Has U.S. employment really polarized? A critical reappraisal

We re-examine whether U.S. workers have become increasingly concentrated in low and high-wage jobs relative to middle-wage jobs, a phenomenon known as employment polarization. We find that the typical occupation-based approach of previous literature has significant weaknesses, leading us to prefer a worker-based approach. At both the occupation and individual level, we do find a decline since 1973 in the share of workers earning middle wages. However, the large increase in the share of high-paid workers is not accompanied by any substantial increase in the share of low-paid workers—inconsistent with employment polarization. In particular, we find that the share of employment in low-wage occupations did not increase during the 1990s, and that the apparent growth (and therefore polarization) found in the previous literature is an artefact of occupation code redefinitions. We therefore do not find support for the view that employment was polarizing during the 1990s (whether because of automation or other factors).