March 17, 2004
Last month, one measure of employment showed a gain of 21,000. Another declined by 265,000. Most economists emphasize the jobs reported from establishments rather than those derived from a sample of households. I would like to explain why.
Every month, all establishments must report jobs and wages on a form that is used to pay unemployment taxes. Any establishments that have employees (but not contract workers) must send in this form.
However, a total enumeration of all the jobs reported on those forms would be too costly. Therefore, the Bureau of Labor Statistics takes a sample each month to help develop the report on jobs by sector that is announced every month in the jobs report.
Every March, a full enumeration of these jobs is made to determine the sample population from which the monthly sample will be derived for the next year. In the meantime, companies that fail cease to send in their forms. Many new companies are not included until a new sample from the next enumeration is made.
Of course, the Bureau of Labor Statistics makes estimates of the creation of new enterprises to adjust for this problem of the births and deaths of firms. The March enumeration then is used to adjust the estimators for the past year and establish a new set for the new year.
The 21,000 job growth mentioned above is the result of this establishment process. Last year’s revision to this process already has occurred. Although most economists expected a significant upward revision in the employment estimate, the adjustment was actually downward. In Atlanta, for example, a 65,000 job gain was reduced to slightly over 20,000 new jobs because of this adjustment.
I am belaboring this point because some commentators have been trying to explain away the job declines during President Bush’s initial three years in office by shifting from the establishment estimates where the job declines are measured to the household survey.
Actually, the household survey does not sample jobs. Instead it interviews people. The purpose of the survey is to determine what type of activity people are doing. They may be out of the workforce because of alternative uses of their time or because they are discouraged from finding jobs. They may be actively seeking work (at least actively getting unemployment compensation) but have not yet found any work. Of course, they also may be actively working.
The 60,000 sample clearly is sufficient to determine the percentage of our population that is unemployed, if they respond truthfully to the survey questions. Therefore, this survey is used to determine the unemployment rate. If we know the size of the total population, we also can estimate the size of the active labor force and derive the number of people employed.
Because of tightened immigration after 9/11, our population apparently has begun to grow more slowly. At least that was the conclusion reached in January, when employment and labor force estimates were revised downward. (I leave it to the statisticians to explain why the unemployment estimate still remains valid even as the size of the population is reduced slightly).
There are many reasons why the household and establishment estimates of employment are not the same. Contract workers do not generate an unemployment compensation form. Instead, they are included in the self employed, which are not captured by the establishment measure.
Also, 7.2 million people have more than one job. Remember, one survey is about people and the other is about jobs. Multiple jobs by people clearly create a difference between the two.
Throughout the 1990s, the job measure grew faster than the implied employment from the survey of people. Since President Bush took office, the people survey shows considerably more job growth than the establishment survey.
So which one is right?
There clearly could be long term trends that are leading to more self-employed. That the self-employed surged by nearly 2 million while all other jobs fell by 2.3 million is highly unlikely, however. We know that those holding multiple jobs have declined. This eliminates about a fourth of the discrepancy between the two series.
Frankly, I believe that unemployment could not have declined by more than half a percentage point without creating more than the 300,000 jobs that the establishment survey shows over the same period of time. That is why I was expecting an upward revision in the establishment estimates.
It is still possible that those revisions will surface from the next benchmark adjustment.
Even so, the household survey is not designed to measure jobs. Just a few unemployed professionals who are too embarrassed to say they are out of work can distort the survey. Never before have so many professionals lost their jobs in such a mild recession.
In short, neither probably captures the full picture of job growth. However, there is method in why economists overwhelming choose the establishment measure over the household measure. Does anyone seriously believe we lost a quarter million jobs last month?
So lets stop trying to solve the sluggish jobs problem by changing the measures. Otherwise, give me a believable reason why self-employment is surging while jobs subject to unemployment compensation are sagging.