Equivalence/differences in UnionStats.com and BLS news release estimates of union membership and density
Both the BLS and UnionStats use the Current Population Survey (CPS) outgoing rotation group monthly files to compile annual estimates of union membership and density. The same methodology and identical definitions of union membership, representation (coverage), and density are used. With the exception of 1994, 2001, and 2002 (see below), estimates for the years 1983 through 2010 have been identical for economy-wide estimates provided by both BLS and UnionStats. As compared to UnionStats, BLS provides fewer significant digits (i.e., more rounding) and does not provide estimates at such disaggregated levels as does UnionStats due to concern about small sample sizes (e.g., detailed industry, detailed occupation, metropolitan areas, and private/public by state).
Beginning in 2011, estimates from BLS and from UnionStats differ trivially due to differences in the internal Census/BLS data files and publicly available CPS files. A new age swapping procedure was adopted in the CPS beginning in 2011 due to concerns about confidentiality. Public release files now have slightly different sample weights attached to individuals as compared to the internal files used by BLS to compile their estimates. Differences in published estimates by BLS and those shown at UnionStats are inconsequential and do not affect reliability. Two noticeable differences are that the “headline” overall density rates in 2012 and 2013 just rounded up to 11.3% in BLS published reports and rounded down to 11.2% in Unionstats using the publicly available CPS data. The chief reliability concern at UnionStats continues to be sampling and reporting error for estimates within narrow categories based on small sample sizes (sample sizes on which estimates are based are reported in all tables in the first column, labeled “obs”).
For the years 1983-2010 (BLS does not provide CPS union estimates for earlier years), there exist trivial differences in estimates from BLS and UnionStats in 1994, 2001, and 2002. Differences In 1994 were due to errors in the internal CPS files used by the BLS (public use files were corrected, but to avoid confusion BLS did not release revised figures). In 2001, BLS compiled union estimates using CPS files with a special half-year over-sample not included in the public use files. In 2002, BLS included revised weights in their internal files not initially included in the public use version (these were subsequently made public).