Micro Estimates of the Union-Nonunion Wage Differential in Canada Using a Sample Selectivity Approach

This paper demonstrates empirically that the limited dependent variable technique for sample selectivity bias correction provides two very different estimates of the gap between union and non-union wages. Our findings, based on the Canadian Survey of Work History, suggest that selectivity correction factors are included, wage gap estimates are conditional differentials and are comparable to OLS estimates. Excluding correction factors leads to larger potential differentials based on wage distributions. This distinction probably explains the volatile estimates of the union-nonunion wage gap for the U.S.