Measuring County-Level Deindustrialization in the United States

Existing measures of county-level deindustrialization in the United States are oftensingle numbers, typically capturing manufacturing job losses over a short time. Thismisses some nuance because job loss trajectories are not always linear and the startingpoint is consequential. I implement a time series clustering algorithm to categorizecounties based on trends in manufacturing jobs over 53 years for the 2,316 countieswith available data. Six distinct clusters represent the major deindustrialization trajectories.Declines have been quite sharp in some clusters, giving those counties lesstime to transition to a post-industrial economy. These cross-cluster differences areconsequential to theories and past findings. For example, I find that TAA claims, atraditional measure of local deindustrialization, meaningfully impact individuals’ viewson investment only when they live in non-deindustrialized counties. In deindustrializedcounties, views are homogeneous across demographics, except across partisanship,where Republicans are more supportive of investment.