbers are the low point for payrolls during each recession, e.g., 76,463(,000) in 1975; and the two weeks marking where initial jobless claims crossed the 500,000 mark.]
IT IS SIMPLY NOT TRUE THAT INITIAL JOBLESS CLAIMS MUST DECLINE TO 400,000 BEFORE NONFARM PAYROLLS TURN POSITIVE.
Note that for all three Recessions/recoveries, I've included jobless claims from the previous month, when nonfarm payrolls troughed. In 1975, payrolls actually troughed at 536,000 new jobless claims. In 1980, they troughed at 580,000. In 1982, they troughed at 559,000 (the 4 week average of the months involved).
That is partly why I have taken the position that in this recession/recovery, payrolls will trough once there is a sustained reading of ~530,000 if there is a slow decline in initial claims, and 500,000 if the decline quickens.
I am utterly confident that there will be job growth long before jobless claims fall to 400,000.
----------
UPDATE: Now, here is the data since I originally posted the above:
Initial claims
2009-04-04 658750 (peak)
2009-09-26 548750
2009-10-03 540500
2009-10-10 533000
2009-10-17 532250
2009-10-24 526750
2009-10-31 524250
2009-11-07 520500
2009-11-14 513000
2009-11-21 495500
2009-11-28 481250
Nonfarm Payrolls
2009-08-01 131257
2009-09-01 131118
2009-10-01 131007
2009-11-01 130996???
2009-12-01 ???
Four days after that post, in a note entitled When Will the Economy Start to Add Jobs (VI), I concluded that it was most likely that jobs would turn positive in November or December, +/- 1 month, and gave a possible "glide path" to that number. That "glide path" is the first column below (in thousands). The current BLS revision is the second column:
2009-09-01 (-145) (-139)
2009-10-01 (- 91) (-111)
2009-11-01 (- 28) (- 11)
2009-12-01 (+ 34) ???
------------------------------
I am sure the data will make me look like a smacked a@@ soon enough, but just at this particular moment, I am smelling like a rose.
