Friday, December 03, 2004

The Employment Situation: November

This being the first Friday in December, the Bureau of Labor Statistics has released the Employment Report for November. Let's take a walk through the key numbers, most of which are provided in Table A in the summary.



1) The top line number is the change in nonfarm employment measured from the establishment (or payroll) survey in November. That came in at +112,000.



2) In each release, the data for the previous two establishment surveys are subject to revision. These revisions were -20,000 in September and -30,000 in October, so the total increase in nonfarm payroll employment in November compared to what was reported last month for October is 112,000 - 54,000 = +58,000, from 132,017,000 to 132,075,000.



3) Continuing with the establishment data, the workweek for private production or nonsupervisory workers fell from 33.8 to 33.7. This may not seem like much, but consider the following. Production and nonsupervisory workers account for four-fifths of nonfarm payroll employment. So every tenth of a decline in the workweek generates a reduction in labor utilization that is equivalent to:



(0.1 hour/prod worker) x (4/5 prod worker/total workers) x (132 million total workers) = 10.56 million hours



Since production and nonsupervisory workers averaged 33.8 hours per week, this is equivalent to a reduction in labor input equal to 10.56/33.8 = 0.312 million or about 312,000 workers. This more than outweighs the increase in the number of payroll jobs added (acknowledging that it is a bit of an approximation, given the rounding of the numbers to get the "four-fifths" and 0.1 numbers).



4) Finishing up the establishment survey, average hourly earnings for production or nonsupervisory workers increased by $0.01, to $15.83. This figure does not adjust for inflation, which would swing it to a decline. In addition, because of the reduction in the workweek, average weekly earnings for these workers fell by $1.25 to $533.47 (again before adjusting for inflation).



5) Turning to the household survey, the top line number is the unemployment rate. It fell by 0.1 percentage point in November, down to 5.4 percent of the workforce. However, this is rounding error again, since the change was really from 5.46 (rounded up to 5.5 in October) to 5.41 in November. Here's how the BLS (appropriately) wrote this up:

The jobless rate has been either 5.4 or 5.5 percent in each month since July. This is slightly below the rates that prevailed in the first half of 2004.


6) It is also of interest to see how the change in the unemployment rate occurred. It breaks down as: 439,000 more people in the labor force, due to higher employment of 483,000 workers and lower unemployment of 45,000 workers. There were also 209,000 fewer people "not in the labor force." The employment-population ratio and the labor force participation rate increased by 0.2 percentage points each, to 62.5 and 66.1 percent, respectively.



7) Note that the household and the establishment survey will often measure different numbers for the change in the number of people employed (here: 483,000 vs. 112,000). Part of this is due to different samples (e.g., household survey includes the self-employed, establishment survey counts multiple job holders). The rest is that they are both estimates of an underlying population number. In general, the establishment survey is to be preferred for the purpose of counting the change in the number of jobs, because that number is estimated more precisely in the establishment survey (it's a bigger survey and it uses administrative data).



8) We also care about the nature of unemployment. In a set of earlier posts (the last one is here), I discussed alternative measures of unemployment. The broadest one included all people who are not in the labor force (i.e., not actively looking for work) but who do "want a job." This group fell from 5,345,000 to 5,118,000 between October and November. If we reclassified them as unemployed (and thus in the labor force), this augmented unemployment rate would have fallen by about 0.2 percentage points, from 8.8 to 8.6 percent.



There are certainly other elements of the report to explore, but these are the highlights in my view.



To summarize, the top line number is payroll jobs, which came in positive but weaker than expected. Combined with a reduction in the workweek and small downward revisions to previous months, this is not a strong report. It shouldn't put any upward pressure on interest rates, via the pace of the Fed's tightening or the market's reaction. The information from the household survey was more positive, but month-to-month changes in the household survey are less reliable.



You can read media coverage at WSJ.com($), as well as see some video from CNBC's broadcast with alternative views (Mankiw, Zandi, Wessel).



UPDATE: Elsewhere in the blogosphere, the employment release is being discussed by Angry Bear, Econopundit, Macroblog, and Wizbang.

2 comments:

Anonymous said...

Turning to the household survey, the top line number is the unemployment rate. It fell by 0.1 percentage point in November, down to 5.4 percent of the workforce. However, this is rounding error again, since the change was really from 5.46 (rounded up to 5.5 in October) to 5.41 in November. Not to mention that the margin for error for the unemployment rate is +/- 0.2%, so the change is well within the margin and statistically insignificant.

Anonymous said...

" In general, the establishment survey is to be preferred for the purpose of counting the change in the number of jobs, because that number is estimated more precisely in the establishment survey "

hmmm I always thought the establishment survey was better than the establishment survey... Guess I wuz wrong, the establishment survey is better after all.