Re-examining the models used to estimate disability-related wage discrimination

We examine how differences in model specifications and econometric methods affect unexplained wage differentials between workers with and without physical disabilities, where the unexplained differentials are estimates of the potential effects of disability-related wage discrimination. We apply an enhanced decomposition method (Neuman and Oaxaca, 2004) to selectivity-corrected wage equations to estimate potential discrimination effects. The decomposition separates observed wage differentials into a part explained by differences in characteristics that affect productivity and the decision to work and an unexplained part potentially attributed to discrimination. In addition to the functional limitation variables used to control for the effects of disability on productivity, we add measures of physical job demands, and interaction effects between functional limitations and job demands, to the wage models to examine how estimates of potential discrimination change. The interaction terms measure the extent to which workers' physical limitations affect important job functions. Data come from the 2004 SIPP merged to measures of job demands from the Occupational Information Network (O*Net). With job demands and interactions included in the model, approximately 10% of the observed wage differential for men, and 20% for women, is potentially attributed to discrimination. Changes in decomposition technique substantially alter the estimates of discrimination effects.