Fed Chair Ben Bernanke a few weeks ago said he saw some green shoots of favorable developments in financial markets. Does today’s Labor Department report that the seasonally adjusted number of initial claims for unemployment insurance fell by 20,000 workers in the most recent week constitute another?
This story in the Wall Street Journal a few weeks ago also caught my attention:
“There’s growing evidence supporting the optimists’ view, and I am surprised at that,” said Robert J. Gordon, an economist at Northwestern University and a member of the National Bureau of Economic Research committee that is the official arbiter of when recessions begin and end. “I was sort of in the pessimists’ camp until I started looking at things.”
He points to one indicator in particular with a remarkable track record: the number of Americans filing new claims for unemployment benefits. In past recessions, it has hit its peak about four weeks before the economy hit a trough and began to grow again. As of right now, the four-week average of new claims hit its peak of 650,000 in the week ended March 14. Based on the model, “if there’s no further rise, we’re looking at a trough coming in April or May,” he said, which is far earlier than most forecasts currently anticipate.
I was curious to take a look at the pattern that Gordon identified. The graph below plots 4-week averages of the initial unemployment claims going back to 1967, with vertical lines drawn at the first week of the month in which the NBER eventually declared that a recovery from the recession began. Gordon’s relation is indeed pretty striking– in each of the last six recessions, the recovery began within 8 weeks of the peak in new unemployment claims.
The graphs below display the 12 weeks preceding and following the business cycle trough for each of those six recessions in more detail. The blue line shows the 4-week average and the black line the raw weekly data themselves. The last of the 7 panels displays the most recent 13 weeks of data, in which one can see the flattening of the blue curve from which Gordon found some grounds for optimism two weeks ago. The number released last Thursday pushed the average back up to a new 2009 peak. But today’s datum brings the 4-week average down, ever so slightly, from the week before, and further reinforces Gordon’s tentative impression of a possible flattening. [Bigger version of the picture is available here]
The series plotted above for all previous recessions are the currently known revised numbers. It would be interesting to see how easy the pattern is to recognize in initial data of the kind that we’re actually trying to use in the last panel.
Perhaps there’s nothing of significance here, as it’s all too easy to read too much into the temporary ups and downs that are part of any broader trend. But I agree with Gordon that these numbers bear watching. If subsequent data confirm that the 4-week average of initial claims did indeed reach its peak in the number reported April 2, and if Gordon’s pattern holds up, the recovery that many of us had assumed would be quarters or perhaps even years away may instead have started by June.
Originally published at Econbrowser and reproduced here with the author’s permission.