How to Fortify Your Town Against Shocks

Deindustrialization is not destiny

San Jose, California.

Austin, Texas.

Raleigh, North Carolina.

Portland, Maine.

All four cities were heavily exposed to trade with China in the 2000s, and yet all are now among the most economically thriving parts of the country — which makes them exceptions to a wider trend.

The key finding of the “China Shock” research is that the cities and towns most in competition with rising Chinese imports not only lost manufacturing jobs but also suffered a variety of trenchant economic problems. Poverty rates, drug and alcohol abuse, and transfers from the government all climbed. Overall employment fell.

As explained by David Autor, David Dorn, and Gordon Hanson, the three economists who pioneered the China Shock literature and coined the term itself:

“Adjustment in local labor markets is remarkably slow with wages and labor-force participation rates remaining depressed and unemployment rates remaining elevated for at least a full decade after the China trade shock commences.”¹

Yet some towns and cities have not just rebounded but also found a way to grow vigorously. Why were they resilient to the economic pressures of deindustrialization, while so many other places were not?

An excuse for protectionism?

One theory is that resilience comes from shielding domestic manufacturers from competition. If manufacturing job loss from global competition is so unavoidably damaging to these communities, then we can help them simply by protecting domestic manufacturers from free trade.

And it is worth noting that the Biden and Trump administrations actually shared this interpretation of the China Shock research. Some of the tariffs enacted by Trump in his first term were kept in place by Biden, with Trump now pushing (much, much) further in his second term. Both administrations have cited the China Shock research in defense of tariffs.

What the research actually shows, however, is that economic resilience isn’t the result of preventing manufacturing job loss. The places that have managed to successfully recover from the China Shock — including our notable examples of San Jose, Austin, Raleigh, and Portland, and others as well — all lost a third or more of their manufacturing jobs from 2000 to 2010.

Nor have their impressive economic recoveries been the result of bringing back all their manufacturing jobs. The number of manufacturing jobs in Austin, for instance, still remains 14 percent below its level in the year 2000. Austin’s overall employment, meanwhile, has doubled in the same period. Its rebound from the China Shock clearly relied on jobs created in non-manufacturing sectors of the economy.

What resilient places have in common

A better explanation for the resilience of certain towns and cities can be found in another paper from David Autor, David Dorn, and Gordon Hanson. Published in 2021, a decade after their first work on the topic, the paper looked at how the places hardest hit by trade with China were doing in the long-run.

They found that an important factor is what economists call “human capital,” a broad term meant to capture a population’s levels of education, skills, ability, and talent. Specifically, their research showed that China-Shocked places with a high share of college-educated workers were much more resilient to the problems that arise from deindustrialization.

Put another way, the costs of failing to adapt to Chinese import competition were concentrated in the places with low levels of education.

Why does having an educated population matter? Again it is clear that this is not a story about preventing manufacturing job loss. High-education places did lose smaller shares of manufacturing jobs than low-education places. But they did lose manufacturing jobs. What really set apart the high-education places is that they did not experience overall job loss.

What human capital delivers is job growth outside of manufacturing.

These results have been confirmed by economists Nicholas Bloom, Kyle Handley, Andre Kurmann, and Philip A Luck. Using confidential Census data on the universe of businesses and workers, they found that while better-educated places did lose manufacturing jobs, the losses were offset by gains in services employment.

The granularity of the data also allowed them to identify instances when a manufacturing business switched to become a service business. In these cases, a plant did not close; it simply stopped manufacturing goods but kept doing Research & Development or other service-oriented work. Think of a computer company that no longer assembles computers but still does R&D, marketing, and sales. This flexibility turns out to be an important component of resilience in more-educated areas.

In places with lower rates of education, by contrast, there were few of these offsetting gains in services. Total employment fell, and that decline brought with it the severe spillover effects discovered in so much of the China Shock research.

The centrality of human capital to a place’s resilience to the China Shock has been found in other research as well. And it is visible in our leading examples of China Shock resilience — San Jose, Austin, Raleigh, Portland — all of which are in the nation’s top 15 percent of commuting zones by education levels.²

The evidence has consistently shown that a dearth of human capital leaves places fragile to economic shocks, while an abundance of human capital offers fortification against them.

And as for adjusting to trade with China, places with higher levels of education behaved as economists believed they would, with expanding sectors absorbing workers who lost jobs in contracting sectors. Places with lower education levels could not make the same adjustment, with terrible consequences for their residents.

Zooming out

Identifying the right lessons from the China Shock research is important, in part because the wrong lessons are so widely promoted, with protectionists leaning so heavily on it to argue that trade is harmful. Yet Chinese imports are hardly the only cause of deindustrialization, with technology and the natural emergence of other sectors also playing a role.

The decline of manufacturing employment is a global phenomenon among developed countries. It can therefore be useful to study other examples of deindustrialization, both in the developed world and in American history.

The economists Luisa Gagliardi, Enrico Moretti, and Michel Serafinelli recently looked at the fates of historical manufacturing hubs across the United States and a handful of other developed countries.³ They confirm, unsurprisingly, that having a high share of manufacturing jobs in the past is correlated with worse economic outcomes down the road. The pressure of deindustrialization is real.

But they also found evidence that deindustrialization is not destiny. Of cities throughout the world that were once manufacturing hubs and then suffered a period of deindustrialization, 34 percent were able to subsequently outperform the overall labor market in their respective countries⁴ — “suggesting that a surprisingly large number of cities was able to adapt to the negative shock caused by deindustrialization,” write the authors.

In the United States, however, only 17 percent of these former manufacturing hubs were able to outperform the American labor market after they deindustrialized.

Why is it that such a high share of American cities could not adapt? What can we learn about the cities that did?

Again we see a central role for education. The top quarter of these manufacturing hubs with the best-educated populations suffered no negative effects on their total employment. The damages were most substantial for the lowest-educated places, as shown in the chart from the paper by Gagliardi, Moretti, and Serafinelli:

Correlation or causation?

Do higher education levels actually cause resilience to deindustrialization, or could there be some other factor that causes both strong resilience and higher education levels.

The local presence of a high-tech industry, for example, could attract more-educated workers to a local area and at the same time reduce the economy’s dependence on manufacturing in the long-run. In this case it would be the industry itself that is responsible for both the strong resilience and the higher education levels.

But the same paper from Gagliaridi, Moretti, and Serafinelli offers a few reasons to believe that education does indeed cause the stronger resilience. They show that while cities with higher education rates did better after manufacturing employment peaked, they did not outperform before then — which suggests that the economic boost provided by education came after the deindustrialization shock, not prior to it.

Their paper also proxies for education level by using the long-standing presence of a university. Importantly, their data allows them to focus only on universities that were built decades before manufacturing peaked, which would help to rule out reverse causality. The results were consistent with the idea that more education causes resilience.

An experiment from history

For even stronger evidence that human capital causes resilience rather than simply coinciding with it, the best (and most fascinating) analysis comes from a natural experiment studied by economists Greg Howard, Russell Weinstein, and Yuhao Yang.

Their work begins with an observation: In the early 20th century, the social reform movement led to a sharp rise in the number of both insane asylums and schools for teachers.

Importantly, these two types of institutions had a lot in common. The locations that policymakers considered for them were similar, as was the process by which these locations were actually chosen. Both the asylums and the teaching schools required large campuses, were often picked by state legislatures and commissions, and were seen as desirable or “prized” by the receiving communities at the time.

As a result of these similarities, counties that received asylums looked a lot like counties that received teaching schools. They looked similar in 1820, before the institutions were put in place, and they still looked similar a century later in 1920.

By 1980, however, there had been a divergence. The asylums had closed or turned into modestly sized health facilities, while the national rise of higher education had led many teaching schools to grow into very large public universities. As a consequence, the counties that initially received teaching schools enjoyed higher rates of education on average.⁵

The research shows that the presence of a regional university provided “full resilience” to the negative effects on total employment from losing manufacturing jobs. The economists also find nearly full resilience against population loss, average earnings loss, and the growth in government transfers.

Further consistent with the China Shock research, the paper finds that education helps not by reducing manufacturing job loss but by helping a city adapt to its loss.

What a better-educated population means

So it is beyond dispute that human capital is a key factor in explaining why some places are more resilient than others. Exactly why this is the case is more uncertain.

In places like San Jose and Austin, it seems likely that the impact of education on innovation, the sharing of ideas, and entrepreneurship plays an important role. And indeed other research, for example a 2023 paper from Michael J Andrews, finds that universities lead to higher local rates of patenting.

Yet the large public universities that began as teaching schools in the natural experiment did not become research universities, indicating other possible channels besides innovation. They find that the university's local spending helps, and also that simply having a more educated population matters.

Another likely part of the story is adaptation. As economists Ed Glaeser and Albert Saiz hypothesize, “Cities are constantly reinventing themselves — moving from one field of specialization to another.” Skilled workers are an important ingredient for speedy reinvention. Highly educated workers can themselves change occupations and develop new skills, but they can also power broader adaptation through their entrepreneurship.

Places that suffer the loss of a large employer and industry need individuals capable of taking the capital and workers that are disemployed and finding the best uses for them.

Policy Lessons

The evidence plainly shows that policymakers should focus on human capital as an essential ingredient to resilience. This does not mean that everyone in a given place should “learn to code.” Resilience is not about sending disemployed manufacturing workers back to college. It is about having skilled individuals around to help put everyone back to work, including those who haven’t learned to code.

Three policy lessons flow naturally from this conclusion.

First, do no harm when it comes to universities.

We are entering a period in which universities are already under stress due to demographically driven enrollment declines. Policies that further threaten university closures will be harming an important source of resilience.

Cutting off foreign-born students, for example, would push many universities into risk of failing. This does not mean propping up universities at all costs or preventing needed reforms, but it does mean understanding the costs to resilience in cities and towns across the country when they do fail.

Second, make universities even more useful to local economies.

There is a wide variance in college and degree quality. “Master’s degrees to nowhere” are real. Policies should push universities and students to focus more on degrees that lead to gainful employment. The role of universities as sources of economic resilience for their local communities only bolsters the case for this goal. Just which policies would best achieve it is a matter for future research.

Third, we need policies that will welcome human capital from abroad into the places that need it most.

Heartland Visas are one important policy that can bring skilled immigrants to deindustrializing places. The era of remote work makes this more possible than ever, as immigrants are better able to work anywhere throughout the country and not just in the traditional employment strongholds.

In short, policy should focus on increasing human capital in struggling places to help them find better futures, not to help them return to the past.

Protectionists must recognize that we cannot always protect manufacturers and the communities that depend on them from the changing tides of the global economy. Rather than trying to prevent economic shocks, a futile and sisyphean task, we would do far better to learn from the places that have become resilient to them.

1. David H. Autor & David Dorn & Gordon H. Hanson, 2016. "The China Shock: Learning from Labor Market Adjustment to Large Changes in Trade," Annual Review of Economics, vol 8(1).

2. China shock resilient commuting zones are identified using Autor et al (2016). Commuting zones are ranked by education levels using the ACS 2023 5-year's data on the share of the population with a bachelor's degree or higher.

3. The geographical unit of analysis for hubs varies by country, but in general they are based on residential and commuting patterns. In the U.S. they use metropolitan statistical areas.

4. Specifically, the paper looked at cities that were manufacturing hubs — defined as “cities with an initial manufacturing employment share in the top tercile” — when manufacturing employment peaked in their home country. The paper found that “cities where manufacturing was initially more important experienced larger negative labor demand shocks,” and then further examined the cities that recovered better from these shocks, which is what I am concerned with in this post.

5. In economics terms, the asylum counties are the control group, while the teaching countries represent places with an exogenous increase in education from having received a teaching school in the prior century. Because the two types of counties had once been highly similar, and because the selection for asylums versus teaching schools had been close to random, the divergence between them can credibly be accounted for by that selection.