Philadelphia Federal Reserve Bank Questions BLS State Level Employment Data Revisions
Kurt Weiss at the Michigan Department of Technology, Management, and Budget (DTMB) reported on March 25th that Michigan’s February 2015 seasonally adjusted unemployment rate dropped to 5.9% from January’s 6.3% figure. This was followed on April 2nd by the DTMB release of non seasonal adjusted employment data for February 2015. This quite large one month drop is great news for Michigan workers, if it indeed reflects our labor market. But does it?
Twelve Month Discrepancies Between Seasonally Adjusted and Unadjusted Data Differentials
The DTMB Michigan non seasonally adjusted (NSA) employment report for February 2015, at the bottom of page 6, showed that Michigan’s civilian labor force declined by 33,000 workers over the twelve preceding months. But the seasonally adjusted (SA) employment report for February 2015 at the top of page 2 showed that Michigan’s civilian labor force increased by 7,000 workers over the same twelve months. The growth of Michiganders employed over the same twelve month period also shows a 5,000 worker NSA/SA discrepancy: 101,000 not seasonally adjusted versus 96,000 seasonally adjusted. And the shrinkage of Michigan’s unemployed worker population over the same twelve month period shows a 46,000 worker NSA/SA discrepancy: 134,000 not seasonally adjusted versus 88,000 seasonally adjusted.
Valid seasonal corrections should produce twelve month differentials which agree with comparable item non seasonally adjusted differentials over the same twelve month period. Here we have a 0.85% discrepancy in Michigan’s total civilian labor force, a 0.1% discrepancy in civilian employment over the same twelve month period, and a 12.5% discrepancy in the shrinkage of Michigan’s unemployed worker population over the same twelve month period.
How did this happen?
The DTMB turns its employment data over to the U.S. Department of Labor’s Bureau of Labor Statistics for X-13ARIMA-SEAT seasonal adjustment. BLS also adjusts underlying data periodically for changes in the Michigan population estimates of the U.S. Census. The two (SA and NSA) February 2015 DTMB employment reports both reflect substantial BLS revisions undertaken in early 2015. A note at the top of page 2 in the DTMB February 2015 SA employment report:
Note: The data in this release reflects recently revised historical estimates. Seasonally adjusted labor force estimates for 1976-2014 for Michigan and 1990-2014 for the Detroit-Warren-Dearborn MSA were revised. Previous published data should be replaced with this new series. In addition, seasonally adjusted payroll job data was revised for 2010-2014. For newly revised data, please contact DTMB at 313-456-3090.
The DTMB February 2015 NSA employment report bears a similar note on its first page noting that it was also subjected to substantial revisions.
So Michigan’s employment data has been subjected to substantial revisions at BLS which produce seasonal adjustments that do not cross check against comparable item non seasonally adjusted data over the most recent 12 month period.