The Polarization of the U.S. Labor Market

Much research (surveyed in Katz and Autor, 1999) documents a substantial widening of the U.S. wage structure since the late 1970s, driven by increases in educational wage differentials and residual wage inequality. The growth in wage inequality was most rapid during the 1980s, and involved a spreading out of the entire wage distribution. Rapid secular growth in the demand for skills, partly from skill-biased technical change (SBTC), combined with a slowdown in the growth of the relative supply of college workers helps explain these wage changes. Eroding labor market institutions—the minimum wage and unions—further contributed to rising wage inequality. Recent work emphasizes a slowing of wage inequality growth over the last 15 years (David Card and John DiNardo, 2002; Thomas Lemieux, forthcoming). This “revisionist” literature views the 1980s surge in wage inequality as an “episodic” event caused by institutional forces and argues that “modest” inequality growth in the 1990s is inconsistent with a key role for SBTC. We reconsider this revisionist view, focusing on a marked change in the evolution of the U.S. wage structure over the past 15 years and divergent trends in upperand lower-tail wage inequality. We first document that wage inequality in the top half of the distribution has exhibited an unchecked secular rise for 25 years, but it has ceased growing since the late 1980s (and for some measures narrowed) in the bottom half of the distribution. We next demonstrate that employment growth differed sharply in the 1990s versus the 1980s, with more rapid growth of employment in jobs at the bottom and top relative to the middle of the skill distribution. Borrowing terminology from Maarten Goos and Alan Manning (2003), we characterize this pattern as a “polarization” of the U.S. labor market, with employment polarizing into highwage and low-wage jobs at the expense of middle-skill jobs. We then show how a model of computerization in which computers complement nonroutine cognitive tasks, substitute for routine tasks, and have little impact on nonroutine manual tasks, can rationalize this polarization pattern.