Working Pauer 93 14 ACCOUNTING FOR EARNINGS INEQUALITY IN A DIVERSE WORK

Groups that are often treated separately in studies of earnings inequality -women and minorities -are included in a general decomposition of sources of such disparity among fulltimelfull-year labor force participants. In order to consider several possible sources of earnings inequality, a new decomposition is developed that allows semilog earnings models to be decomposed into inequality factors defined by regression coefficients. Simultaneous estimation of inequality factors is justified if correlations between the effects of factors or between factors and the earnings of subgroups exist. Results show that incorporating the covariances between earnings factors and the earnings of subgroups is necessary for an accurate picture of the sources of earnings inequality. Thenew decomposition confirms that with the complete labor force included, education remains the largest observable factor in the rise of U.S. earnings inequality in the 1980s. The decomposition also indicates that industry wage differentials have played a contributing role. A large decline in raceand sex-related earnings inequality hides factors capable of explaining the rise in earnings inequality. Finally, a surprising amount of the change in measured inequality is traced to shifts in the supply of workers with given characteristics rather than to shifts in relative earnings. http://www.clevelandfed.org/Research/Workpaper/Index.cfm