Thursday, May 24, 2007

Problems With the Employment Numbers?

From the WSJ:

• The News: Recent signs indicate the job market may be weaker than monthly data show, particularly in construction.

• The Background: The economy has slowed, but job growth has remained robust.

• The Upshot: The cause of the statistical disparity remains something of a puzzle.

Let me caution here: the primary reason for the disparity is based on statistical sampling which is not my strong suit. I'm good at explaining final numbers, not generating the numbers used in analysis. Now that we have that caveat out of the way:

Those signs are particularly stark in the home-building industry, which has been hurt by the slump in the housing market. Housing starts in April fell 33% from their recent peak in January 2006. Yet, the number of residential-construction jobs has dropped by only about 3% over the same period.

Economists cite several possible explanations for the disparity. One is that layoffs have lagged behind the housing slump and will weaken further.

In addition, some economists say the monthly figures from the Labor Department's Bureau of Labor Statistics may be overestimating employment, perhaps by misclassifying construction workers or by failing to count large numbers of laid-off illegal immigrants.

The bureau releases two monthly employment figures: the unemployment rate, which is based on a household survey, and a tally of nonfarm payrolls, based on a survey of employers. Both are conducted through sampling and depend on voluntary responses.

A lesser-known employment snapshot, based on a quarterly census of state unemployment insurance records, shows the economy created about 19,000 private-sector jobs in the third quarter of 2006, the most recent data available. That contrasts with the 500,000 indicated in the monthly figures for that period. It also shows the number of construction jobs dropped by 77,000, in contrast with the increase of 19,000 jobs shown in the monthly surveys.

The data suggest that "maybe the labor market behaved a little more like we thought it should in times of a slowing economy," said Michael Feroli, an economist at J.P. Morgan Chase & Co.

There has been a lot of discussion about the employment reports this economic cycle. The discussions have centered on a few problems:

1.) The birth/death model. The BLS uses a model (called the birth/death model) that attempts to account for new businesses that aren't in their employment sample and businesses that go out of business whose closing is not reflected in the final employment numbers. I can't speak for the veracity of this analysis, but it has come under fire from several commentators.

2.) The continual revisions of the jobs numbers. The BLS has continually recalculated the monthly figures, making the initial release an almost moot points. In addition, the BLS magically found about 800,000 jobs in an annual revision last year. The point is there appears to be a pretty big problem with the way BLS is conducting business. There is always a certain amount of lee-way with economic numbers and analysis. However, the size of the revisions have led some people (myself included) to question the way BLS does business.

There have been several articles lately which noted the divergence between the overall growth rate (1.3%) and the continued low unemployment rate. These two numbers just don't add up from an analysis perspective. Usually every economic cycle creates analytical problems because (surprise) reality does not add up like economic models say they should. This cycles are not more pronounced than other cycles. But the divergence between theory and reality has emerged again.