Employment Data ABCs
by ilene - February 9th, 2010 4:56 pm
Employment Data ABCs
Courtesy of John Lounsbury at Piedmonthudson’s Weblog
Comments on employment data, especially over the past few days, indicate that many are confused about what the data means, how reliable/unreliable it is, and whether or not it is politically motivated. I have written past articles in which many details of how the DOL (U.S. Dept. of Labor) numbers should and should not be used. One of these (here) was a very detailed examination of various aspects of DOL processes. In this article, I want provide a summary update of the earlier work and add some new details.
Political Manipulation
I will address the third concern first. I can find no evidence that there is any political influence whatsoever in the collection and analysis of the employment data. The processes are well defined and stable over long periods of time. In fact, this very stability can produce errors in analysis models that need to be corrected when the historical form of the models starts to deviate from new economic reality.
The Birth/Death Adjustment
The most notable example of model drift is the birth/death adjustment, which has recently overstated the estimate of jobs created by new business formation before the new companies actually show up in the DOL’s Establishment Survey. These estimates are adjusted for 12 month data intervals almost a year after the fact, when new business formation can be more accurately accounted for from state tax and business records.
The ex post facto corrections made necessary by this “wandering model” were applied this month to data for April 2008 through March 2009. A total of 930,000 non-farm payroll job losses were added to those twelve months. Interim adjustment corrections were also made this month for April through December 2009. This will presumably reduce the corrections needed in February 2011 for the period April 2009 through March 2010.
The Establishment Survey
This is one of two monthly surveys conducted by the DOL. It covers approximately 140,000 businesses and government agencies (~410,000 work locations). The output of this survey consists of employment analysis of various segments of the economy, such as retail, manufacturing, construction, mining, transportation, education and health services, government, etc. The average weekly hours worked is also determined by this survey. It is the data used to produce the non-farms payroll report.
In spite of the fact that the non-farms…