Effect of union decline on male wage differentials, 1978–2011

		1978	1989	2000	2011
Percent of workers in union (“union coverage”)
By occupation	White collar	14.7%	12.1%	11.2%	10.3%
	Blue collar	43.1%	28.9%	23.1%	17.8%
	Difference	-28.4	-16.7	-11.9	-7.5
By education	College	14.3%	11.9%	13.1%	12.1%
	High school	37.9%	25.5%	20.4%	14.9%
	Difference	-23.6	-13.6	-7.4	-2.9
Union wage effect*
By occupation	White collar	0.2%	0.0%	-0.2%	-0.2%
	Blue collar	11.5%	6.7%	4.3%	3.5%
	Difference (change in differential)	-11.3	-6.8	-4.5	-3.6
By education	College	0.9%	0.5%	0.9%	0.6%
	High school	8.2%	5.5%	3.1%	2.6%
	Difference (change in differential)	-7.3	-5.0	-2.3	-2.0
		1978–1989	1989–2000	2000–2011	1978–2011
Change in wage differential**	White-collar/blue-collar	5.0	4.2	0.9	10.1
	College/high school	13.0	8.0	2.8	23.9
Change in union wage effect	White-collar/blue-collar	-4.6	-2.3	-0.9	-7.7
	College/high school	-2.3	-2.5	-0.3	-5.1
Deunionization contribution to change in wage differential***	White-collar/blue-collar	-90.5%	-55.2%	-91.8%	-76.1%
	College/high school	-17.8	-30.7	-10.2	-21.2

* Union wage effect is "union wage premium" (estimated with simple human capital model plus industry and occupational controls) times union coverage; negative values in the difference row show how much unionization narrowed the wage gaps.

** Log wage gaps estimated with a simple human capital model

*** Change in union wage effect on wage differential divided by overall change in differential

Source: Authors' update of Freeman (1991) using Current Population Survey Outgoing Rotation Group microdata

Documentation and methodology

This analysis replicates, updates, and expands on Freeman (1991), Table 2, using the CPS-ORG sample used in other analyses (see Appendix B). The year 1978, rather than 1979, is the earliest year analyzed because we have no union membership data in our 1979 sample. “Percent union” is the share covered by collective bargaining. The “union wage premium” for a group is based on the coefficient on collective bargaining coverage in a regression of hourly wages on a simple human capital model (the same one used for estimating education differentials, as described in note to Table 4.13), with major industry (12) and occupation (9) controls in a sample for that group. The change in union premium across years, therefore, holds industry and occupation composition constant. Freeman’s analysis assumed the union premium was unchanged over time. We allow the union premium to differ across years so changes in the “union effect” on wages (the union wage premium times union coverage) are driven by changes in the unionization rate and the union wage premium. The analysis divides the percentage-point change in the union effect on wage differentials by the actual percentage-point change in wage differentials (regression-adjusted with simple human capital controls plus controls for other education or occupation groups) to determine the deunionization contribution to the change in the wage gaps among men, which, as a negative percent, indicates contribution to the growth of the wage gaps.