Saturday, October 06, 2012

The Two Labor Market Surveys

If you go to the recent release from the BLS, you can find these two sentences a few paragraphs apart:

Total employment rose by 873,000 in September.

Total nonfarm payroll employment increased by 114,000 in September.

To a layman, this may seem confusing.  The first statement suggests a robust labor market, the second a more lackluster one.  What is going on?

The issue is that there are two surveys.  The first estimate of employment comes from the survey of households; the second is from the survey of establishments.  I thought readers might like to hear what my favorite intermediate macro textbook says about this issue.  Here is an excerpt:


Because the BLS conducts two surveys of labor-market conditions, it produces two measures of total employment. From the household survey, it obtains an estimate of the number of people who say they are working. From the establishment survey, it obtains an estimate of the number of workers firms have on their payrolls.

One might expect these two measures of employment to be identical, but that is not the case. Although they are positively correlated, the two measures can diverge, especially over short periods of time. A particularly large divergence occurred in the early 2000s, as the economy recovered from the recession of 2001. From November 2001 to August 2003, the establishment survey showed a decline in employment of 1.0 million, while the household survey showed an increase of 1.4 million. Some commentators said the economy was experiencing a “jobless recovery,” but this description applied only to the establishment data, not to the household data.

Why might these two measures of employment diverge? Part of the explanation is that the surveys measure different things. For example, a person who runs his or her own business is self-employed. The household survey counts that person as working, whereas the establishment survey does not because that person does not show up on any firm’s payroll. As another example, a person who holds two jobs is counted as one employed person in the household survey but is counted twice in the establishment survey because that person would show up on the payroll of two firms.

Another part of the explanation for the divergence is that surveys are imperfect. For example, when new firms start up, it may take some time before those firms are included in the establishment survey. The BLS tries to estimate employment at start-ups, but the model it uses to produce these estimates is one possible source of error. A different problem arises from how the household survey extrapolates employment among the surveyed households to the entire population. If the BLS uses incorrect estimates of the size of the population, these errors will be reflected in its estimates of household employment. One possible source of incorrect population estimates is changes in the rate of immigration, both legal and illegal.

In the end, the divergence between the household and establishment surveys from 2001 to 2003 remains a mystery. Some economists believe that the establishment survey is the more accurate one because it has a larger sample. Yet one recent study suggests that the best measure of employment is an average of the two surveys. [George Perry, “Gauging Employment: Is the Professional Wisdom Wrong?,” Brookings Papers on Economic Activity (2005): 2.]

More important than the specifics of these surveys or this particular episode when they diverged is the broader lesson: all economic statistics are imperfect. Although they contain valuable information about what is happening in the economy, each one should be interpreted with a healthy dose of caution and a bit of skepticism.