## RISING WAGE INEQUALITY: Allowing for Differences Across Skill Groups 3

The calculations summarized in Table 5 use equation (6) together with the data in Table 4 to re-estimate the contribution of changing unionism to rising wage inequality. I have computed the effects two ways: using the unadjusted union-nonunion wage gaps presented in Table 4; and using the adjusted wage gaps derived from my 1996 paper. For women, the results are quite similar to the results of the naive calculations in Table 3: changes in unionism have had a negligible effect on wage inequality. For men, the results are qualitatively similar to the naive calculations but the magnitude of the union effect is reduced: from about 30 percent of the overall rise in inequality under the naive calculation to 18 percent using the more sophisticated model with the unadjusted wage gaps to 12 percent using the adjusted gaps.

In principle, it is also possible to implement equation (6) using longitudinally-based estimates of the union variance effect (Av) rather than the simple differences in the variances of wages between the union and nonunion sectors shown in Table 4. Card (1992) and Lemieux (1992) both present estimates of the effect of unions on the variance of wages based on the wage outcomes of union status changers. The longitudinal variance gap estimates presented in Card (1992) are relatively noisy, and on average only slightly smaller in absolute value than the corresponding cross-sectional estimates. Lemieux’s estimates are also noisy but tend to be noticeably smaller (in absolute value) than the cross-sectional estimates. If the cross-sectional variance gaps in Table 4 are viewed as bounding the likely effect of unions on wage dispersion, then the estimates in Table 5 should be interpreted as upper bound estimates of the contribution of changing unionization to rising wage inequality. Taken as whole, then, it appears that the effect of unions on widening wage inequality may be relatively modest.