BIG PICTURE: Labor Market Data - All Messed Up
13 Dec 07:30
(This article was originally published Thursday)
By John McAuley
Of DOW JONES NEWSWIRES
NEW YORK (Dow Jones)--Just as anxiety about the labor market's struggle to
get out of its soft spot intensifies, the reliability of the latest data is
being called into question.
First came the Labor Department's announcement Monday that there were errors
in the November jobs report issued last Friday. Although the errors were
confined to September and didn't affect the November data, the revision so
quickly after release cast doubt on the reliability of the report.
Indeed, the Department had already announced that payroll employment in March
2002 - the benchmark month to be incorporated in the May 2003 report - was
284,000 lower than the level of 130,701,000 currently in the record books.
Then on Thursday, the Labor Department reported a sharp 83,000 jump in
initial jobless claims which reflected the unwinding of an understatement in
November. Wall Street economists say that the volatility seen in the past few
weeks' claims data - the result of problems created by a change in the
seasonally adjustment methodology that statisticians use to adjust for
distortions in the actual data due to holidays and other factors - has made the
series extremely hard to read.
As legendary baseball manager Casey Stengel once asked of the early Met's:
"Can't anyone play this game?" It's starting to look as if the Labor Department
simply can't get its data right.
That's especially unfortunate now, given the crucial role that employment
data hold in determining the outlook for the economy and the sputtering
recovery.
The snafu with the latest employment data was the most egregious error, but
probably the least significant.
The Labor Department, following normal practice, had announced and released
on Monday Dec. 2, updated seasonal adjustment factors to be used for the
payroll survey data for November 2002 through April 2003. In addition, revised
factors were issued for September and October. However, when the November jobs
report was issued just four days later on Friday, the statisticians used the
updated seasonal factors for the October and November data, but used the old
factors for September.
As a result, total employment was overstated by 80,000 in September and, so,
the Department had to resort to the unprecedented alternative of issuing a
corrected release on Monday, Dec. 9.
Ray Stone, of Stone and McCarthy Research Associates, was the one who spotted
the error. "When I applied the revised seasonal factors to the not-seasonally
adjusted data for September, I didn't get the results I expected to find, so,I
spoke to people at the Labor Department and we figured out the problem," he
said. Stone had expected a 90,000 decline rather than the 4,000 decline
contained in the original November report for September.
There's nothing unusual about the Labor Department already knowing the
magnitude of its annual benchmark revision, which it is due to report next June
6 along with the May employment report. It takes the Labor Department that much
time to apply the benchmark to the detailed breakdown of the data and then to
compute revised data going forward and backward from that date.
It is just unfortunate that the size of the benchmark revision is being
reported at the same time other problems with labor market data are being
exposed. Indeed, next year's revision, which the Department notes is roughly
0.2% of the total, is less than the average revision for the past decade of
0.3%. However, as Stone points out, this is the first downward benchmark
revision since 1991.
Seasonal Adjustments Gone Awry
Perhaps the greatest concern, however, is with the seasonal adjustment
problems with the initial jobless claims data. These data are released on a
very timely basis - covering the week ended five days earlier - and, because
they measure the new filings for unemployment insurance benefits, are at the
cutting edge of labor market information.
The 83,000 jump in claims to 441,000 looks worrying. Moreover, Stone expects
little decline - only to about 435,000 - in the next weekly report, which would
could send a signal that the market was seriously deteriorating.
The problem has to do with a new seasonal adjustment methodology that the
Labor Department is using this year. Instead of mapping the data against those
from 1985, 1991, and 1996 - years with the same holiday calendar as this year -
the new method also gives weight to the most recent year. The problem is that
the blip up in claims after Sept. 11 meant that the model predicted a larger
increase than actually occurred and as such understated the adjusted numbers in
October and November.
For their part, Labor Department statisticians are putting a positive spin on
things.
"It's true that there were some distortions, but on the whole, the four-week
moving average tended to converge," said Labor Department statistician Tom
Stengel.
There is, moreover, an additional problem with the data in December.
California changed the law on eligibility and benefits in December of last year
and, so, many job losers delayed filing until after Jan. 1, 2002 in order to
receive the higher benefits. That is likely to overstate claims this December.
"I wonder whether we might not be better off with data that were unadjusted
for seasonality," said Ian Morris, chief economist at HSBC Securities. "The
problems with the data might be a sign that cutbacks at the Bureau of Labor
Statistics have cut into the job they do. (Federal Reserve Chairman) Greenspan
has expressed concern about the role of economic data."
Indeed, the Federal Open Market Committee's statement after its meeting on
Tuesday referred to "the limited number of incoming economic indicators since
the November meeting." While the FOMC did see a full monthly round of data
during theperiod, the statement may have referred to doubts about the data.
However, many economists don't like the idea of using unadjusted data, which
would arguably make it harder to identify trends, despite all the problems with
the adjustment process. "We want to be able to interpret things at the margin,"
said Stone and McCarthy's Stone
-By John McAuley, Dow Jones Newswire, 201-938-4425; john.mcauley@dowjones.com
(END) Dow Jones Newswires
12-13-02 0730ET |
