Did the Rust Belt Become Shiny? A Study of Cities and Counties That Lost Steel and Auto Jobs in the 1980s

Abstract

Did the Rust Belt Become Shiny? A Study of Cities and Counties That Lost Steel and Auto Jobs in the 1980s James Feyrer Dartmouth College Bruce Sacerdote Dartmouth College Ariel Dora Stern Federal Reserve Bank of New York [Comments] How do counties, cities, or regions respond to adverse economic shocks? How quickly does an area recover and through which adjustment mechanisms? These questions touch on many different areas of social science and economics and are relevant to our understanding of economic growth, income gaps across regions (for example, North and South in the United States or Italy), and the plight of individual laid-off workers and their families. In this paper we undertake a study of one of the biggest negative shocks to affect the U.S. economy in the past fifty years, namely, the massive loss of steel- and auto-related jobs in the early 1980s, which we refer to collectively as the Rust Belt shock. In the decade between 1977 and 1987 the United States shed about 500,000 jobs in the auto industry and 350,000 jobs in the steel industry, far outstripping any other job losses in recent U.S. history. These job losses were concentrated in roughly 140 of the 3,000 counties in the United States. Kahn as well as Black, McKinnish, and Sanders discuss the size of the manufacturing shocks and accompanying job losses.1 For the first section of our paper, we assemble total employment, industry-level employment, population, labor force participation, and income data at the level of the county and the metropolitan statistical area (MSA). Our basic approach is to regress short- and long-run changes in outcomes on the size of [End Page 41] the Rust Belt shock. Consistent with the state-level analysis in Blanchard and Katz, we find very rapid recovery in the unemployment rate in Rust Belt cities and counties.2 Within five years, unemployment rates in the Rust Belt areas returned to the U.S. average.3 The adjustment took place entirely through out-migration of people rather than in-migration of jobs or a change in labor force participation. Each steel or auto job lost in a county led to a net decrease in population of 1.8 persons. In the long run, population in Rust Belt counties tended to stay flat or slightly below 1977 levels, likely due to the Glaeser and Gyourko effect of durable housing.4 After establishing the basic facts of shock and recovery, we attempt to distinguish among successful and unsuccessful shock places based on the long-run growth of population in these counties. The most successful post-shock counties tend to be those that are located in warm sunny climates such as Jefferson County (Alabama) or near a major city, for example, Carbon County (Pennsylvania).5 The Glaeser and Saiz relationship between city growth and human capital holds within our general sample of all counties, but we do not have enough statistical power to identify the effect within our sample of shock counties.6 Like Kahn as well as Glaeser, Kolko, and Saiz, we believe that city- or region-level amenities are a key input to an area's desirability and growth.7 We find that, while Rust Belt counties and MSAs recovered quickly on certain dimensions like unemployment and income per capita, amenities in these places experienced a negative shock that did not diminish over time and, if anything, worsened. This effect may be a simple result of declining population. However, it is also possible that our Rust Belt counties (which are disproportionately in the North) have underutilized infrastructure and abandoned capital that are particularly ugly, even though deindustrialization has made the air and water cleaner and safer.8 We hypothesize that the lack of amenities, [End Page 42] more than crime or unemployment per se, is what contributes...