U.S. labor markets are increasingly diverse and persistently unequal between genders, races/ethnicities, educational groups and age. In the spirit of Chari, Kehoe, and McGrattan (2007), we use a structural model approach to decompose the observed differences in labor market outcomes across demographic groups in terms of underlying wedges, or frictions. Of particular interest is the potential role of discrimination, either taste- based or statistical. Our prototype model is a version of the Diamond- Mortensen-Pissarides model extended to include a life cycle, learning by doing, a non-participation state, and informational frictions. The prototype exhibits group-specific wedges in initial human capital, returns to experience, matching efficiencies, and hazard rates. We use the model to reverse engineer group-specific wedges which we then feed back into the model to assess the fraction of various disparities they account for. Applying this methodology to 1998-2018 U.S. data reveals that differences in initial human capital, returns to experience, and in hazard rates, account for most of the demographic disparities; wedges in matching efficiencies play a secondary role. Our results suggest a minor aggregate impact of taste-based discrimination in hiring and an important role for statistical discrimination affecting particularly female groups and Black males. Our approach is macro, structural, unified, and comprehensive.