Comments
Albert Saiz: The paper examines the aftermath of a major negative economic shock: the massive loss of auto and steel jobs in a number of U.S. counties in the early 1980s. The authors consider three time periods: the shock period (1977–82), the recovery period (1982–87), during which presumably no further major job losses occurred, and the long run in the aftermath of the auto and steel crisis (1987–2004). The authors basically compare the evolution of a number of outcomes in "shock" counties (that is, counties affected by the auto and steel job losses during the 1977–82 period) and all other counties in the United States.
The results in the paper are consistent with three important ideas in urban and regional economics. Confirming previous research by Blanchard and Katz, the economic adjustment to the negative industrial shocks did not entail higher unemployment, even in the short run.1 In fact, most of the local adjustment to a rapidly shrinking local labor market seems to have occurred through the out-migration of individuals or reduced in-migration of prospective workers. These findings are important, given the relatively large magnitude of the shock under study: U.S. local labor markets respond quickly in a flexible manner, reflecting high mobility and, possibly, relatively low reliance on long-term unemployment insurance schemes.
The study also emphasizes how important "lifestyle" amenities have become to explain population growth. Even in a context where population levels seem to have been driven by production-side locational advantages in the industrial sector, amenities may have been important to retain workers locally. Furthermore, the results are a reminder that amenities are also endogenous to population levels and growth, probably due to agglomeration, scale, and scope economies in the nontradable sectors. This is important because models of city growth that are based on amenities (like some of my previous work [End Page 90] with Edward Glaeser and Jed Kolko) should take into account the endogeneity of some of the urban amenities.2
Finally, the results also point to the importance of regionwide agglomeration economies: proximity to urban areas helped some of these counties to cushion the impact of the shock. Externalities (such as knowledge and educational spillovers) and the advantage of a "diversified" industrial composition may manifest themselves at the broader regional level.
However, we cannot possibly learn about other important issues pertaining to local economic development and urban economics using the methodology that Feyrer, Sacerdote, and Stern propose.
What is the economic shock? The very nature of the economic shock that the authors study is not clear. The counties that experienced substantial losses in employment in the steel and auto industries during the years 1977–82 were probably very different from other counties in the United States. We do not know the "counterfactual" evolution of employment in these counties, and therefore the "shock" dummy variable may simply be capturing the impact of other characteristics of these cities that predict decline more generally. Even if we were confident that the "shock dummy" is not simply capturing the impact of other omitted variables in the relevant period, but stands for a real economic shock, it is not clear that the response of the counties under study is representative. Similar economic shocks that are more general in scope could have very different impacts on more diversified local economies. This is, of course, not a problem if we want to focus on the history of the "shock" counties, as the authors define them, but it is more problematic if we want to learn about how local economies respond to negative economic shocks more generally.
Is the shock exogenous? The way in which the authors define the "shock" counties is problematic due to potential reverse causality: plants that were located in worse environments may have had more of an incentive to shut down. Therefore, selection into the "shock" (as measured by the number of jobs lost) may be a symptom more than a...
