Comment

Kerwin Kofi Charles, Erik Hurst, and Mariel Schwartz provide an insightful and comprehensive empirical examination of the link between the transformation of the US manufacturing sector and the substantial decline in the employment rates and average annual hours worked of prime-age adults (those aged 21–55), especially men and less educated women, since 2000. Charles et al. document that US manufacturing experienced a massive decline in employment of 5 million jobs, a large increase in capital intensity, and substantial skill upgrading from 2000 to 2017. They exploit geographic variation in manufacturing employment decline across commuting zones (CZs) in the 2000s using a Bartik (shiftshare) instrument to assess the “causal” impact of local manufacturing employment demand shocks on employment outcomes. Charles et al.find that CZs with larger manufacturing employment declines have larger declines in employment rates, hours worked, and wages for prime-age workers from2000 to 2016, regardless ofwhether the adversemanufacturing shocks were related to China trade shocks or other sources. Charles et al. conclude that manufacturing decline can account for about half of the prime-agemale employment rate decline since 2000. Geographic areas with bigger manufacturing employment losses in the 2000s also experience greater social problems as seen in more severe drug and opioid addiction problems. Finally, Charles et al. show that adversemanufacturing employment shocks have generated less geographicmobility (less of a regional migration response) and more persistent impacts on employment rates in the 2000s than in the 1980s. I find the analysis of Charles et al. to bewell crafted and quite convincing. Their findings of adverse labor demand shocks against less educated workers being a driving force in declining employment rates in the 2000s complement other work using cross-area variation in China trade exposure (Autor, Dorn, and Hanson 2013) and automation shocks