自由贸易总是好的吗？
Comparative advantage says yes. The workers who lost their jobs say it’s more complicated.
The case for free trade
“We are right now taking in $billions in Tariffs. MAKE AMERICA RICH AGAIN. I am a Tariff Man.”
— @realDonaldTrump, December 2018
Shared millions of times. The claim: tariffs make America richer. Most economists disagree. But is it that simple?
To evaluate that claim, you need the most powerful result in all of economics: comparative advantage. David Ricardo showed in 1817 that trade makes both countries better off in aggregate, even when one country is better at producing everything.
The insight is counterintuitive. Suppose the United States can produce both wheat and textiles more efficiently than Bangladesh. Shouldn’t the US produce both? No. What matters is not absolute cost but opportunity cost. If producing one more ton of wheat costs the US 2 units of textiles forgone, but costs Bangladesh only 0.5 units, then Bangladesh has a comparative advantage in wheat — even if it’s less productive at both goods.
The result: both countries are better off if each specializes in the good where its opportunity cost is lower and trades for the rest. Under competitive markets, full employment, and no externalities, free trade maximizes total surplus.
So what happens when a country opens to trade? The domestic price converges toward the world price. If the world price is below the domestic price, the country imports. Consumer surplus rises (cheaper goods), producer surplus falls (domestic firms face competition), but the net effect is positive. The gains to consumers exceed the losses to producers.
Now: what about those tariffs the “Tariff Man” is so proud of? A tariff is a tax on imports. It raises the domestic price above the world price, reduces imports, and protects domestic producers. But it creates deadweight loss — the same triangle of wasted potential we see with any price distortion.
If the world price is $P_w$ and the tariff raises the domestic price to $P_w + t$, the deadweight loss consists of two triangles:
$$DWL = \frac{1}{2}t \cdot \Delta Q_{\text{consumption}} + \frac{1}{2}t \cdot \Delta Q_{\text{production}}$$Every dollar of protection comes at a cost that no one receives.
A tariff is a tax that makes imports more expensive. Consumers pay more. Some of that extra money goes to domestic producers, some to the government — but some just evaporates. It’s pure waste: the cost of producing things at home that could have been made more cheaply abroad, plus the purchases consumers would have made but can’t at the higher price.
The standard model is unambiguous: tariffs reduce total welfare. The “Tariff Man” is taxing his own citizens and calling it income. Want the full derivation with the interactive graph? Ch 2 §2.6.
“We are right now taking in $billions in Tariffs. MAKE AMERICA RICH AGAIN. I am a Tariff Man.”
— @realDonaldTrump, December 2018
“Do tariffs actually work?”
The “Tariff Man” says tariffs make America richer. Economists say they’re a tax on American consumers. The 2018 trade war provides the first major test case in modern data.
The bedrock and the blind spot
“Comparative advantage is the best example of an economic principle that is undeniably true yet not obvious to intelligent people.”
— Paul Samuelson, quoted by Douglas Irwin in Free Trade Under Fire
Samuelson is making the strongest possible case: comparative advantage isn’t a conjecture, it’s a mathematical theorem. Under the standard assumptions, free trade maximizes total surplus. The result is robust to extensions — multiple goods, multiple countries, varying factor endowments. It survives every generalization Ricardo could not have imagined. This is the bedrock.
“We estimate that the full incidence of the tariffs fell on domestic consumers and importers. Tariffs, as of 2018, amounted to one of the largest tax increases in decades.”
— Mary Amiti, Stephen Redding & David Weinstein, Journal of Economic Perspectives, 2019
This is the first rigorous empirical study of the 2018 tariffs. Amiti, Redding, and Weinstein used customs microdata to show near-complete pass-through: foreign export prices didn’t fall, so the tariff was paid entirely by American buyers. The “Tariff Man” wasn’t taxing China. He was taxing Iowa farmers, Texas manufacturers, and every consumer who bought imported goods. The theory predicted exactly this.
Where this leaves us
The case for free trade is strong in the aggregate. Comparative advantage is real, the gains from trade are measurable, and tariffs genuinely destroy welfare. The 2018 tariffs confirmed it with modern data. But notice the caveat: “in the aggregate.” Total surplus goes up. But the gains are spread thinly across millions of consumers — each saving a few cents — while the losses are concentrated on specific workers, specific factories, specific towns. The model says the winners could compensate the losers. In practice, they never do.
The standard trade model assumes displaced workers smoothly transition to new jobs. A factory closes in Ohio; the workers become software engineers in California. How long does that adjustment take? The answer, when economists finally measured it, changed the entire debate.
The China shock
“Adjustment in local labor markets is remarkably slow, with wages and labor-force participation rates remaining depressed and unemployment rates remaining elevated for a full decade or more after a China trade shock commences.”
— David Autor, David Dorn & Gordon Hanson, Annual Review of Economics, 2016
The “China shock” paper changed the debate. Free trade has winners — but the losers don’t just “adjust.”
The intro model from Stage 1 treats labor as perfectly mobile between sectors. A displaced factory worker costlessly becomes a barista, a coder, or a nurse. Total surplus rises; everyone lands on their feet. Autor, Dorn, and Hanson tested that assumption — and demolished it.
What happened. Between 1999 and 2011, China’s share of world manufacturing exports surged from 5% to 15%. US communities most exposed to Chinese import competition experienced persistent job losses, depressed wages, increased disability claims, and higher mortality. These weren’t temporary adjustment costs that faded after a few years — they were still visible a decade later.
The scale. Autor, Dorn, and Hanson (2013) estimated that the China trade shock eliminated 2.0–2.4 million US manufacturing jobs between 1999 and 2011. The affected communities didn’t recover by moving into service jobs or tech. They experienced secular decline — falling labor force participation, rising opioid use, and political radicalization. The rust belt didn’t “adjust.” It collapsed.
Why this matters for the model. Stage 1 said trade creates net gains and losers can be compensated. Two things went wrong. First, the adjustment was far slower and more painful than anyone expected — labor markets are geographically sticky and skill mismatches are real. Second, the compensation never happened. Trade Adjustment Assistance programs in the US were chronically underfunded, poorly targeted, and largely ineffective.
Paul Krugman’s new trade theory adds another dimension. His 1980 work showed that trade generates gains beyond comparative advantage when markets are imperfectly competitive. With economies of scale, each country specializes in different varieties of the same good. Consumers get more variety; firms serve larger markets and move down their average cost curves. The gains from trade here come from variety and scale, not just Ricardian specialization. But the same logic means that when trade displaces an industry with scale economies, you don’t just lose jobs — you lose the accumulated learning and the cluster effects that made the industry possible.
“The ‘China shock’ of the 2000s destroyed 2.4 million American jobs. The affected communities still haven’t recovered.”
— David Autor, David Dorn & Gordon Hanson, American Economic Review, 2013
Did China kill American manufacturing?
Economists spent decades saying trade adjustment would be smooth. Autor, Dorn and Hanson showed it wasn’t. Import competition from China hit specific communities with devastating, persistent effects — job loss, wage decline, opioid addiction, political radicalization.
Aggregate gains vs. concentrated losses
“The China shock operates through a narrow set of manufacturing industries concentrated in specific communities. The majority of American workers are in sectors unaffected by import competition and benefit from cheaper goods. The aggregate gains from trade with China are large and positive.”
— Douglas Irwin, Free Trade Under Fire, 5th ed., 2020
Irwin is the leading historian of trade policy, and he’s making the nuanced version of the pro-trade argument. He doesn’t deny the China shock — the data is too strong. Instead, he argues that the aggregate gains still exceed the losses and the right response is better domestic policy (retraining, relocation assistance, wage insurance), not trade restriction. The question is whether that response will ever actually materialize.
“Regions more exposed to Chinese imports saw sharp declines in manufacturing employment, lower wages, and higher uptake of disability, retirement, and death benefits. The labor market costs of trade are real, persistent, and geographically concentrated.”
— David Autor, David Dorn & Gordon Hanson, American Economic Review, 2013
This is the paper that broke the consensus. Before Autor, Dorn, and Hanson, the profession treated trade adjustment as a short-run friction that theory could safely ignore. After it, the distributional costs of trade became impossible to dismiss. The paper didn’t say trade was bad — it said the assumption of costless adjustment was catastrophically wrong, and that real people bore the consequences of that assumption for decades.
Where this leaves us
The China shock didn’t disprove comparative advantage. Trade with China genuinely made the average American better off through cheaper goods. But averages lie. The costs were concentrated on communities that had no political voice and received no compensation. The profession’s consensus shifted: distributional effects are not a footnote — they are the central policy challenge of trade. The theory said “trade creates gains and losers can be compensated.” The practice said: “the losers were never compensated, and they voted accordingly.”
But the China shock story is incomplete without the macro dimension. The US didn’t just import Chinese goods — it imported Chinese savings. The trade deficit everyone panics about has a mirror image that changes the entire picture.
Trade deficits and capital flows
“America doesn’t have a trade deficit because foreigners are beating us. We have a trade deficit because foreigners want to invest in America.”
— A common reframing in international economics
The trade deficit looks scary until you understand capital flows. Every dollar that leaves to buy foreign goods comes back as foreign investment in American assets.
The micro-level trade debate — comparative advantage, tariffs, the China shock — misses half the picture. Every trade flow has a financial mirror image. Here is the fundamental identity of international economics:
The current account measures the flow of goods, services, and income across borders (roughly, net exports). The capital account measures the flow of financial assets. If the US imports \$500 billion more in goods than it exports, that “trade deficit” means foreigners are investing \$500 billion in US assets — Treasury bonds, real estate, equities. This isn’t a theory. It’s an accounting identity, true by definition.
What the trade deficit actually is. When politicians wave the trade deficit around as proof that foreigners are “winning,” they’re looking at one side of the ledger. The other side says: foreigners are pouring capital into America because they view it as a safe, profitable place to invest. The US runs a trade deficit because it runs a capital surplus — because the world wants to hold dollar assets. You can’t fix one without eliminating the other.
Exchange rates and adjustment. In a floating exchange rate regime, trade imbalances tend to self-correct. If a country runs persistent trade deficits, its currency depreciates, making exports cheaper and imports more expensive. But China managed its exchange rate for decades, keeping the yuan undervalued. This made Chinese exports artificially cheap — a de facto subsidy financed by Chinese consumers. That’s not free trade. It’s trade distorted by exchange rate policy.
The impossible trinity. A country cannot simultaneously maintain (1) a fixed exchange rate, (2) free capital flows, and (3) independent monetary policy. It can pick any two. China chose a managed rate and independent monetary policy, requiring capital controls. The US chose free capital flows and independent monetary policy, accepting a floating rate. These choices shape trade outcomes as much as tariff schedules do.
The Stolper-Samuelson theorem predicts that trade with a labor-abundant country hurts the scarce factor (unskilled labor) in the rich country. Formally, an increase in the relative price of the labor-intensive good raises the real wage and lowers the real return to capital. Trade with China did the reverse in the US: it lowered the relative price of labor-intensive manufactures, reducing real wages for unskilled workers while benefiting capital owners and skilled workers. This isn’t a market failure — it’s the model working exactly as predicted — but it means the gains from trade are distributed in a way that increases inequality.
Trade with a country that has lots of cheap labor pushes down wages for workers who compete with that labor, and pushes up returns for the people who own the capital and high-skill workers who complement it. The pie gets bigger, but the workers who were already struggling get a smaller slice. The model predicts this — it’s a feature, not a bug.
Global imbalances and the 2008 crisis. In the 2000s, the US ran massive current account deficits while China, Germany, Japan, and oil exporters ran surpluses. Ben Bernanke called it a “global saving glut” — excess savings flowing to the US, financing American consumption and the housing bubble. The 2008 financial crisis was partly a story of these imbalances unwinding. Capital inflows can finance productive investment — or they can finance bubbles.
“Trade wars are good, and easy to win.”
— Donald Trump, Twitter, March 2, 2018
Are trade wars easy to win?
The U.S.-China trade war imposed tariffs on hundreds of billions of dollars of goods. The theory says both sides lose. The politics says someone has to blink first. What actually happened?
Are trade deficits a problem?
“The current account deficit is a sign of strength, not weakness. It means the rest of the world wants to invest here. Countries that attract capital are the winners, not the losers.”
— Summary of the capital-flows perspective (Bernanke, Summers)
This is the optimistic macro view: the trade deficit reflects capital inflows, and capital flows to where returns are highest. America’s deficit reflects the world’s confidence in its economy. The problem: capital inflows also appreciate the dollar, making US exports less competitive and accelerating deindustrialization. Strength for the financial sector can mean weakness for the manufacturing sector. And if the capital finances bubbles rather than productive investment, the “strength” is illusory.
“China’s managed exchange rate amounted to a massive subsidy for its exporters, paid for by Chinese consumers who received less purchasing power for their labor. This was a deliberate strategy to build industrial capacity at the expense of trading partners.”
— Summary of the currency-manipulation critique (Bergsten, Krugman)
The currency-manipulation argument adds a critical nuance: the free-trade framework assumes market-determined exchange rates. When a major trading partner manages its currency to maintain undervaluation, the price signals that comparative advantage depends on are distorted. American manufacturers weren’t competing against Chinese productivity — they were competing against the People’s Bank of China. This isn’t a free-market outcome.
Where this leaves us
Trade deficits are neither inherently good nor bad. They reflect intertemporal trade — borrowing from abroad to consume or invest at home. A country can run deficits sustainably if the capital inflows finance productive assets. But persistent large imbalances can fund bubbles, appreciate the currency to the detriment of exporters, and create strategic vulnerabilities. The partial-equilibrium analysis of Stage 1 is dangerously incomplete — you need exchange rates, capital flows, and savings behavior to understand trade outcomes. The politician waving the trade deficit number is usually wrong about what it means. But the economist dismissing it as irrelevant is ignoring real structural effects.
We’ve now seen the micro case, the labor market case, and the macro case. But there’s one more dimension that transforms the entire debate. The most successful development stories of the last century — Japan, South Korea, Taiwan, China — all involved strategic trade policy, not pure free trade. Were they right?
Industrial policy and development
“I don’t know any economist who would say that Korea or Taiwan would have been better off with free trade from the start. They used industrial policy, and it worked.”
— Dani Rodrik, The Globalization Paradox, 2011
East Asia’s success is awkward for free traders. The countries that grew fastest did it by strategically violating the free-trade playbook.
In the 1960s, South Korea was poorer than Ghana. By 2000, it was a high-income country with world-leading firms in semiconductors, shipbuilding, and automobiles. Japan, Taiwan, and China followed similar paths. None of them did it with free trade. They all used targeted protection, export subsidies, managed exchange rates, and state-directed credit.
The infant industry argument. If production involves learning-by-doing — costs fall with cumulative output — then a new entrant faces a chicken-and-egg problem. It can’t compete with established foreign firms at current costs, but it would become competitive if it could accumulate enough experience. Temporary protection allows the infant firm to survive the learning phase.
where $C_0$ is the initial unit cost, $Q$ is cumulative output, and $\beta > 0$ is the learning rate. If the present value of future competitive profits exceeds the cost of the temporary protection, the intervention is welfare-improving.
A new industry is expensive at first — but every unit it produces teaches it to produce the next one more cheaply. If you let foreign competition kill it before it finishes learning, you never find out how cheap it could have become. Temporary protection buys time to learn. The question is whether the firm will actually learn fast enough to justify the cost.
The Brander-Spencer model. Under oligopoly, strategic subsidies can shift profits from foreign firms to domestic ones. Consider a global market with two firms competing in a Cournot oligopoly. If the domestic government subsidizes its firm, the firm becomes more aggressive, the foreign firm retreats, and profits shift homeward. If the profit shift exceeds the subsidy cost, the domestic country wins.
South Korea’s POSCO illustrates the infant industry argument at its best. When the state-owned steel company was established in 1968, the World Bank refused to fund it — Korea had no comparative advantage in steel. By the 1990s, POSCO was one of the most efficient steel producers in the world. The learning curve worked as theory predicted.
But for every POSCO, there are dozens of failures. Latin American import substitution in the 1960s–80s created sheltered, uncompetitive industries that never grew up. India’s “License Raj” produced bureaucratic stagnation. Africa’s state-led industrialization efforts largely failed. The East Asian successes are the exception, not the rule. The difference may be institutional quality, state capacity, and — crucially — export discipline. East Asian firms received protection but were expected to export. The discipline of global competition prevented the rent-seeking that killed infant industries everywhere else.
The 2020s revival. Industrial policy is back. The US CHIPS Act, the Inflation Reduction Act, and the EU Green Deal all involve targeted subsidies to strategic industries — exactly what Brander-Spencer contemplated, justified now by supply chain resilience, national security, and the climate transition.
“I don’t know any economist who would say that Korea or Taiwan would have been better off with free trade from the start.”
— Dani Rodrik, 2011
Should we protect strategic industries?
East Asia proves industrial policy can work. Latin America proves it usually doesn’t. The CHIPS Act bets that the US can tell the difference. The 2018 tariffs suggest otherwise.
Strategic engagement vs. free trade default
“The right model for development is not free trade or protectionism. It’s strategic engagement with the global economy: selectively open, with industrial policies calibrated to institutional capacity.”
— Dani Rodrik, The Globalization Paradox, 2011
Rodrik’s “industrial policy 2.0” is the most influential heterodox position in the trade debate. He doesn’t reject comparative advantage — he argues it’s incomplete. Countries can create comparative advantage through strategic investment, and the standard free-trade prescription ignores this possibility. The key qualification: institutional capacity is the binding constraint. Countries without strong, accountable institutions shouldn’t try industrial policy. Which, Rodrik admits, describes most of the developing world.
“The historical record is more favorable to free trade than the infant industry narrative suggests. Countries that grew fastest in the 19th century — including the US — did so during periods of falling trade barriers. The East Asian cases are real but they are the exception, not the rule.”
— Douglas Irwin, Free Trade Under Fire, 5th ed., 2020
Irwin is the leading trade historian, and his counterargument is data-driven. He shows that the US’s own 19th-century tariffs were more about revenue than industrial strategy, and that US growth accelerated as trade barriers fell. East Asia’s success may have been despite industrial policy, not because of it — the real drivers were high savings, human capital investment, and macroeconomic discipline. The selection bias is real: we remember POSCO, we forget the failures.
Where this leaves us
Pure free trade doctrine was too strong. The most successful development stories involved strategic engagement with global markets, not passive acceptance of comparative advantage. But strategic intervention works under demanding conditions that are uncommon. Most countries that tried industrial policy failed. The honest answer: free trade is the right default for most countries most of the time. Strategic intervention can work when institutions are strong, export discipline is enforced, and protection has a credible expiration date. The climate transition and supply-chain reshoring are creating a new trade regime that neither Ricardo nor Brander-Spencer anticipated — one where national security, environmental externalities, and geopolitical risk matter as much as efficiency.
Where this leaves us
We started with the “Tariff Man” claiming that tariffs make America richer. Four stages later, here’s what you now know:
- Comparative advantage is real (Stage 1). Trade based on comparative advantage increases total surplus, and tariffs genuinely destroy welfare. The 2018 tariffs confirmed it: American consumers paid the bill, not China. But “total surplus increases” is a weaker statement than it sounds when the losers are never compensated.
- The losers don’t just adjust (Stage 2). The China shock showed that displaced workers suffer persistent, concentrated, and devastating losses. The profession assumed frictionless adjustment for decades. The data said otherwise. This isn’t a footnote to trade theory — it’s the central policy challenge.
- Trade deficits are about capital, not competitiveness (Stage 3). Every trade deficit has a capital-account mirror image. The US deficit reflects global demand for dollar assets, not foreign countries “beating” America. But exchange rate manipulation, capital flow distortions, and global imbalances create real problems that the simple Ricardian framework doesn’t capture.
- Industrial policy can work — but usually doesn’t (Stage 4). East Asia proves strategic trade policy is possible. Latin America proves it’s usually a disaster. The conditions for success — strong institutions, export discipline, hard budget constraints — are demanding and rare. The climate transition is creating new rationales, but the historical failure rate should humble anyone proposing to pick winners.
The next time someone tells you “tariffs bring back jobs” or “free trade is always optimal,” you have the tools to evaluate both claims. The tariff advocate is ignoring deadweight loss, consumer costs, and the historical failure rate of protectionism. The free trade purist is ignoring distributional devastation, exchange rate distortions, and the East Asian development experience. The honest answer lives in the conditional: free trade is the right default, but it’s not a universal law. The conditions matter — market structure, exchange rates, institutional quality, and whether anyone is actually compensating the losers.