i (3)
During the past 25 years the fraction of U.S. workers belonging to trade unions has fallen dramatically (see e.g. Farber, 1990; Riddell, 1992) while the level of wage inequality has risen (e.g. Blackburn, Bloom, and Freeman, 1990; Bound and Johnson, 1992). Since wage dispersion is generally lower among union workers than their nonunion counterparts (e.g. Freeman and Medoff, 1984) there is a presumption that the fall in unionization may have contributed to the rise in wage inequality. Indeed, a number of recent studies — including Freeman (1993) and DiNardo, Fortin, and Lemieux (1996) — estimate that the fall in union membership can account for about one-quarter of the rise in male wage inequality over the 1980s.

This paper re-assesses the connection between changing unionization and wage dispersion for men and women over the period from 1973 to 1993. The methodology departs from the two-sector framework developed by Freeman (1980) for studying the equalizing effect of trade unionism in two important ways. First, explicit attention is paid to the fact that unionization rates vary systematically across the wage distribution. While union members were concentrated at the “middle” of the wage distribution in the 1970s, union densities have fallen most rapidly among the least-paid workers, and have actually risen among the most highly-paid. These trends have reduced the equalizing effect of unionism in the economy. Second, the methodology accounts for differences in the relative wage effect of unions for more- and less-skilled workers. Conventional estimates of the union wage gap for low-skilled workers are large and positive (over 30 percent) while estimates for highly-skilled men are significantly negative. Taken at face value these estimates imply a substantial equalizing effect of unions. Evidence presented in Card (1996), however, suggests that some part of the large positive union-nonunion wage gap for low-skilled workers is attributable to the higher unobserved productivity characteristics of union workers with low observed skills, while the negative gap for high-skilled workers is attributable to the higher unobserved characteristics of nonunion workers with higher observed skills. Estimates of the equalizing effect of unions that these ignore differential selectivity biases (e.g. Freeman’s two-sector method, or the re-weighting method used by Dinardo, Fortin and Lemieux) may therefore overstate the role of unions in compressing wage differences across the skill distribution.

Tags: , ,