A Meta-Analysis of the Union-Nonunion Wage Gap

The authors use meta-regression analysis, a new method for synthesizing empirical results reported in the economic literature, to corroborate and extend H. Gregg Lewis's landmark 1986 research on the union-nonunion wage gap. Analyzing 152 observations from 114 studies, they investigate how different model specifications, data sets, and variable definitions can affect the estimates of the wage gap. An important finding is that many adjustments for differences among studies, such as the use of actual or weekly earnings rather than logarithmic wages, are themselves time-dependent. The authors' annual estimates of the union-nonunion wage gap for 1967–79 vary directly with the national unemployment rate and range from 8.9% to 12.4%, lower than Lewis's range of 9.6% to 16.4%.