Two Cleveland boys -- Mark Schweitzer and Guhan Venkatu -- were cited in a Greg Ip article published in Monday's Wall Street Journal. (The article was on page A2. Unfortunately web links are available to subscribers only.)
In their original article, Mark and Guhan note the strikingly different picture about the US jobs picture that emerges if attention is focused on a measure derived from a survey of households versus one based on a survey of businesses. They describe several differences across the two surveys that might account for the discrepancy.
The household survey (officially known as the Current Population Survey) limits its sample to individuals over the age of 16 in the civilian, noninstitutional population (that is, those not in the military, prisons, or long-term care or nursing-home facilities). The establishment survey (known officially as the Current Employment Survey and often unofficially as the payroll survey), by contrast, surveys a sample of nonagricultural work sites each month. The household survey thus includes agricultural workers, the self-employed, and paid and unpaid family workers, whereas the establishment survey does not.
There is also a difference in terms of the size of the samples from which information is taken.
The household survey’s monthly sample (of individuals in about 60,000 households) constitutes approximately one person for every 3700 people in the population... The establishment survey’s monthly sample (about 400,000 worksites) covers approximately one-third of all of the workers in the population that it attempts to cover... The survey’s broad coverage of its target population is a key advantage of its approach.
The two authors go to some considerable length to explain other differences, and to make the data contained in each of the series as comparable as possible. In the end, they come to this conclusion:
The recent debate over the state of the labor market during this recovery has centered on the question of whether the employment estimates from the household survey or the establishment survey are the more accurate. There seem to be good arguments for concluding that... the employment estimates produced by the establishment survey are more accurate than those produced from its counterpart, the household survey.
Having acknowledged this, however, it is not necessary to disregard the household survey altogether. When used appropriately, there is much that it can tell us. And, moreover, when used in a way that minimizes the role of population estimates, it tends to paint a picture of the labor market that is very
much in keeping with that presented by the establishment survey.