From a Merrill Lynch Research Report, Macro-Economics: The Employment Benchmark Revision: A New Event
The full report can be found at http://www.cwes01.com/9093/24013/ds/03513252.pdf
The October 8th employment report will include a preliminary estimate of the size of the annual benchmark revision to the payroll survey set to take place next February. To be sure, there is some scope for an upward benchmark revision. However, the size of the revision is likely to fall short of expectations. We see the revision as no more than 50k-100k (which would be 4k-8k per month), with even a possibility that the revision will be negative, as it has been in the past three years.
Every year, the BLS benchmarks the payroll establishment survey with a more complete data set from the state unemployment insurance programs. These data more accurately capture the hiring of newly created firms and firms that go out of business (known as the Birth / Death adjustment). The net creation of new firms tends to move with the business cycle, with more firms being created and less destroyed during business cycle upturns.
Previous benchmark revisions, from 1994 to 2000 were quite sizable. In fact, payrolls were revised higher by almost 0.7% (almost 800k) in 1994, with much of the revision stemming from poor handling of the birth / death adjustment. Since then, new methodology by the BLS has made the benchmark revisions much smaller. Over the past three years there have been small downward revisions ranging from 0.1% to 0.2%.
Since net business creation turned up over 2003 it is possible that the BLS's birth / death adjustment has not fully incorporated the pace of current economic activity. For that reason, we expect a small upward revision to be announced on Friday. However, extrapolating the state unemployment program employment data, for which data are only currently available up to December 2003, and using the recent trends in private sector payroll growth, leaves open the possibility of a small downward benchmark revision.
Many are hoping this Friday's revisions will help resolve the puzzling and persistent gap between the job growth reported in the household survey (3.3 million new non-farm jobs created since November 2001) and the payroll survey (604k new jobs created over that same period). This gap, however, is more likely explained by an over counting in the household survey. The household survey uses Census Bureau population controls that may be over counting immigration flows into the US. We believe that the gap in employment trends between both surveys will likely remain relatively wide until the population issues are resolved.
thanks to doug henwood at lbo-talk for making this available
http://mailman.lbo-talk.org/pipermail/lbo-talk/Week-of-Mon-20041004/022451.html