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.
Technorati Tags: macroeconomics,
employment,
economics,
recession
Why is it that an economic recession is judged by negative GDP. The US has population growth of about 0.8% a year so surely the US economy should be growing at 0.8% a year just to stand still in real terms. I mean, we should expect to see GDP per capita fall on the basis of the 0.6% figure. Or am I confused (highly plausible scenario)?
This has puzzled me for a while.
Yes indeed, Finsense, slow growth means that real income per person is declining. But the question is whether we have entered a distinct phase of the business cycle in which the dynamic behavior of these series undergoes an abrupt shift. Those dynamics are characterized not by slow growth but by sharp outright declines.
There’s an interesting report here from Wachovia, which contrasts the current slowdown with the 2001.
http://mediaserver.fxstreet.com/Reports/f94cca42-c3fa-47e4-88dd-b4c17a0cdced/3018fdda-e404-41a8-974e-af719ba5ceed.pdf
Essentially the argument is that the strong world economy will support exports and by so doing limit the scope of recession. The other thing to throw in is the increasing re-patriation of profits from abroad – foreign profits have doubled as a proportion of GDP over the last couple of decades and this when combined with the re-financing of financial institutions via soverign Wealth Funds, very significantly offsets the scope of any downturn.
This one isn’t like 2001. It’s more like 1991: high energy prices, collapsing housing markets, and a big fat credit blowup.
Me, I waffle. Things look pretty ugly, but not so ugly that I am quite convinced. Yet.
In addition, I believe that it is essential to understand the extent to which the “senior policy makers” in this administration are totally committed to the control of expectations (of all sorts) so as to engender levels of “confidence” in the public which are inconsistent with reality.
This results (to varying degrees) in the intentional selection of compillation methodologies for the data series that are “self serving” and result in reportage always presenting a picture somewhat or substantially more rosy than reality,
a “if the people knew how bad things really are, they would just go out and make them worse” sort of mind set.
You can obtain a glimpse of this process in the speeches and press releases of Paulson, Bernanke, other FOMC members, the CEO of FNMA, hell, the entire gang.
I suppose the best way to refer to this is as a “benign corruption” of the facts (and data).
Very soon, if not right now, a new (animated) emoticon needs to make an appearance … the one with the tears running down the cheeks.
We have had weakness in certain states for some time: California and Florida sales tax receipts have been running below year-ago levels since July.
Well, the weakness is now at the national level: Federal receipts in January were below year-ago levels. If it happens again in February and March, that will be another strong indicator that all is not well.
Eyeballing the historical Federal collection data since ’80, it appears that Federal collections are a coincident/lagging indicator of recession.
http://www.fms.treas.gov/mts/mts0108.pdf
I not only believe that recession began in December 2007, I actually think that it is a classical ‘V’ shape, and a steep one at that.
Thus, we might already be near the half-way point, and could see recovery as soon as Sept or Oct.
Job losses, of course, are a lagging indicator, and most occur in the 2nd half of a recession and even after the recession ended.
If it is a reasonable assumption that a recession has already begun, should we begin to see the recession probability index leap above 50 relatively soon and continue to climb? How important really is the non-farm payroll data?
RWW, that index is based on GDP alone. Won’t change until the new GDP numbers come out in April.
Finnsense: One point about not including population growth into account might be that those actually making up the growth are mostly just being born, and not really contributing to the economy at all yet (unless you consider the role of consumer contributing). And it would probably get somewhat hairy trying to establish a better measure, trying to take into account the rate of people turning 16 or 18, those who have going to college or not, the rate at which people are retiring, the changing ages at which they’re retiring, etc…. I’d suppose some “they” at some point decided it might just be best to consider it a plain, unadorned 0% change in GDP.
A recent paper by Tim Kane of the Joint Economic Staff on Employment Numbers as Recession Indicators argues that the labor market data that is most useful as recession indicators is the unemployment rate and weekly claims for unemployment insurance. Here’s the abstract:
This paper investigates the value of employment data as real-time recession indicators. Among popular monthly labor measures, the unemployment rate is the most useful as an indicator of recession, whereas two top measures of employment growth – payroll jobs and civilian employment – have little value. Two other series, the labor force participation rate and the employment-population ratio, also provide little or no value in anticipating a recession. The best pre-recession employment indicator is actually weekly claims for unemployment insurance (UI).
If one believes Kane’s analysis, the labor market data released last week argues for a much lower probability of recession than is commonly perceived. That’s because the unemployment rate fell for the second month in a row, while growth in the 4-week moving average of weekly claims has been very modest. If I read his tables correctly, his interpretation of the data would point to something like p(recession) = .1 or less.
Since Kane is apparently a good Bayesian who based his methodology largely on work by Chauvet and Hamilton (Dating Business Cycle Turning Points in Real Time) it would be very interesting get an appraisal from JDH as to whether analysts might be putting too much emphasis on the payroll data.
Interestingly, the recession futures contracts traded at Intrade.com agreed with JDH and jumped from about p=.59 to p=.69 on the payroll data. But I wonder if that is simply due to the huge significance the media puts on the payroll data, notwithstanding the kind of evidence Kane has mustered in support of other labor market data being more useful.
The wholesale sales figures released today don’t suggest a recession began as of January.
Guys,
You can’t pick each of the statistics and see if they indicate a recession. The data is full of problems. If wholesale sales figures are OK (and not indicating recession) then it doesnt mean we are not in recession.
Mr. Buffet has said commonsense wise, we are in a recession. So plan and act accordingly.
Economists and NBER let them do the useless work of confirming the recession after 9 months.
the other point to consider is that the stock market is not reacting to easier monetary policy or lower rates. The stock market PE is still falling despite significantly lower rates. That is not a good sign.
Bill S, I’d be interested in JDH’s take on that as well.
As a quick and dirty question about how the Unemployment Rate and NFP (M/M changes) move, I plotted the normalized series since 1985 here. You can see in the picture some of why the Unemployment Rate is a better recession predictor than NFP.
sjp: I can’t replicate your picture of NFP. What is the “normalized” NFP? Did you HP-filter the series?
Menzie: My mistake and thanks for catching it — for the underlying NFP data, I was taking the annual difference, not the month-to-month (the number for Jan 2000 is NFP_{Jan,2000} – NFP_{Jan, 1999}, not NFP_{Jan,2000} – NFP_{Dec, 1999}). So these underlying numbers are quantities like -2000 or 3000.
For series X, I calculate (from Jan 1985 to Jan 2008) mean(X) and std(X). Then the normalized series is (X-mean(X))/std(X). I normalize so that the units don’t obscure the time series movement.
And also this is the real-time NFP data, as first reported.
Geesh! Sorry for not knowing what I was drawing!