Working Paper: NBER ID: w31948
Authors: Luisa Gagliardi; Enrico Moretti; Michel Serafinelli
Abstract: We investigate the employment consequences of deindustrialization for 1,993 cities in France, Germany, Great Britain, Italy, Japan, and the United States. In all six countries we find a strong negative relationship between a city's share of manufacturing employment in the year of its country’s manufacturing peak and the subsequent change in total employment, reflecting the fact that cities where manufacturing was initially more important experienced larger negative labor demand shocks. But in a significant number of cases, total employment fully recovered and even exceeded initial levels, despite the loss of manufacturing jobs. Overall, 34% of former manufacturing hubs--defined as cities with an initial manufacturing employment share in the top tercile--experienced employment growth faster than their country's mean, suggesting that a surprisingly large number of cities was able to adapt to the negative shock caused by deindustrialization. The U.S. has the lowest share, indicating that the U.S. Rust Belt communities have fared relatively worse compared to their peers in the other countries. We then seek to understand why some former manufacturing hubs recovered while others didn't. We find that deindustrialization had different effects on local employment depending on the initial share of college-educated workers in the labor force. While in the two decades before the manufacturing peak, cities with a high college share experienced a rate of employment growth similar to those with a low college share, in the decades after the manufacturing peak, the employment trends diverged: cities with a high college share experienced significantly faster employment growth. The divergence grows over time at an accelerating rate. Using an instrumental variable based on the driving distance to historical colleges and universities, we estimate that a one standard deviation increase in local college share results in a rate of employment growth per decade that is 9.1 percentage points higher. This effect is in part explained by faster growth in human capital-intensive services, which more than offsets the loss of manufacturing jobs.
Keywords: No keywords provided
JEL Codes: J0
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| initial manufacturing employment share (L69) | subsequent change in total employment (J63) |
| initial share of college graduates (I23) | employment growth per decade (J21) |
| initial manufacturing employment share (L69) | local college share (I23) |
| strong local human capital bases (J24) | mitigate impacts of manufacturing decline (O14) |
| local college share (I23) | employment growth per decade (J21) |