What this covers

David Autor of MIT joins Russ Roberts to examine the effects of U.S. trade with China on workers and communities, drawing on Autor's research into labor-market outcomes in the wake of rapid import competition. The conversation centers on a paradox: while trade with China increased aggregate American wealth, the shock was exceptional in its scale, speed, and geographic concentration. Rather than the smooth reallocation classical economics predicted, displaced manufacturing workers in affected regions fell into long-term non-employment, disability programs, and early retirement—outcomes that upended conventional assumptions about market adjustment.

Autor traces the empirical argument through several layers. China's share of world manufacturing exports rose from roughly 2% in 1990 to 17% by 2012, flooding global markets with labor-intensive goods at a pace and volume no prior trading partner had achieved. His research isolates this effect by mapping Chinese import categories to U.S. counties and instrumenting Chinese export growth using simultaneous Chinese sales to other wealthy nations—a methodological move to separate China's falling costs from U.S. demand shifts. The findings reveal that total employment decline in affected areas ran about a third larger than manufacturing job loss alone, a local multiplier effect where lower household incomes suppressed spending across restaurants, retail, and services. Rather than moving to new sectors, workers often cycled through successive labor-intensive industries—textiles to leather to furniture—each subsequently undercut by Chinese competition. The conversation also addresses where the adjustment flowed: not mainly through transfer payments like unemployment insurance, but predominantly through Disability, early retirement, Medicare, and Medicaid—programs that shield income but lack incentives to return people to work.

Sharpest takeaway

Autor argues that while trade with China raised aggregate U.S. welfare, the 'China shock' was uniquely large, rapid, and geographically concentrated, producing severe and unexpectedly persistent harm to low-skilled workers and communities that conventional economic theory had assumed would adjust smoothly.

  • China's unprecedented size and speed of development flooded global markets with labor-intensive goods that directly competed with low-skilled U.S. manufacturing
  • Displaced workers in affected communities did not reallocate to other sectors or migrate, instead falling into long-term non-employment and disability programs
  • The gains from trade go to consumers and other sectors while concentrated losses fall on specific workers and regions who lack good outside options

The argument · threads8 threads · 38 claims
0.85

Displaced manufacturing workers suffered durable earnings losses, did not relocate geographically, and fell into benefit programs rather than other jobs.

6 pointscentrality 5/5
  • The United States has seen a long-run decline in demand for less-educated workers since 1979-1980, so the bottom has been dropping out for non-college workers for decades; the manufacturing workers hit by the China shock were in a relatively privileged position of stable, well-paying jobs not requiring high education, and when those jobs disappeared they had no equally good outside option to fall back on.

    “the bottom has been dropping out for a long time for non-college workers in the U.S. labor market, really since 1979, 1980.”

  • Displaced workers who lose career jobs do very badly: the Jacobson, Lalonde, and Sullivan study of Pittsburgh steel plant closings in the early 1980s found displaced workers lost on average about 30% of earnings over the next 10 years, many were never re-employed, those re-employed earned significantly less, and even their mortality increased—demonstrating that displacement is durably costly.

    “they lost on average about 30% of earnings for the next 10 years. Many of them were not re-employed many years later; and those who were reemployed were reemployed at significantly lower wages.”

  • Displaced workers often move to another sector using a similar skillset (from textiles to leather to furniture to dolls) that subsequently also gets hit by Chinese competition, so the same workers are repeatedly unpleasantly surprised by the next shock because all these labor-intensive industries are moving to China.

    “people often move to another sector initially that then often gets hit. So, they say, 'Things aren't so good in textiles; I think I'll move to leather'... 'I think I'll move to doll manufacturing.'”

  • Much of the adjustment to trade shocks flowed onto benefit programs not designed as labor-market programs: while Unemployment Insurance and Trade Adjustment Assistance responded proportionally, the much larger transfers were Disability payments, early retirements, Medicare, Medicaid, TANF, and SNAP—programs that shield against income shocks but lack the incentive properties to return people to work, with Disability being effectively the opposite of employment.

    “The much bigger transfers spurred by these trade shocks are actually Disability payments... And early retirements. And Medicare, Medicaid, TANF... and SNAP”

  • Contrary to expectation, displaced workers did not move from manufacturing into other sectors at the local level, nor did they show the anticipated geographic mobility to growing places—they tended to stay put, partly because of localized benefits, family ties, uncertainty about opportunities, and the lack of compelling alternatives for low-educated, middle-aged workers even elsewhere.

    “we expected to see people moving out of manufacturing into non-manufacturing--everything else. We don't see that happening at the local level. Nor do we see a lot of mobility--geographic mobility”

  • Using longitudinal Social Security earnings data, the long-run impact on trade-exposed workers is bifurcated: the top-earning third (likely best-educated) adapt fine, moving to other employers or sectors without significant wage loss, while the bottom tercile lose employment and earnings they cannot recover—earning less whether they stay with the same employer, move within manufacturing, or leave manufacturing—because the high earners are simply more adaptable with better outside options.

    “the top third of them, the highest-earning third... they do fine... But if you sort of look at the bottom tercile of workers... those guys, they just seem to lose employment and earnings in a way that they are not able to recover from”

0.80

Trade theory predicts and justifies aggregate welfare gains, but always predicts strongly redistributive effects across individuals.

5 pointscentrality 5/5
  • Comparative advantage is not about absolute skill but about what you can produce that the market wants more cheaply than others; it is crucially dependent on prices and opportunities, so the thing you are 'best at' under one set of prices may not be your comparative advantage at another set of prices when trade changes prices and lowers the value of your skill.

    “the thing that you are "best at" under one set of prices--let's say that thing is--it may not be the thing you are best at, at another set of prices, as the value of that thing that you are good at, that you have absolute advantage in, may fall.”

  • Free trade allows countries to specialize in what they produce most cheaply or better than others, sell to the world, and buy what others produce better, which under almost any conventional economic theory improves aggregate welfare for all countries—and if it didn't, they would have no reason to trade.

    “the ability to focus on your comparative advantage makes all countries better off in aggregate. And if it didn't they would have no reason to trade.”

  • Trade increases the total size of the economic pie (GDP) in aggregate, but it has always been understood in theory to be strongly redistributive: the pie grows but some slices shrink enough that certain individuals end up worse off, because owners of a country's abundant factors gain while owners of its scarce factors lose.

    “trade both increases the size of the pie; but it's also always been understood in theory to be strongly redistributive.”

  • Paul Krugman argued in 1997 that the economists' case for free trade is essentially unilateral—a country serves its own interests by pursuing free trade regardless of what other countries do—which is correct for aggregate country welfare but not for the perspective of every individual within the country.

    “we actually quote from Paul Krugman in 1997, who writes: 'If economists ruled the world there would be no need for a World Trade Organization. The economists' case for free trade is essentially a unilateral case.'”

  • Before the China shock, the conventional wisdom held that trade's distributional harm to low-skilled workers 'could happen in theory but doesn't happen in practice,' treating it as a theoretical curiosity like a Giffen good—known on the blackboard but not seen in real life—because the Bretton Woods era simply did not make it visible.

    “we know trade is not Pareto-improving in theory'--meaning, Pareto-improving doesn't make everybody better off all at once. But in practice, that sort of doesn't show up so much. Well, it showed up. It showed up a lot.”

0.80

U.S. manufacturing employment fell primarily because the country stopped producing labor-intensive goods domestically when imports became cheaper.

5 pointscentrality 5/5
  • The decline of U.S. manufacturing from China was not about losing export markets but about the U.S. ceasing to produce labor-intensive goods (leather, textiles, plastics, toys, commodity furniture) for its own market when it became cheaper to import them, wiping out import-competing plants and causing a conservative loss of roughly 1.5 million manufacturing workers.

    “It was that it was producing them for itself... And all of a sudden, it found: Wow, it's actually cheaper just to buy it abroad than to make it here.”

  • U.S. manufacturing employment has been on the decline since the end of World War II, roughly seventy years, and much of that long contraction is due to technology rather than trade, while manufacturing output measured in value added has continued rising.

    “manufacturing in the United States has been on the decline since the end of WWII.”

  • About 25% of the decline in U.S. manufacturing employment between 1990 and 2007 was due to China's rising competitiveness, and about 45% between 2000 and 2007—conservative estimates capturing only the productivity-driven variation, amounting to roughly 1.5 million manufacturing jobs plus another half million from the one-third local multiplier.

    “about 25% of the decline in U.S. manufacturing employment between 1990 and 2007 was due to China's rising competitiveness; and about 45% between 2000, 2007.”

  • The large U.S. merchandise trade deficit (3-4% of GDP in the 2000s) loomed large because, instead of trading one type of manufacturing for another (which would have reallocated labor between manufacturing activities), the U.S. simply stopped doing a set of activities and got the goods on loan, producing a contractionary inward shift in labor demand for the affected workers.

    “it's like we took a set of activities that we were engaged in, that employed, millions of people actually to do them, and we just stopped doing them. And instead just got the goods on loan from another country.”

  • Autor's team isolated China's causal effect by mapping the ~400 categories of Chinese imports to the U.S. counties (grouped into 122 commuting zones) where those goods would have been produced, then instrumenting U.S. import growth with simultaneous surges of the same Chinese goods exported to 8 other wealthy countries—a component very likely driven by China's falling costs and rising productivity rather than U.S. demand changes—and relating that predicted import growth to local manufacturing employment changes.

    “what we do is we look at the same goods being imported to 8 other wealthy countries... when exports from China in a particular category surged to all those countries simultaneously... we use that to predict the import growth in the United States”

0.75

China's scale and speed of export growth was unprecedented and fundamentally different from prior trade patterns.

4 pointscentrality 5/5
  • Although economists understood theoretically that trade is not Pareto-improving and imposes disruptive losses on low-skilled workers, in the Bretton Woods era of rich-country-to-rich-country trade these costs appeared to be second-order and did not show up much in practice—a view that was correct at the time but was overturned by China's rise.

    “we know trade is not Pareto-improving in theory'--meaning, Pareto-improving doesn't make everybody better off all at once. But in practice, that sort of doesn't show up so much. Well, it showed up. It showed up a lot.”

  • China's share of world manufacturing exports in dollar terms rose from approximately 2% in 1990 to about 17% in 2012, and its share of world manufacturing value added rose from 5% to 25% over the same period—an unprecedented and remarkable shift.

    “China's share of world manufacturing exports, measured in dollar terms, went from approximately 2% in 1990 to about 17% in 2012.”

  • The China shock is distinctive not because trade theory changed but because of China's unique scale and speed: as the world's largest country with roughly 150 million surplus rural laborers migrating to export zones, it had latent capacity that could convert rapidly like a rock poised atop a mountain—a scale no smaller country like Vietnam could match because it would have run out of labor and space before supplying much of world demand.

    “China was like the proverbial rock poised atop a spiring mountain, waiting to come down. It just--all that potential energy can convert into kinetic energy in very short order.”

  • China's accession to the WTO in 2001 caused a dramatic inflection in U.S.-China trade even though it did not nominally change tariffs (China already had Most Favored Nation status); the best interpretation, per Peter Schott and coauthors, is that it resolved policy uncertainty, making firms confident that property rights and market access were secure enough to move production overseas permanently.

    “the best interpretation of what actually happened goes to Peter Schott at Yale and his coauthors--is that all of a sudden the uncertainty was resolved.”

0.61

Trade shocks created multiplier effects where manufacturing losses contracted entire local economies and labor markets beyond the initial shock.

4 pointscentrality 4/5
  • China's purchase of U.S. bonds allowed the United States to borrow at very cheap, near-zero interest rates—an opportunity the country may or may not have used well.

    “it allowed for very, very cheap interest rates in the United States. You know--we could borrow for essentially nothing.”

  • The surprising finding is that the net decline in employment across an affected commuting zone is actually larger—about a third again larger—than the decline in manufacturing alone, indicating a general slowdown ('Detroit effect') where lost manufacturing reduces household incomes, spending on restaurants, movies and consumer goods, pushing people out of the labor force entirely rather than reallocating to other sectors.

    “the net decline in employment, community-zone wide, is actually larger than just the decline in manufacturing... about a third again larger.”

  • Every $1000 of increased import penetration leads to approximately a $58 increase in transfer payments—about $4 in Unemployment, $8 in Disability, $10 in Retirement, $15 in other federal income assistance, and $18 in government medical benefits—which is only about 6% but not small, and these transfers carry efficiency costs from the taxes needed to fund them plus the cost of involuntary unemployment itself.

    “every $1000 of increased import penetration leads to approximately--this is kind of interesting--about a $58 increase in transfer payments.”

  • The China episode differs partly because it was so big and rapid: displaced workers compete with each other for new jobs, so when many are laid off simultaneously and geographically concentrated (as manufacturing always is), there is a multiplier effect on the local economy, evidenced by Austria's early-retirement buyout of older steelworkers immediately raising employment for slightly younger workers competing for the same jobs.

    “the rate of change matters because workers compete with each other when they are unemployed for new jobs.”

0.43

Technology, not trade, was the dominant force reducing demand for low-educated workers over the long run since 1979.

3 pointscentrality 3/5
  • Of the two forces hurting low-educated men, technology is the more important one—it created a lot of redundancy as there is far less dexterous physical labor to do in manufacturing and warehousing—while trade was not a big deal for low-educated male unemployment in the 1980s except through the dollar shock.

    “if I say which of the two forces is the most important, it is the former--it is technology.”

  • The decline in work among less-educated men is driven less by improved outside options of non-work and more by the worsening of available job options: earnings for non-college workers fell 20-30% since 1980, jobs have less security and fewer health benefits, and many are personal-services work less psychically appealing—especially to men—rather than tangible work.

    “I'm not sure the incentive, the outside option of being a non-worker, has been really improved in the time such that people don't have an incentive to work. I think it's more that the set of options that are available to them have really gotten worse.”

  • The ironic paradox is that even though low-educated workers are only about a third as large a share of the population in 2015 as in 1980, demand for them has shifted inward even faster, so their wages have fallen and employers view them as even less scarce despite their reduced supply.

    “even though there's only a third as many as a share of population in 2015 as there were in 1980, seemingly demand has moved inward even faster such that their wages have fallen, such that employers view them as even less scarce.”

0.43

Policymakers should phase adjustment to trade shocks gradually while simultaneously investing in worker skills and social insurance infrastructure.

3 pointscentrality 3/5
  • Helping workers requires doing two things simultaneously: investing in people to create opportunities (K-12 education, better vocational training systems like Germany, Austria, and Sweden, plus infrastructure to enable more manufacturing) and providing social insurance because not everyone will win—and the speed of adjustment matters since people have finite careers, so even if barriers should fall, a 25% tariff might be phased out over 10 years rather than overnight.

    “There's always going to have to be two things you do simultaneously. One is going to be investing in people to create opportunities... And then of course we are also going to have to have social insurance programs”

  • The real point of the research is not net job losses—the aggregate gains from trade are very likely positive—but the degree of concentrated, devastating loss to specific subsets of people and specific regions that went from robust manufacturing centers to blighted areas, with those workers unable to find good alternatives.

    “it's not about the net job losses. It's really about the degree of concentrated loss.”

  • The most trenchant criticism of the research is that it measures concentrated relative losses well but cannot equally measure the diffuse gains (e.g. jobs indirectly created in Silicon Valley), so the loss estimates may be conservative relative effects mislabeled as absolute; Autor concedes this but argues that even accounting for the missing gains, the episode was still strongly redistributive because the gains accrue to entirely different people.

    “you are measuring kind of a relative effect but you are calling it an absolute effect... It's that they might miss other important margins. And I'm happy to concede that point.”

0.24

China's rise lifted hundreds of millions out of poverty and spread wealth to developing economies through commodity demand.

2 pointscentrality 2/5
  • China's rise has been great for China, bringing hundreds of millions of people out of poverty faster than any time in world history, and has spread wealth to the developing world through China's demand for commodities in Africa, Central America, and Brazil—so the episode is not all cost.

    “It's brought hundreds of millions of people out of poverty. The fastest anyone has moved out of poverty any time in world history.”

  • Manufacturing is growing again in the United States but is much less labor intensive; the U.S. has relatively low wages, low energy prices, and pretty good infrastructure, so there are ways to invest that increase manufacturing opportunities.

    “manufacturing is growing in the United States again. It's much less labor intensive. But we do have relatively low wages; low energy prices; and pretty good infrastructure”