A country can be worse at making every single thing, worse at growing wheat, worse at sewing shirts, worse at building aircraft, and still get richer by trading with the country that beats it at everything. That sentence sounds like a trick, and for two centuries it has been the most counterintuitive true statement in economics. It is where trade theory begins, and the rest of the field is the long project of working out exactly when it holds, what it predicts about who exports what, and who inside each country wins and loses when the borders open.
You met the headline at intro level in section 2.6: the world price, the gains-from-trade triangle, the deadweight loss of a tariff. This chapter is the formal home of the apparatus. We rebuild comparative advantage from opportunity-cost ratios, generalize it to factor endowments, trace trade's effect on wages and rents through the magnification effect, and then layer on the models the classical theory could not produce: increasing returns and intra-industry trade, the gravity equation that predicts trade flows with eerie accuracy, and the firm-level micro that explains which producers actually export. The models accumulate. Each one answers a question the previous one could not, and none of them is overthrown by the next.
We follow one trade relationship the whole way through. Call the two economies Home and its partner; watch the same pair re-explained by six successive lenses. By the end the chapter turns to policy, the one door the theory leaves open for tariffs and why it usually slams, and to the empirical record the efficiency-focused thread under-weighted: the China shock, where the distributional prediction the models made in the abstract arrived as concentrated, lasting damage to particular American towns.
Prerequisites: intro comparative advantage and the world-price tariff diagram (section 2.6), consumer and producer surplus and deadweight loss (Chapter 3), monopolistic competition and the Cournot/Stackelberg games (Chapter 6). The gravity discussion connects to the econometrics of structural estimation (Chapter 10); the closing section borrows the balance-of-payments identity from open-economy macro (Chapter 17).
Home and its partner each make two goods, cloth and wine, using one input: labor. Suppose Home is more productive at both: it takes fewer worker-hours per bolt of cloth and fewer per cask of wine. The eighteenth-century mercantilist would conclude that Home should make everything and the partner nothing. David Hume had already undercut that view with his price-specie-flow mechanism, the argument that a country running a permanent trade surplus would import gold, raise its own prices, and price itself back out of its markets, so a country cannot simply out-produce the world forever. But the decisive move came from Ricardo, and it turns on the difference between two kinds of advantage.
Write $a_{LC}$ for the worker-hours Home needs per unit of cloth and $a_{LW}$ for the hours per unit of wine; the partner's requirements carry an asterisk, $a_{LC}^*$ and $a_{LW}^*$. Home has a comparative advantage in cloth when it gives up less wine per bolt of cloth than the partner does. The comparison is of ratios, not levels, which is exactly why absolute advantage drops out.
Eq. 22.1 is the comparative-advantage condition: Home's opportunity cost of cloth (in wine forgone) is lower than the partner's. Note that it can hold even when $a_{LC} > a_{LC}^*$ and $a_{LW} > a_{LW}^*$, with Home worse at both in absolute terms. Productivity levels cancel out of a ratio of ratios; only the relative slope of each country's frontier survives.
Pourquoi c’est important : Think of a surgeon who also happens to type faster than her assistant. She is better at both surgery and typing, yet she should still hand the typing to the assistant, since every hour she spends typing is an hour not spent in the operating room, where her edge is enormous. The assistant gives up almost nothing by typing instead of operating. That is comparative advantage: each party specializes where its sacrifice is smallest, and the country that is worse at everything is simply the assistant. It has the least to lose by doing the thing it is least-bad at, so the world hands it that job and everyone produces more.
With constant unit labor requirements, each country's production possibility frontier is a straight line whose slope is the opportunity cost. In autarky, with no trade, each country must also consume on its own frontier, and competition forces the relative price of cloth to equal its opportunity cost: $P_C/P_W = a_{LC}/a_{LW}$ at Home, and the partner's analog abroad. The two autarky price ratios differ precisely because the two slopes differ.
Open the border and a single world relative price emerges. For both countries to gain, that price must sit strictly between the two autarky ratios.
Eq. 22.2 is the gains-from-trade range. At any world price strictly inside the interval, Home can sell cloth for more wine than it could get at home, and the partner can sell wine for more cloth than it could get at home; each specializes in its comparative-advantage good and trades for the other. The wage ratio that supports both countries producing follows from the same logic: $w/w^*$ must lie in $\left(a_{LW}^*/a_{LW},\, a_{LC}^*/a_{LC}\right)$, high enough that Home's absolute productivity edge is not undercut, low enough that the partner stays competitive somewhere (Eq. 22.3).
Pourquoi c’est important : Why must the world price fall between the two home prices? Because a price outside the range gives one country a worse deal than it already had at home, and it simply refuses to trade. If cloth is cheaper on the world market than Home could make it for, Home buys cloth instead of selling it, and then nobody is selling Home the wine it wants. Trade only clears when the price is good news for both: each country gets more of the other good per unit of its own than it could squeeze out of its own backyard. Drag the world-price slider on the figure past either endpoint and watch one country drop out; the "no trade" flag is the range condition made visible.
Figure 22.1. Ricardian production frontiers for Home and its partner, with the world-price line. When the world relative price of cloth sits inside the gains range, both countries specialize and consume at a point beyond their own frontier (the consumption markers). Push the price slider past either autarky ratio and the figure flags "no trade — price outside gains range." Drag the sliders.
Problem. Home needs $a_{LC}=2$, $a_{LW}=4$ worker-hours per unit; the partner needs $a_{LC}^*=6$, $a_{LW}^*=3$. Each has 120 worker-hours. (a) Who has absolute advantage in each good? (b) Who has comparative advantage in cloth? (c) Find the gains range and pick a world price inside it. (d) Show Home consumes beyond its frontier.
Solution.
(a) Home is absolutely more productive in both: 2 < 6 in cloth, 4 > 3 in wine, so Home has absolute advantage in cloth only; the partner is better at wine. (b) Home's opportunity cost of cloth is $a_{LC}/a_{LW}=2/4=0.5$ wine; the partner's is $6/3=2$ wine. Since $0.5 < 2$, Home has comparative advantage in cloth, the partner in wine.
(c) The autarky price ratios are $0.5$ (Home) and $2$ (partner), so the gains range is $0.5 < P_C/P_W < 2$. Pick $P_C/P_W = 1$. (d) In autarky Home produces and consumes on its frontier (max cloth $120/2=60$, max wine $120/4=30$). Specializing fully in cloth gives 60 cloth; selling 20 cloth at price 1 buys 20 wine, leaving the bundle $(40\text{ cloth}, 20\text{ wine})$. Home's own frontier line is $W = 30 - 0.5\,C$, so 40 cloth would permit only $30 - 20 = 10$ wine in autarky. Twenty wine beats ten: the traded bundle lies outside the frontier. Home gains.
The intellectual descent of these ideas, from Hume's price-specie-flow argument against permanent surpluses to Ricardo's 1817 statement of comparative advantage as the reply to mercantilism, is told as a history of ideas in the history-of-economic-thought book: Mercantilism, physiocracy, Hume, and Classical political economy: Ricardo on comparative advantage.
The Atlantic and Indian-Ocean network Ricardo wrote against, the Iberian, Dutch, and English trading systems of the mercantilist era, is mapped in the economic-history book's account of early modern globalization; the formal models here are what came after.
The Ricardian model you just built is where the free-trade argument starts: both countries gain, even the less productive one. This is the apparatus the free-trade walkthrough compresses at its first stage.
Comparative advantage (section 22.1) makes the efficiency case for free trade at its strongest: specialization moves both countries' consumption beyond their production frontiers, and the gain does not require either country to be good at anything in absolute terms. The tariff deadweight-loss triangle you met in section 2.6 is the mirror image: a tax on imports throws away part of that gain. This is the floor of the free-trade case, and it is not seriously contested.
Ricardo establishes that the country as a whole gains. He is silent on who inside the country gains: there is only one factor, labor, so there are no internal winners and losers to speak of. That silence is filled in section 22.3, and it is where the live argument over free trade actually lives.
Ricardo tells us that countries gain and that comparative advantage drives the pattern, but he leaves the origin of comparative advantage as a brute fact about technology. Eli Heckscher and Bertil Ohlin offered a deeper answer: comparative advantage comes from what a country has, not just what it knows how to do. Give production two factors, call them capital and labor, and let countries differ in their endowments rather than their technologies. Now a clean prediction falls out.
The Heckscher-Ohlin theorem joins these two ideas. A country exports the good that uses its abundant factor intensively. The capital-abundant country exports the capital-intensive good; the labor-abundant country exports the labor-intensive good. Technology is assumed identical across countries, so the entire pattern of trade is read off endowments.
Pourquoi c’est important : A country that is drowning in capital and short on workers finds capital cheap and labor dear; whatever uses a lot of capital is cheap to make there. A country with the opposite mix finds labor cheap. Each ends up cheapest at the good that leans on its plentiful resource, so each exports that good. Trade then becomes a roundabout way of shipping the abundant factor abroad: when a capital-rich country exports machinery, it is really exporting the services of its surplus capital, and importing, embedded in apparel, the labor it is short of. Slide the endowment lever on the figure across the intensity threshold and the export good flips, which is the whole point: the pattern is not fixed by what a country is good at, but by what it happens to have a lot of.
A striking implication follows. If both countries make both goods with the same technology and trade freely, the goods-price equalization that trade produces forces the underlying factor prices into line as well. The wage in the labor-abundant country rises toward the wage in the labor-scarce one; the return to capital converges the other way.
Eq. 22.5 holds only inside the cone of diversification, the range of endowments over which both countries produce both goods. Once a country is so extreme in its endowment that it specializes completely, the link breaks and factor prices need not equalize. In practice trade costs, technology gaps, and complete specialization mean factor prices converge partially, not exactly; FPE is the limiting theorem, not a description of the data.
Pourquoi c’est important : You do not need workers to cross the border for wages to chase each other. When a labor-rich country starts exporting labor-intensive shirts, its factories bid for more workers and wages there rise; meanwhile the labor-scarce country buys those shirts instead of making them, its garment factories shrink, and demand for its workers eases. Shipping the goods does the work that immigration would have done. The theorem is a clean statement of a real worry: opening trade with a labor-abundant partner can press on wages at home, not because anyone moved, but because the goods did.
Problem. Home has $K/L = 4$; its partner has $K^*/L^* = 1$. Aircraft use $K/L = 5$ at common prices; apparel uses $K/L = 0.5$. (a) Which factor is Home abundant in? (b) Which good is capital-intensive? (c) Predict the export pattern. (d) Whose real wage does opening trade put under pressure, and in which country?
Solution.
(a) Home's capital-to-labor ratio (4) exceeds the partner's (1), so Home is capital-abundant and the partner labor-abundant. (b) Aircraft, at $K/L=5$ versus apparel's 0.5, is the capital-intensive good. (c) By Heckscher-Ohlin, Home exports aircraft (capital-intensive, its abundant factor) and imports apparel; the partner does the reverse.
(d) Opening trade raises the relative price of aircraft at Home (it now sells to the world) and lowers the relative price of apparel. As section 22.3 shows, the factor used intensively in the falling-price good, labor at Home, sees its real return fall. So it is the scarce factor inside the capital-rich country (Home labor) that comes under pressure: the same prediction that the China-shock evidence in section 22.7 confirmed for the United States.
Heckscher's 1919 essay and Ohlin's 1933 book, which turned this from an intuition into a model, sit inside the marginalist formalization of economics traced in the history-of-economic-thought book: The marginalist revolution and formalization.
Figure 22.2. Heckscher-Ohlin endowment box. The dark ray is Home's capital/labor endowment; the two lighter rays are the factor intensities of good X (capital-intensive) and good Y (labor-intensive). Home exports the good whose intensity ray its endowment leans toward. Cross the threshold and the export good — and the "abundant factor" badge — flips. Drag the sliders.
The Heckscher-Ohlin pattern looks like unambiguous good news: both countries gain, and the gains rest on nothing more than getting more of the abundant factor's services out into the world. Wolfgang Stolper and Paul Samuelson showed in 1941 that the good news has a sharp distributional edge, and that the edge is not an accident or a market failure — it is the model functioning exactly as designed.
When trade opens, the relative price of a country's export good rises and the relative price of its import-competing good falls. The Stolper-Samuelson theorem traces those goods-price changes down to factor prices: the real return to the factor used intensively in the rising-price good goes up, and the real return to the other factor goes down, not just relatively but absolutely.
Trade amplifies the change as it transmits it. A given percentage change in a goods price produces a larger percentage change in one factor price and a fall in the other. This is the magnification effect, and it is what makes trade's distributional consequences politically heavy out of proportion to the price moves that cause them.
Pourquoi c’est important : This is the result that turns "trade is good for the country" into "trade is good for some people and bad for others, predictably." When a capital-rich country opens to a labor-rich one, the price of labor-intensive imports falls, those industries shrink, and the workers in them lose in real terms, not just relative to capital owners. The model does not treat this as a glitch to be corrected; it is the mechanism by which the country captures its gains. The same equations that promise the aggregate gain also forecast the losers. Drag the price-change slider on the figure and watch the factor-return bar swing further than the price bar that drove it; that overshoot is the magnification, and the scarce factor's real-return readout going negative is the loss the theory predicts.
Figure 22.3. Stolper-Samuelson magnification. A percentage change in the relative price of the labor-intensive good (the price bar) maps into a larger change in the wage and a fall in the real return to capital (the factor bars overshoot the price bar in opposite directions). The dashed 45° reference marks one-for-one pass-through; the wage bar passing it is the magnification. Drag the slider.
Stolper-Samuelson runs from goods prices to factor prices. Tadeusz Rybczynski's 1955 theorem runs the other way: hold goods prices fixed and ask what happens to outputs when a country's factor endowment changes, say immigration raises the labor supply. The answer is again magnified and again counterintuitive.
Pourquoi c’est important : When a wave of workers arrives, the labor-intensive industries do not just expand; they expand by pulling capital out of the capital-intensive industries, which actually shrink. The new workers need some capital to work with, and the only place to get it at unchanged prices is from the other sector. So a country that accumulates labor tilts hard toward labor-intensive goods, and a country that accumulates capital tilts toward capital-intensive ones. This is why fast-growing economies that build up capital see their export mix shift toward heavier, more capital-intensive products over time: the Rybczynski tilt is the engine behind the changing composition of a country's exports as it develops.
Problem. A country produces apparel (labor-intensive) and machinery (capital-intensive) at fixed world prices. Immigration raises the labor force by 10%, capital unchanged. Use the magnified output response (suppose the Rybczynski elasticity of apparel output to labor is 1.8). (a) What happens to apparel output? (b) To machinery output? (c) Why does machinery output fall when nothing was taken from it directly?
Solution.
(a) Apparel output rises by about $1.8 \times 10\% = 18\%$, more than the 10% labor increase, the magnification. (b) Machinery output falls; with capital fixed and apparel absorbing more than the new labor's worth of capital, machinery contracts (here roughly $-6\%$ on plausible shares). (c) Nothing is taken from machinery by decree, but the new workers each need some capital. At fixed prices the wage cannot fall to absorb them in machinery, so apparel, which needs little capital per worker, expands and bids capital away from machinery. The contraction is the mirror image of apparel's magnified boom.
Heckscher-Ohlin made a flat, testable prediction, and in 1953 Wassily Leontief tested it on the obvious case. The United States was the most capital-abundant economy on earth; the theory said its exports should be more capital-intensive than its imports. Leontief computed the factor content of US trade from his input-output tables and found the reverse: American exports were more labor-intensive than the goods America imported. The most capital-rich country in the world was, by the measure, exporting labor.
The result was a genuine break, the kind that disciplines a field rather than ending it. The resolutions did not discard Heckscher-Ohlin; they enriched what counts as a factor. Treating skilled labor (human capital) as a third factor showed that US exports were intensive in skill, which America was abundant in, even if not in raw physical capital. The Heckscher-Ohlin-Vanek formulation generalized the theorem to many factors and predicted the factor content of trade rather than which goods cross the border, and measurement work showed that allowing for technology differences and home bias in demand removed much of the residual paradox.
Pourquoi c’est important : The paradox is a lesson in how good theories survive bad news. The clean two-factor story said capital-rich America should ship out capital-heavy goods; the data said otherwise, and rather than throw the model away, economists asked what "capital" was leaving out. The answer was the trained engineer, the skilled machinist, the research lab: America's real surplus was in human capital, and once you count skill as a factor, the country was exporting exactly what it had most of. The episode is why modern trade work measures the factor content of trade carefully and never assumes "capital" means only machines.
Stolper-Samuelson is where the free-trade argument gets hard. The model that promises aggregate gains also predicts that a definite group of workers loses, in real terms. This is the apparatus stage 2 of the free-trade walkthrough rests on.
Trade raises the real return to a country's abundant factor and lowers the real return to its scarce factor (section 22.3). In a rich country trading with a labor-abundant one, that means lower real wages for the workers who compete most directly with imports. This is not a market failure to be fixed — it is the mechanism by which the country captures its gains. Free-trade advocates who deny the losers exist are arguing against the model they are invoking.
The aggregate gain is real and the concentrated loss is real; both are predictions of one model. The live disagreement is whether the winners actually compensate the losers, a question the theory leaves to politics, and which section 22.7's China-shock record answers in the negative for the United States.
Second stop of the thread that traces the trade idea from Hume's specie-flow through to the gravity equation. This is the factor-endowment turn: Heckscher-Ohlin and the distributional theorem that came with it.
The thread enters its second phase here. Ricardo (section 22.1) explained that countries gain; Heckscher and Ohlin (section 22.2) explained what each exports, grounding comparative advantage in endowments. Stolper and Samuelson (section 22.3) then read the distributional consequences off the same model, and the Leontief paradox marked where the pure-endowment story met its empirical limit. This is the apparatus the thread previously had no A-book home for; it lives here now.
The thread's verdict is cumulative, not corrective: each model adds a layer the previous could not produce. By the end of this stop the reader holds the factor theory and its distributional edge, plus the open question (rich countries trading similar goods) that the increasing-returns models at the next stop were built to answer.
Look at who actually trades the most with whom and a problem appears. Germany and France, with similar endowments, similar technology, and similar wages, are each other's largest trading partners, and much of what crosses the border is the same kind of good going both ways: German cars to France, French cars to Germany. The classical models cannot explain this. Comparative advantage and Heckscher-Ohlin both predict that similar countries have little reason to trade and that whatever trade occurs is inter-industry, cloth for wine, machinery for apparel. Roughly half of world trade is the opposite: similar countries exchanging similar goods.
Paul Krugman's 1979 and 1980 papers supplied the missing engine. Drop the classical assumption of constant returns and let production have a fixed cost: average cost then falls as output rises. Combine increasing returns with product differentiation, where every firm makes a slightly different variety and consumers value variety, and you get monopolistic competition, the Dixit-Stiglitz structure developed in Chapter 6. Each firm produces one variety at a scale large enough to spread its fixed cost; free entry drives profit to zero and pins down the number of varieties a market can support.
With CES demand of elasticity $\sigma > 1$, each monopolistically competitive firm sets a constant markup over marginal cost, $p = \frac{\sigma}{\sigma-1}\,c$ (Eq. 22.10). Free entry drives profit to zero, which fixes each firm's output and therefore the number of varieties the market sustains. Opening trade is, in this structure, equivalent to enlarging the market: the integrated market supports more varieties, each produced at larger scale and hence lower average cost.
Pourquoi c’est important : Take two countries that are carbon copies of each other: same endowments, same technology, no comparative advantage anywhere. Classical theory says they have no reason to trade. Yet merging their markets makes both richer, for two reasons that have nothing to do with who is better at what. First, the combined market supports more car models, more software, more types of everything, and people value the choice. Second, each surviving variety is now sold to twice as many buyers, so factories run longer and average cost falls. Trade here is not about specializing where you are relatively good; it is about pooling demand so that fixed costs get spread thin and variety blooms. Slide the market-size lever on the figure with the "no comparative-advantage difference" banner pinned on, and watch varieties rise and average cost drop: gains appear where Ricardo and Heckscher-Ohlin both predict none.
Figure 22.4. New trade theory: the average-cost curve $AC(Q)=F/Q+c$, with the zero-profit equilibrium output and the number of varieties the integrated market sustains. Growing the combined market (trade between identical countries) raises the variety count and slides each firm down its AC curve — with the comparative-advantage difference pinned at zero. Drag the sliders.
Problem. Two identical countries, each of market size $L=1000$, marginal cost $c=1$, CES elasticity $\sigma=4$, fixed cost $F=100$. (a) Find the number of varieties and the markup price in autarky for one country. (b) After integration ($L=2000$), find the variety count. (c) State the source of the gain and contrast it with the Ricardian gain.
Solution.
(a) Markup price $p = \frac{4}{3}(1) = 1.33$, independent of market size. Varieties $n = L/(\sigma F) = 1000/400 = 2.5 \approx 2$ per country. (b) Integrated market $n = 2000/400 = 5$ varieties, more than the 2 each country had alone, and each is produced at larger scale, so average cost falls. (c) The gain is more varieties at lower average cost. It is not Ricardian: there is no comparative-advantage difference between the countries, no specialization by relative productivity. The Ricardian gain comes from differences; this gain comes from size. Both are real, and they add.
Krugman's 1979 increasing-returns model is treated as a counter-revolution moment in the history of trade thought, the point where economists stopped deriving trade only from differences: see Krugman's new trade theory on the timeline. The underlying monopolistic-competition model (the Dixit-Stiglitz CES structure used here in compressed form) is developed in full in Chapter 6, Market Structures and Game Theory.
In 1962 Jan Tinbergen noticed that bilateral trade flows obey a law that looks like Newton's. Trade between two countries rises with the product of their economic sizes and falls with the distance between them: gravity, with GDP playing the role of mass. The fit is uncanny and has survived sixty years of better data and tougher tests: the gravity equation is the most stable empirical regularity in all of trade economics.
The naive equation works in practice but for a long time had no theory behind it; it was a curve fit. James Anderson and Eric van Wincoop showed in 2003 that the equation falls out of a structural model once you include what they called multilateral resistance: trade between $i$ and $j$ depends not only on the bilateral distance between them but on how far each is from all its other potential partners. A country surrounded by close, large partners trades less with any one of them than an isolated country of the same size would, because it has more attractive alternatives.
Eq. 22.12 is structural gravity: bilateral trade as a share of world output $Y_W$, scaled by bilateral trade cost $\tau_{ij}$ relative to the multilateral-resistance terms $\Pi_i$ (outward) and $P_j$ (inward), with $\sigma$ the trade elasticity. A deep reason the gravity form fits regardless of the underlying mechanism is that it emerges from almost any trade model with trade costs: Armington, Krugman, Eaton-Kortum, and Melitz all produce a gravity equation. Gravity is the reduced form many micro-foundations share. This is why it became the workhorse of the credibility revolution in trade empirics (the structural-estimation connection to Chapter 10).
Pourquoi c’est important : Big economies trade a lot and distant ones trade little; that part is obvious. The non-obvious part is multilateral resistance: how much two countries trade depends on who else is nearby. Imagine a shop that is the only one in a remote town versus the same shop in a crowded mall. The remote shop captures all of the town's custom; the mall shop, surrounded by rivals, captures less even though the town and the mall have the same number of shoppers. Countries are the same. New Zealand trades a lot with Australia partly because New Zealand has few close alternatives; a country in the middle of Europe, ringed by large neighbors, spreads its trade thin. On the figure, shrink the distance between one pair and their trade grows; but then grow a third nearby partner and watch the original pair's trade fall, even though nothing changed between them. That fall is multilateral resistance, and it is invisible in the naive equation.
Figure 22.5. Structural gravity for a small country set. The bars show Home's predicted trade with each partner. Shortening the Home–A distance raises Home–A trade; growing the nearby third partner B raises Home's multilateral resistance and reduces predicted Home–A trade even though the Home–A distance is unchanged. Drag the sliders.
Problem. Two countries have GDPs $Y_i = 100$, $Y_j = 50$ and distance $D_{ij} = 2000$ km, with $G$ calibrated so naive predicted trade is 25. (a) Predict bilateral trade from the naive equation. (b) A large new partner $k$ opens up close to $i$. Qualitatively, what happens to predicted $i$–$j$ trade, and why? (c) Which equation captures the effect in (b)?
Solution.
(a) Naive gravity: $X_{ij} = G\,Y_i Y_j / D_{ij}$, giving the calibrated 25. (b) Predicted $i$–$j$ trade falls: the close, large partner $k$ raises $i$'s multilateral resistance, so $i$ now has an attractive alternative, so the same bilateral distance buys less trade with $j$. Nothing changed between $i$ and $j$, yet their trade shrinks. (c) The naive equation (Eq. 22.11) cannot see this; it has no term for third parties. The structural equation (Eq. 22.12) captures it through the $\Pi_i$ multilateral-resistance term, the central correction Anderson and van Wincoop supplied.
Gravity describes flows between countries; Marc Melitz's 2003 model describes who inside a country actually does the exporting. The micro fact that motivated it is stark: within any narrowly defined industry, only a minority of firms export, and the exporters are systematically the most productive. Exporting carries a fixed cost (establishing distribution abroad, meeting foreign standards), and only firms productive enough to earn back that cost choose to pay it.
Pourquoi c’est important : Trade does not lift all boats evenly inside an industry; it sorts them. There is a price of admission to exporting (setting up shop abroad, clearing foreign regulations), and only the most productive firms clear it. When trade costs fall, the price of admission drops, more firms step over the threshold, and the strongest firms grab a larger share of the market while the weakest, now facing import competition at home, shrink or close. The industry's average productivity rises not because every firm improves but because resources move from the laggards to the leaders. This within-industry reshuffling is a gains channel the country-level models miss entirely, and it is also why trade liberalization can be brutal for particular firms even as it makes the industry as a whole stronger. On the figure, drag the trade-cost slider down and watch the cutoff line slide left, shading in the newly-exporting productive tail.
Figure 22.6. The Melitz productivity distribution. The curve is the density of firm productivity (Pareto-shaped); the vertical line is the export cutoff $\varphi^*$. Firms to the right of the cutoff (shaded) export. Lowering trade cost slides the cutoff left, pulling more — and more productive — firms into exporting and reallocating share toward them. Drag the slider.
The estimation of gravity and heterogeneous-firm models, turning these structures into regressions that survive endogeneity and omitted-variable problems, belongs to the credibility revolution in econometrics: Chapter 10, Econometrics Foundations. The shift toward firm-level data and the empirical turn that produced Melitz's model is traced as a history of ideas in the history-of-economic-thought book: the modern empirical turn.
The thread that began with Hume's specie-flow argument reaches its empirical terminus here: the gravity equation and the firm-level micro beneath it.
The thread closes on the two results that turned trade theory into a quantitative science: the gravity equation (section 22.5), the most stable regularity in the field and the reduced form almost every trade model shares; and the Melitz heterogeneous-firm model, which explains who actually exports. From Hume's eighteenth-century worry about gold flows to a structural equation estimated on millions of customs records: that is the distance the thread travelled.
No model on this thread was overthrown. Comparative advantage explains inter-industry trade; factor endowments explain its pattern and distribution; increasing returns explain intra-industry trade; gravity describes the volume; Melitz describes the firms. They are layers, not rivals, and gravity's mechanism-agnosticism is the proof, since it emerges from all of them.
The gravity equation is the apparatus that turns "trade with the EU matters more than trade with faraway partners" from a slogan into a prediction. This is the case_brexit walkthrough's apparatus beat.
Gravity (section 22.5) says trade is dominated by large, close partners, and the EU is both, on the UK's doorstep. Raising trade costs with the EU (the $\tau_{ij}$ term) cannot be offset by deals with distant economies, because distance enters the equation directly and multilateral resistance means faraway partners are poor substitutes for a large near one. The same multilateral-resistance logic explains why disrupting deeply integrated supply chains (the gravity model's modern extension to global value chains) is costlier than a tariff-revenue calculation alone suggests.
Gravity gives the economic cost of raising trade barriers with a large near neighbor; it does not weigh that cost against the sovereignty and political goals that drove the decision. The apparatus quantifies the trade-off; the walkthrough argues it.
Start with the case you already know. A small country, one too small to affect the world price, that imposes a tariff simply raises its domestic price by the tariff, shrinks imports, transfers some surplus to its producers and some to the treasury, and destroys the rest. The two deadweight-loss triangles from section 2.6 are pure waste: the small country's tariff makes it unambiguously poorer. For a price-taking country there is no economic case for a tariff on efficiency grounds, and pretending otherwise on the settled small-country case is the mistake to avoid.
The large country is different, and this is the one door the trade model genuinely leaves open. A country large enough that its import demand moves the world price faces an upward-sloping foreign export supply. When it taxes imports, part of the tariff is borne by foreign exporters, who cut their price to keep selling, so the importing country's terms of trade improve. Up to a point, that terms-of-trade gain outweighs the deadweight loss, and a positive tariff raises the large country's welfare.
The optimal tariff is a beggar-thy-neighbor policy: the importing country's gain comes partly at the expense of its trading partner, whose export prices are pushed down. That sets up the problem with it. If both large countries play the optimal-tariff game, each retaliates, and the Cournot-style Nash equilibrium of the resulting trade war leaves both with high tariffs, shrunken trade, and lower welfare than free trade, a classic prisoner's-dilemma structure (the oligopoly-game apparatus is in Chapter 6).
Pourquoi c’est important : A country big enough to be a major buyer has market power, just like a large customer who can squeeze a supplier. By taxing imports it forces foreign sellers to eat part of the tax, and it pockets the difference as cheaper imports, a genuine gain and the only one the theory grants. But it is a gain taken from the partner, and the partner can play the same game. When both do, both end up with high walls and little trade, each worse off than if neither had started. On the figure, raise the tariff and watch the terms-of-trade gain grow alongside the deadweight-loss triangles; net welfare climbs to a peak at the optimal rate and then falls. Flip the retaliation toggle and the inset payoff grid drops both countries into the bottom-right cell: the trade-war Nash, worse for everyone than the free-trade corner.
Figure 22.7. The large-country import market. As Home raises its tariff, the terms-of-trade gain (foreign price falls) grows alongside the two deadweight-loss triangles; net welfare peaks at the optimal tariff $t^*=1/\varepsilon^*$ and then declines. Switch the scenario slider to "Retaliation" and the inset payoff grid highlights the trade-war Nash cell, where both countries are worse off than under free trade. Drag the sliders.
Problem. A large country faces foreign export supply with elasticity $\varepsilon^* = 4$. (a) Compute the optimal tariff. (b) Explain why a higher tariff than this reduces welfare. (c) If the partner retaliates at the same rate, sketch why both end up worse than free trade.
Solution.
(a) $t^* = 1/\varepsilon^* = 1/4 = 25\%$. (b) At the optimum, the marginal terms-of-trade gain just equals the marginal deadweight loss. Beyond 25%, the deadweight-loss triangles grow faster than the terms-of-trade rectangle, so net welfare falls. As $\varepsilon^* \to \infty$ (small country, perfectly elastic foreign supply), $t^* \to 0$: the small-country result drops out as a special case.
(c) Each country's optimal tariff is a best response that improves its terms of trade at the partner's expense. When both impose tariffs, each loses the terms-of-trade benefit (the partner is doing the same thing back) but keeps the deadweight loss. The Nash equilibrium of the tariff-setting game is mutual high tariffs and shrunken trade, a Pareto-inferior outcome relative to mutual free trade and the trade-war analog of the prisoner's dilemma.
A second, narrower door opens once markets are oligopolistic. James Brander and Barbara Spencer showed in the 1980s that when a domestic and a foreign firm compete in a third market, a government subsidy to the domestic firm can shift profit toward it. The subsidy lets the domestic firm credibly commit to a larger output; the foreign rival, best-responding, contracts; the profit captured can exceed the subsidy's cost. It is a real result, and a fragile one.
The fragility is the whole story. The Brander-Spencer result is built on Cournot (quantity) competition; under Bertrand (price) competition the optimal policy flips to a tax, not a subsidy. It assumes the rival's government does not retaliate, that the targeted industry actually earns rents, and that policymakers can pick the winner. Drop any of these and the case collapses. Strategic trade policy is a coherent theoretical possibility, not a general license to subsidize.
The oldest interventionist argument is the infant-industry case, and the new-trade-theory apparatus is what gives it teeth. If an industry has increasing returns or learning-by-doing, with costs that fall as cumulative output grows, a country could be locked out of an industry it would eventually dominate, simply because incumbents abroad got there first and now operate at a scale newcomers cannot match. Temporary protection lets the domestic industry climb its learning curve until it can compete unaided; comparative advantage, on this view, is partly created rather than given.
The conditions are demanding, and stating them honestly is the point. The learning externality must be real and must spill outside the protected firm (otherwise the firm would invest in learning itself); protection must be temporary and credibly removed (otherwise it shelters permanent inefficiency); and the state must be able to identify the industry and resist capture. Where those hold, and parts of the East Asian growth record are the standard exhibits, the argument has force; where they do not, it becomes a rationalization for permanent rents. The development application of this argument is treated in Chapter 20.
The small-country tariff diagram recapped here is developed at intro level in Chapter 2, section 2.6 (International Trade); the Cournot and Stackelberg games behind the retaliation Nash and the Brander-Spencer profit-shift are in Chapter 6, section 6.5; and the infant-industry case as a development strategy is treated in Chapter 20, Development Economics.
This section is where the question gets its in-model answer. The optimal-tariff and strategic-trade results are the two cases the theory grants, and the conditions on each are the catch.
For a small country the answer is no: a tariff is pure deadweight loss (section 2.6). For a large country, yes, with a ceiling: the optimal tariff $t^*=1/\varepsilon^*$ captures a terms-of-trade gain from foreign exporters (section 22.6). And in oligopolistic industries, a strategic subsidy can shift profit to a domestic firm (Brander-Spencer). Both are genuine results; neither is a general license.
The optimal tariff is beggar-thy-neighbor: it invites retaliation, and the trade-war Nash leaves everyone worse than free trade. Strategic trade policy reverses under price competition, assumes no foreign retaliation, requires real rents, and demands that the government pick winners. The settled small-country case and the conditional large-country case are different animals, and conflating them is the most common error in the tariff debate.
Tariffs are "good" in exactly two narrow, condition-laden senses, and the conditions (large country, no retaliation, identifiable rents) are precisely what real trade wars violate. The apparatus does not say tariffs are always bad; it says the cases where they help are specific, fragile, and easy to mistake for the general one. The live policy verdict is the walkthrough's to argue; the apparatus is here.
The infant-industry / dynamic-comparative-advantage apparatus here is the formal core of the Smith-versus-List dispute over whether a nation should protect young industries.
Adam Smith's case for free trade rests on static comparative advantage: specialize where you are relatively good now. Friedrich List's national-economy reply rests on what section 22.6 calls dynamic comparative advantage: with learning-by-doing and increasing returns, advantage can be built, and a latecomer may need temporary protection to climb the learning curve incumbents already descended. The new-trade-theory machinery is what turns List's intuition into a model with stated conditions.
The dispute is not settled in the abstract; it turns on whether the infant-industry conditions hold: a real, external learning spillover; credibly temporary protection; a state that can pick the industry and resist capture. Where they hold, List has a point; where they do not, Smith does. The apparatus converts a clash of doctrines into a checklist of conditions.
The strategic-trade and scale-economy results are the most sophisticated case for intervention the theory offers, and the conditions on them are where the free-trade walkthrough's stage 4 does its work.
Once increasing returns and oligopoly are in the model, the first-best free-trade benchmark loses its monopoly on the argument: strategic trade policy (Brander-Spencer) can shift rents, and scale economies plus learning give the infant-industry case a formal footing (section 22.6). These are the strongest in-model arguments a free-trade skeptic can field.
Each exception is real but conditional and fragile, sensitive to the form of competition, to retaliation, and to the government's ability to identify rents and resist capture. The honest position is that free trade is the right default precisely because the exceptions are hard to implement well, not because they do not exist. The walkthrough argues where the default should bend; the apparatus marks where it can.
For most of the twentieth century the distributional prediction of section 22.3 stayed an abstraction: true in the model, modest in the data, easy to wave away with the promise that the winners could compensate the losers. Then China entered world manufacturing on a scale and at a speed the models had never been stress-tested against, and the abstraction arrived as a fact in particular American towns.
David Autor, David Dorn, and Gordon Hanson's research on what they named the China shock found that the surge in Chinese import competition after roughly 2000 produced losses that were concentrated, persistent, and geographically local. Communities whose factories competed most directly with Chinese imports saw manufacturing employment fall and not recover; wages stayed depressed, labor-force participation dropped, and the adjustment that theory assumed, with workers moving to new regions and industries, happened far more slowly and incompletely than the textbook supposed. The aggregate gains from trade with China were real; so was the bill, and it landed on a small set of places.
This is Stolper-Samuelson realized. The factor that competed most directly with the labor-abundant entrant, less-skilled manufacturing labor in a capital- and skill-abundant economy, saw its real prospects fall, exactly as the magnification effect (Figure 22.3) predicts when the import-competing good's relative price collapses. The China shock did not refute trade theory; it confirmed the part of it the efficiency-focused thread had under-weighted. The model always said there would be losers. What the political system never honored was the other half of the standard defense: that the winners would compensate them.
Pourquoi c’est important : The economics textbook had two sentences about trade and workers. The first, "the country as a whole gains," got repeated endlessly. The second, "some specific workers lose, in real terms, and they need to be compensated," got treated as a footnote. The China shock is what happens when you act on the first sentence and ignore the second. The gains were real and diffuse: slightly cheaper goods for everyone. The losses were real and concentrated: a factory town that never came back. Both were in the model from the start. The reason the China shock felt like a betrayal of economics rather than a confirmation of it is that the profession had spent decades quoting half the prediction. This is not "trade is bad." It is a parameter-magnitude split inside a frame everyone agreed on: the gains and the losses are both larger and more uneven than the casual version admitted.
One reframe closes the loop with the macro side. The "China is stealing our jobs through its trade surplus" framing misreads the accounting: a trade deficit is the mirror image of a capital-account surplus: the dollars that buy Chinese goods come back as investment in American assets, by the balance-of-payments identity $CA + KA = 0$ developed in Chapter 17. The labor-market damage is real and is a distributional story about which workers bear the adjustment, not a story about the aggregate trade balance draining the economy.
The accounting behind "a trade deficit is a capital-account surplus," and why the trade balance is not the right scoreboard for the China-shock debate, is developed in Chapter 17, section 17.1 (the balance of payments).
The historical record of China's integration into the world economy is narrated in the economic-history book: China's reform era, the wider globalization of the 1990s–2000s in the globalization decades, and the backlash that followed the 2008 crisis in the GFC and after.
You now hold the full positive apparatus (six models, layered not competing), the policy corollaries, and the distributional record. This is where the free-trade question gets its in-chapter resolution.
Free trade raises aggregate real income: comparative advantage (section 22.1), factor endowments (22.2), and increasing returns (22.4) each add a distinct source of gain, and gravity (22.5) shows how reliably trade follows size and proximity. The efficiency case is as solid as anything in economics. On the small-country tariff it is decisive: a tariff is pure waste.
Three in-model qualifications survive. Distribution: Stolper-Samuelson (22.3) guarantees concentrated losers, and the China shock (22.7) shows the losses can be large, lasting, and unpaid. Terms of trade: a large country has an optimal tariff (22.6). Dynamics: strategic trade and infant-industry arguments (22.6) show advantage can be built. None of these is a fringe objection; each is a result inside the same apparatus that makes the free-trade case.
The discipline did not reverse on free trade; it conditioned it. The consensus moved from "free trade is unambiguously optimal" to "free trade maximizes aggregate income, but the distribution and the dynamics require policy attention": adjustment assistance, not protection, was the textbook remedy, with the China-shock evidence forcing a hard look at whether that remedy was ever actually delivered.
Free trade is good for aggregate efficiency and the right default for policy, but "always good" is the wrong frame. The honest statement is that trade reliably enlarges the pie and reliably reslices it, that the reslicing can be brutal and local, and that the case for free trade has always depended on a compensation promise the political system did not keep. The interesting question is not "is trade good?" but "good for whom, by how much, and who bears the adjustment?", and that is a parameter-magnitude question inside a frame the apparatus settles.
The apparatus tells you the gains and the losses exist and how large each can be; it cannot tell you how much weight to put on a concentrated loss versus a diffuse gain: that is a value judgment about distribution, not a theorem. Nor can it settle when the dynamic and strategic exceptions are worth the implementation risk. These are exactly the questions the free-trade, tariff, and Brexit walkthroughs take up in argument mode.
The China shock was real and the losses were local and lasting, but "stole our jobs" misreads a distributional story as an aggregate-balance one. Stolper-Samuelson predicted the losers; the compensation never came.
AvancéComparative advantage proves the country gains in aggregate. It is silent on who inside the country gains, and quoting it as a verdict on the distributional question is the most common misuse of the theorem.
AvancéThe trade-deficit-as-capital-inflow reframe above is the light apparatus this case borrows when it connects payment imbalances to the monetary order's collapse.
Trade pattern and gains-from-trade are this chapter's domain; exchange-rate dynamics and the balance-of-payments accounting are open-economy macro's (Chapter 17). The China-shock reframe (a trade deficit is a capital-account surplus) is the one piece of trade-side apparatus this case needs, and it sits at the boundary the two chapters share.
We watched one pair the whole way. Ricardian (22.1): Home and its partner gain by specializing along comparative advantage (Home in cloth, the partner in wine) even though one of them is worse at both. Heckscher-Ohlin (22.2): why each exports what it does: Home is capital-abundant and ships the capital-intensive good; the partner is labor-abundant and ships the labor-intensive one.
Stolper-Samuelson (22.3): who inside each country wins and loses: Home's scarce factor, labor, sees its real wage fall, a seed that pays off in the China-shock section. New trade theory (22.4): the same pair also trades similar goods both ways once we allow increasing returns: German cars for French cars, not just cloth for wine. Gravity and Melitz (22.5): the volume of their trade obeys size and distance, and only the most productive firms in each country actually carry it across the border.
Policy (22.6): if Home is large, it has an optimal tariff, but if the partner retaliates both end up worse than free trade. The China shock (22.7): the real-world version of the pair, where the distributional bill the §22.3 seed warned about came due: concentrated, lasting losses in the towns that competed most directly with the labor-abundant entrant. Six models, one relationship, no model overthrown.
David Ricardo (1817). On the Principles of Political Economy and Taxation gave the cloth-and-wine example of England and Portugal that has anchored trade theory for two centuries, and supplied the answer to mercantilism that Hume's specie-flow argument had only gestured at.
Heckscher (1919) and Ohlin (1933). The Swedish economists grounded comparative advantage in factor endowments; Ohlin shared the 1977 Nobel Prize. Samuelson formalized the factor-price-equalization and Stolper-Samuelson results in the 1940s.
Wassily Leontief (1953). Using the input-output tables he won the 1973 Nobel for, Leontief found US exports more labor-intensive than its imports, the paradox that disciplined the field into refining what counts as a factor.
Paul Krugman (1979/1980). The increasing-returns models of intra-industry trade and the home-market effect earned Krugman the 2008 Nobel and turned trade theory toward geography and scale.
Autor, Dorn, and Hanson (2013 onward). The China-shock research program documented that trade's distributional losses were larger, more local, and more persistent than the profession had assumed: the empirical record that closes this chapter.
| Libellé | Équation | Description |
|---|---|---|
| Eq. 22.1 | $a_{LC}/a_{LW} < a_{LC}^*/a_{LW}^*$ | Comparative-advantage condition (Ricardian) |
| Eq. 22.2 | $a_{LC}/a_{LW} < P_C/P_W < a_{LC}^*/a_{LW}^*$ | Gains-from-trade range for the world price |
| Eq. 22.3 | $w/w^* \in (a_{LW}^*/a_{LW},\, a_{LC}^*/a_{LC})$ | Relative-wage range (Ricardian) |
| Eq. 22.4 | $K/L > K^*/L^* \Rightarrow$ export capital-intensive good | Heckscher-Ohlin theorem |
| Eq. 22.5 | $w = w^*,\ r = r^*$ (under FPE conditions) | Factor-price equalization |
| Eq. 22.6 | $\partial(w/P)/\partial p_{\text{labor-int.}} > 0$ | Stolper-Samuelson theorem |
| Eq. 22.7 | $\hat{w} > \hat{P} > 0 > \hat{r}$ | Magnification effect |
| Eq. 22.8 | $\hat{Q}_i > \hat{V}_j > 0,\ \hat{Q}_k < 0$ | Rybczynski theorem |
| Eq. 22.9 | $AC(Q) = F/Q + c$ | Increasing returns (average cost) |
| Eq. 22.10 | $p = \frac{\sigma}{\sigma-1} c$ | Monopolistic-competition markup (CES) |
| Eq. 22.11 | $X_{ij} = G\, Y_i Y_j / D_{ij}$ | Naive gravity equation |
| Eq. 22.12 | $X_{ij} = \frac{Y_i Y_j}{Y_W}(\tau_{ij}/\Pi_i P_j)^{1-\sigma}$ | Structural gravity (multilateral resistance) |
| Eq. 22.13 | $\varphi^*: \pi(\varphi^*) = 0$ | Melitz export-productivity cutoff |
| Eq. 22.14 | $t^* = 1/\varepsilon^*$ | Large-country optimal tariff |
| Eq. 22.15 | subsidy shifts oligopoly profit (Cournot) | Brander-Spencer profit shifting |
Ricardo (1817); Heckscher (1919); Ohlin (1933); Stolper & Samuelson (1941); Samuelson (1948, 1949); Rybczynski (1955); Leontief (1953); Vanek (1968); Tinbergen (1962); Krugman (1979, 1980, 1991); Dixit & Stiglitz (1977); Helpman & Krugman (1985); Anderson & van Wincoop (2003); Melitz (2003); Brander & Spencer (1985); Autor, Dorn & Hanson (2013, 2016).