Reporting/Market Focus from the April 2005 Edition of the SGS Newsletter. 

My friend John Crudele, financial columnist with the New York Post, recently reminded me of a further long-term problem the Bureau of Labor Statistics has with its reporting consistency, beyond the myriad issues already discussed in the SGS. Separate from the BLS’s inability to reconcile to within 1,000,000 jobs the household survey of employed people with the payroll survey of jobs, the BLS cannot even reconcile the sub-components within the payroll survey to within 500,000 jobs.
With irregular timing, ranging from weeks to over a month after the regular employment report, the BLS publishes regional and state employment on a monthly basis. The state-by-state data are compiled from the same household and payroll surveys used in national reporting, so one might assume that a total of employment in the fifty states plus the District of Columbia would equal the U.S. total. Such is not the case.
The March 31st release showed that seasonally-adjusted February payrolls by state totaled 132.285 million. The April 1st release of March payroll data showed a revised national total for February of 132.816 million, which had been 132.843 million in original reporting. The difference of 531,000 to 558,000 jobs is meaningful, especially considering that both sets of reporting encompassed the recent major benchmark revision to the payroll or establishment survey.
Part of the reporting difference is due to different assumptions on the bogus bias factors (the net birth/death model of company formations) and difference in seasonal factors. Nonetheless it is hard to believe that the BLS, after decades of these inconsistencies, still cannot get the same national total when jobs are added up by state or estimated solely at the national level.
The BLS explains that, "State estimation procedures are designed to produce accurate data for each individual state." The BLS then goes on to caution against aggregating the state data and comparing them with the independently derived national employment series. The problem, in theory, is that "each state series is subject to larger sampling and nonsampling errors than the national series."
The funny thing, though, is that those big individual state errors seem to balance out when the state numbers are added to together. The aggregated numbers, instead of jumping all over the place, show a consistent bias to the downside against the "national" estimates. There seems to be some missing political component from the state totals that is needed to come up the national growth being reported by those great statistical manipulators in Washington, D.C.
Indeed, the state totals offer an alternate version of what is happening in the labor market. As a rule of thumb, the less a series is followed by the popular press, the less likely it is to suffer political manipulation. Accordingly, the national totals by state may actually offer a more accurate, less-politicized version of employment change than the glamorized and highly publicized and politicized "official" nonfarm payroll totals.
For example, per the most recent national report, seasonally-adjusted payrolls in February 2005 were up by 2,350,000 from February 2004. The same change calculated by summing all the states shows payrolls up by 1,919,700 for the year ended February 2005. The latter gain is 18.3% less than the official one.
In like manner, the revised and official monthly payrolls gains for December 2004, January 2005 and February 2005 are 155,000, 124,000 and 243,000, respectively, versus 102,900, 61,300 and 197,300 based on state totals. The three-month jobs gain based on state totals was 361,500, 30.7% less than the official 522,000 gain. The trend in discrepancies is getting worse, not better.
Beyond suggesting that the national data may be overstating economic growth, the state data also give a rough picture of varying regional economic activity. Rarely are all sections of the country experiencing the same business environment at the same time. The aggregated state payrolls in February 2005 were 1.5% ahead of February 2004 (versus 1.8% for the official national growth). States that have jobs growth more than one percent above (2.5%) or below (0.5%) the national state average probably are seeing respectively relatively good and bad times. Keep in mind, however, that these state numbers are far from perfect and the national state average is not showing a booming economy.
The only state on the boom side is Nevada, with annual jobs growth of 6.4%. States with generally stronger growth of 2.5% to 4.0% are Utah, Arizona, Colorado, Idaho, Oregon, Hawaii, Florida and Delaware. Michigan is in recession, with employment down 0.2% year-to-year. States with minimal 0.0% to 0.5% growth include Ohio, Illinois, South Carolina, Louisiana and the District of Columbia.