International Business

The big picture

Pick up whatever you are reading this on. If it is a phone, the processor was probably designed in California or Cambridge, England, etched in Taiwan on machines built in the Netherlands from optics ground in Germany, packaged in Malaysia, fitted with a screen from South Korea and a camera module from Japan, assembled in China or India, and shipped in a steel box that spent three weeks crossing the Pacific. The company that sells it may be American. The company that made almost none of it also has its name on the back. Somebody in a treasury office hedged the currency risk on those component purchases, somebody in a compliance office checked whether any supplier appeared on a restricted party list, and somebody in a customs brokerage classified the finished device under a tariff heading that determined how much duty you indirectly paid.

Every one of those people works in international business, and none of them does the same job. That is the first thing to understand about this subject: it is not one discipline. International business is the study of commercial activity that crosses national borders, and it is a field defined by its problem rather than its method. Because borders are simultaneously economic lines, legal jurisdictions, currency zones, political units, and cultural regions, working across them pulls in economics, law, finance, politics, sociology, and logistics at once. A specialist in any one of those can be blindsided by another.

An honest note about what a text course can and cannot do. It can teach you the models, work the numbers with you, and show you what the evidence says, including where the evidence is contested. It cannot give you the experience of sitting in a room in Guangzhou at eleven at night realizing your interpreter has been summarizing rather than translating. Treat this like ground school before flight training: necessary, cheap, and not the same as flying.

This lesson does three things. It sorts out what counts as international business, it puts real numbers on the shape of the world economy, and it teaches you to read those numbers without being fooled by them.

The forms cross-border business takes

Students often use "international business" and "trade" interchangeably. Trade is one form among several, and the differences matter enormously to a firm because they differ in control, capital at risk, and legal exposure. Here is the ladder, from arms-length to deeply committed:

• Trade in goods: physical products crossing a border. This is what customs data measures and what most trade politics is about.
• Trade in services: consulting, software, tourism, education, shipping, insurance, royalties. Harder to measure, growing faster than goods trade, and often invisible in political debate even though the United States runs a large services surplus.
• Licensing and franchising: selling the right to use your intellectual property or business system. Low capital, low control, and the fastest way to lose your brand if you pick badly.
• Contract manufacturing and outsourcing: paying a foreign firm to make or do something for you without owning it.
• Foreign direct investment (FDI): acquiring a lasting interest in an enterprise in another country. The conventional statistical threshold is ten percent or more of voting shares, which distinguishes FDI from portfolio investment, where you own the stock but not the strategy.
• Joint ventures and wholly owned subsidiaries: FDI at its most committed. Building a plant abroad, or buying a company that already has one.

A multinational enterprise (MNE) is a firm that owns and controls value-adding activity in more than one country. That definition is deliberately about ownership and control, not about selling abroad. A firm that exports to sixty countries from one factory is an exporter. A firm with a single plant in Mexico is a multinational. Once you see the distinction you notice that most trade politics targets exporters while most tax and labor controversy targets multinationals.

Key idea: International business spans trade, services, licensing, contracting, and direct investment, and these differ mainly in how much control and capital the firm commits abroad.

How big is all this, actually?

Numbers keep you honest. World merchandise exports in the early 2020s ran in the neighborhood of twenty-four trillion US dollars a year, with commercial services exports adding roughly seven to eight trillion more. Put differently, world trade in goods and services, counting exports plus imports, has run at somewhere around fifty-five to sixty percent of world GDP for most of the last fifteen years. The exact figure moves with commodity prices and the dollar, which is one reason you should always check the current release rather than trusting a number you memorized.

The three largest merchandise traders are China, the United States, and Germany, with the ordering depending on whether you count goods alone or goods plus services. The United States is by a wide margin the largest services exporter. For the United States specifically, the Census Bureau reports goods and services exports of roughly three trillion dollars a year in the mid-2020s against imports somewhat under four trillion, producing a persistent goods deficit partly offset by a services surplus. Mexico and Canada trade more with the United States than any other partners do, and in 2023 Mexico passed China as the largest single source of US goods imports, a change that tells you something about the reorganization we will study in Lesson 3.

Investment is smaller in flow but arguably larger in influence. Global FDI flows run in the range of one to one and a half trillion dollars a year, an order of magnitude below trade flows, yet the OECD and UNCTAD estimate that multinational enterprises account for roughly a third of world output and something like half of world exports. That second statistic is the one to sit with. A large fraction of what customs officers record as trade between countries is actually trade within companies: a Ford engine plant in Mexico shipping to a Ford assembly plant in Michigan. Roughly a third of US goods trade is intra-firm, meaning the buyer and seller have the same ultimate parent. Those transactions have prices, but the prices are set by internal policy rather than by a market, which is why transfer pricing is a live tax controversy.

One more figure worth holding: participation in exporting is far more concentrated than people expect. In the United States, only a small percentage of firms export at all, and among those that do, a few hundred very large firms account for the great majority of export value. Exporting is not a normal business activity that most companies do a bit of. It is an activity that a small number of unusually productive firms do a lot of, a fact whose theoretical explanation we will meet in Lesson 5.

Key idea: World trade runs at roughly fifty-five to sixty percent of world GDP, multinationals account for around a third of output and half of exports, and exporting is concentrated in a small minority of unusually productive firms.

The gravity model: the most reliable pattern in the field

Suppose I tell you nothing about two countries except their economic size and how far apart they are, and ask you to predict how much they trade. You can do surprisingly well. The gravity model, named for its resemblance to Newton's law, says that trade between two countries rises roughly in proportion to the product of their economic sizes and falls roughly in proportion to the distance between them. Written loosely: trade flow is proportional to GDP of country A times GDP of country B, divided by distance between them.

Empirically this works well enough that economists call it one of the most robust relationships in the discipline. A common estimate is that doubling the distance between two countries roughly halves the trade between them, after controlling for size. Try it against your intuitions. The United States trades enormously with Canada, a moderately sized economy next door, and modestly with Australia, a similar economy on the other side of the planet. Germany's largest trade partners are almost all within a day's drive.

Distance in the gravity model is not only kilometers. Estimates consistently show extra resistance from crossing a border at all, beyond what physical distance explains. The Canadian economist John McCallum found in 1995 that Canadian provinces traded roughly twenty times more with each other than with comparably distant and sized American states. That number was later revised down substantially by better methods, but the qualitative finding survived: national borders reduce trade a great deal even between rich, friendly, contiguous countries speaking mostly the same language. Shared language, shared colonial history, shared currency, and trade agreements all show up as reductions in effective distance.

Why should a business student care about a regression? Because gravity is a discipline against wishful market selection. A team that has fallen in love with a distant, unfamiliar, mid-sized market is fighting the strongest pattern in the field. Sometimes that is correct, and you should be able to say why your case beats the base rate.

Key idea: Trade rises with economic size and falls with distance in a relationship so robust it functions as a base rate, and borders themselves add resistance beyond physical distance.

Reading trade statistics without being fooled

Now the part that separates a careful reader from a careless one. Conventional trade statistics record the gross value of a good every time it crosses a border. If a Korean firm ships a five hundred dollar display panel to China, and China assembles it into a device exported to the United States for eight hundred dollars, the world records five hundred dollars of Korean exports and eight hundred dollars of Chinese exports, thirteen hundred dollars in total, for a device with eight hundred dollars of final value. The five hundred dollars is counted twice. When production is chopped into stages spread across countries, gross trade statistics inflate and misattribute.

The correction is trade in value added, an approach developed jointly by the OECD and the WTO. It asks how much value each country actually contributed. Applied to consumer electronics, the results are striking. Studies of the iPhone in its early generations found that although each unit added its full wholesale value to China's measured exports to the United States, the value actually captured by Chinese assembly operations was on the order of a few dollars per phone, with the rest accruing to component suppliers in Japan, Korea, Taiwan, Germany, and the United States, and to the American firm that designed and marketed it. A bilateral deficit computed on gross flows can therefore attribute to one country value created in six others.

Three more cautions worth internalizing:

• Bilateral balances are close to meaningless as a scorecard. You run a large deficit with your grocery store and a large surplus with your employer. Neither is a problem. A country's overall current account balance reflects its saving and investment decisions far more than its trade policy.
• Services are undercounted. There is no customs officer stamping a software license or a consulting hour. Services trade is estimated through surveys and is widely believed to be understated, which matters because rich countries tend to run services surpluses.
• Imports and exports are recorded on different bases. Exports are typically valued free on board at the port of departure while imports may include cost, insurance, and freight, so the world as a whole appears to import more than it exports, which is impossible.

Key idea: Gross trade statistics double-count intermediate goods and misattribute value in global supply chains, so value-added accounting, not bilateral balances, tells you where value is actually created.

Try it

A German machine tool maker asks which of four markets to enter first: Poland (GDP about 800 billion dollars, 500 kilometers away, EU member), Brazil (GDP about 2.1 trillion, 9,500 kilometers, no EU agreement in force), Japan (GDP about 4.2 trillion, 9,000 kilometers, EU-Japan agreement in force), and Austria (GDP about 500 billion, 400 kilometers, EU member). Rank them by what gravity alone predicts, then name one factor gravity ignores that could reorder the list.

Answer: Gravity multiplies size and divides by distance, so Poland and Austria dominate: Poland has 800 divided by 500, Austria 500 divided by 400, both far above Japan's 4,200 divided by 9,000 and Brazil's 2,100 divided by 9,500. Between the two near markets, Poland's larger size gives it the edge. Japan outranks Brazil on size and slightly shorter distance, and the trade agreement lowers effective distance further. Factors gravity ignores: the composition of demand, since a country with heavy machinery-using industry may buy far more machine tools than its GDP suggests; competitive intensity, since the nearby market may already be saturated by domestic rivals; and tariffs and standards, which act like extra distance and are why the Japan agreement matters.

Common misconceptions

• "International business is basically exporting." Exporting is one mode. Licensing, franchising, contract manufacturing, joint ventures, and wholly owned subsidiaries all cross borders with very different control and risk profiles.
• "A bilateral trade deficit means we are losing." Bilateral balances mostly reflect who makes what. Overall current account balances reflect national saving and investment, not the toughness of trade negotiators.
• "China exported eight hundred dollars of value in that phone." Gross trade statistics record the full value at each crossing. Value-added accounting shows assembly often captures a small fraction, with the rest spread across many countries.
• "Globalization means distance no longer matters." The gravity relationship is as strong as ever. Distance, borders, and language still shape trade powerfully.
• "Most companies export." In the United States and most rich countries only a small minority of firms export, and a small number of very large exporters account for most of the value.
• "FDI and portfolio investment are the same." FDI implies a lasting management interest, conventionally ten percent of voting shares or more. Portfolio investment is ownership without control and behaves very differently in a crisis.

Recap

• International business is defined by its problem, crossing borders, and therefore pulls in economics, law, finance, politics, and logistics together.
• Cross-border activity ranges from arms-length trade through licensing and contracting to full foreign direct investment, differing in control and capital at risk.
• World trade runs at roughly fifty-five to sixty percent of world GDP, and multinationals account for about a third of output and around half of exports.
• A large share of measured trade is intra-firm, which is why transfer pricing is a tax battleground.
• The gravity model, trade rising with size and falling with distance, is the field's most reliable empirical pattern and a useful check on wishful market selection.
• Gross trade data double-count intermediates; the OECD and WTO value-added approach shows where value is really created, and bilateral balances are a poor scorecard.

Sources

1. World Trade Organization. (2024). World trade statistical review. WTO. wto.org
2. Organisation for Economic Co-operation and Development. (2023). Trade in value added (TiVA). OECD. oecd.org
3. U.S. Census Bureau. (2025). Foreign trade statistics. U.S. Department of Commerce. census.gov
4. United Nations Conference on Trade and Development. (2024). World investment report. UNCTAD. unctad.org
5. Encyclopaedia Britannica. (2024). International trade. britannica.com

The big picture

In 1913 a Londoner with money could telephone an order for goods from anywhere on earth, invest in a Buenos Aires railway or a Malayan rubber plantation without asking anyone's permission, and travel to almost any country without a passport. Wheat moved from Chicago to Liverpool for a fraction of what it had cost fifty years earlier. Capital crossed borders under a shared monetary standard. Roughly sixty million people had left Europe for the Americas and Australasia in the preceding century. By any reasonable measure, the world economy was profoundly integrated.

Within about thirty years, essentially all of that had been destroyed. Trade collapsed, capital controls became universal, migration was slammed shut, currencies became inconvertible, and two world wars and a depression intervened. It took until roughly the 1970s for trade as a share of world output to climb back to where it had been in 1913.

Hold that in your head, because it is the single most useful fact in this course. Globalization is not a ratchet. It is not a natural process that only goes one way. It is a policy choice made repeatedly by governments, enabled by technology, and reversible under political pressure. Anyone who tells you the current system is permanent, or that its unraveling would be unprecedented, has not looked at the twentieth century.

This lesson walks the first wave, the collapse, the postwar rebuild, and the shipping container, which did as much for trade as any treaty ever signed.

The first wave, 1870 to 1914

Three technologies did the heavy lifting. Steamships with compound and then triple-expansion engines cut ocean freight costs dramatically, and the Suez Canal, opened in 1869, took thousands of miles off the Europe-to-Asia route. Railways pushed interior grain and ore to ports. The telegraph, transatlantic from 1866, collapsed the time to send a price quotation from weeks to minutes, which is what made genuinely integrated commodity markets possible. Refrigerated shipping from the 1880s let Argentina and New Zealand sell meat in London.

The economic historians Kevin O'Rourke and Jeffrey Williamson measured the result the right way: not by trade volumes, which can rise for many reasons, but by price convergence. If markets are truly integrating, the price gap between two places for the same good should shrink. Their figures for wheat show the Liverpool price standing roughly fifty-eight percent above the Chicago price in 1870 and roughly sixteen percent above it by 1913. Similar convergence appeared in cotton, iron, and other commodities. That is integration you can measure.

Money and people moved too. Under the classical gold standard, most major currencies were fixed to gold and therefore to each other, which made long-term cross-border lending far less risky. Britain ran current account surpluses and exported capital on a scale no country has matched since as a share of its own economy. And migration was extraordinary by modern standards: the foreign-born share of the United States population hovered near fourteen percent in the decades around 1900, roughly where it stands today, but with almost no legal barriers to entry for Europeans.

Now the part that gets left out of the celebration. This integration rested substantially on empire and coercion. Much of the era's trade ran along colonial lines under rules the colonies did not write. After the abolition of slavery, roughly a million and a half indentured laborers were shipped from India alone to plantations from the Caribbean to Fiji, under contracts that were formally voluntary and practically not. Colonial administrations pushed cash-crop cultivation in ways that raised export volumes and lowered food security. China's ports were opened by war. When you read a passage praising the openness of 1913, note whose openness it was.

And even at the peak, the backlash was already running. As cheap American and Russian grain arrived, Germany passed protective tariffs in 1879, France followed with the Meline tariff of 1892, and the United States restricted immigration progressively, culminating in the quota acts of 1921 and 1924. Openness produced concentrated losses, concentrated losses produced politics, and politics produced tariffs. That sequence will return in Module 2 with modern data.

Key idea: The first wave of globalization was real and measurable in converging prices, it was built substantially on empire and coerced labor, and its political backlash began well before the shooting started in 1914.

The collapse, 1914 to 1945

The war shattered the system, and the attempt to rebuild it made things worse. Britain returned to gold in 1925 at the prewar parity, a decision Keynes attacked at the time as forcing painful deflation on British industry, and he was right. When the American slump began in 1929, countries on the gold standard could not use monetary policy to fight it without abandoning the peg, a trap the economist Barry Eichengreen called golden fetters.

Into that came the Smoot-Hawley Tariff Act, signed in June 1930, which raised US duties to among the highest in the country's history, with average rates on dutiable imports climbing toward sixty percent. More than a thousand economists signed a petition urging President Hoover to veto it. He did not. Canada, France, Spain, and others retaliated. World trade contracted by roughly two thirds in nominal value between 1929 and 1933, a spiral the economist Charles Kindleberger rendered in a famous inward-spiraling chart.

Here is where a careful course differs from a slogan. Smoot-Hawley is often described as having caused the Great Depression. Most economic historians do not believe that. US imports were only about four to five percent of national output at the time, so the direct arithmetic is too small to explain a contraction of that magnitude; the monetary collapse, bank failures, and the gold standard's constraints do far more explanatory work. What Smoot-Hawley did do was deepen and internationalize the crisis, poison trade relations, trigger retaliation against US exporters, and demonstrate that tariff escalation is easy to start and hard to stop. That is a serious enough charge without inflating it, and being able to state the honest version is what makes your argument credible when you later criticize a tariff.

The 1930s finished the job: competitive devaluations, exchange controls, bilateral clearing arrangements, and imperial preference blocs after the Ottawa conference of 1932. By 1939 the world economy was a set of walled gardens.

Key idea: Between 1914 and 1945 the open world economy was destroyed by war, monetary rigidity, and tariff retaliation, and Smoot-Hawley worsened and spread the Depression without being its principal cause.

Bretton Woods and the GATT

In July 1944, with the war still on, delegates from forty-four nations met at a hotel in Bretton Woods, New Hampshire, with an explicit goal: design a postwar economic order that would not repeat the 1930s. The intellectual duel was between John Maynard Keynes for Britain, who proposed an international clearing union with its own unit of account and pressure on surplus as well as deficit countries, and Harry Dexter White for the United States, whose more modest and dollar-centered plan prevailed because the United States held the money.

Three institutions came out of the effort, though only two came out of the conference:

• The International Monetary Fund, to lend to countries with temporary balance of payments problems so they would not have to devalue or impose controls in a panic, and to oversee a system of exchange rates pegged but adjustable, with the dollar convertible to gold at thirty-five dollars an ounce.
• The International Bank for Reconstruction and Development, now the core of the World Bank Group, initially for European reconstruction and soon for development lending.
• The International Trade Organization, negotiated later in the Havana Charter of 1948, which would have handled trade, employment, commodity agreements, and restrictive business practices. The US Senate never took it up, and the ITO died.

The trade gap was filled by an improvisation. Twenty-three countries had already signed the General Agreement on Tariffs and Trade in 1947 as an interim tariff-cutting deal pending the ITO. The interim arrangement ran the world trading system for forty-seven years. Its core rules are still the architecture you must know:

• Most-favored-nation treatment: a concession granted to one member must be granted to all. The name is misleading; it means non-discrimination, not favoritism.
• National treatment: once imported goods have cleared customs, they must be treated no worse than domestic goods on taxes and regulation.
• Bound tariffs: members commit to ceiling rates in published schedules, which turns tariffs from a lever pulled at will into a negotiated commitment.
• Reciprocity through rounds: countries trade concessions in packages, because a package can pass a legislature when a unilateral cut cannot.

It worked. Across eight rounds, average tariffs on manufactured goods in industrial countries fell from roughly twenty percent after the war to under five percent by the mid-1990s. What it did not do is equally important: agriculture was substantially carved out, textiles and clothing were governed by a separate quota regime from 1974 to 2004, and services and intellectual property were not covered at all. Those gaps set the agenda for the Uruguay Round and the WTO, which is Lesson 3.

The monetary half unraveled first. Financing the Vietnam War and domestic programs while running the world's reserve currency produced more dollars abroad than the United States could plausibly convert to gold, a contradiction the economist Robert Triffin had predicted. In August 1971 President Nixon suspended convertibility. After a failed attempt to reset the pegs, the major currencies were floating by 1973, and the 1976 Jamaica agreement made that official.

Key idea: Bretton Woods produced the IMF and World Bank and, after the ITO failed, the GATT, whose non-discrimination rules and negotiating rounds cut industrial tariffs from about twenty percent to under five percent while leaving agriculture, textiles, and services outside the system.

The box that changed everything

Treaties get the credit; a steel box deserves a large share of it. Before containers, cargo moved as break bulk: barrels, sacks, crates, and bales, loaded piece by piece by gangs of longshoremen. A ship could spend as long in port as at sea. Pilferage was routine. Malcom McLean, an American trucking executive who was frustrated watching his trailers unloaded item by item, bought a shipping company and on 26 April 1956 sent a converted tanker, the Ideal-X, from Newark to Houston carrying fifty-eight truck bodies. In Marc Levinson's history of containerization, the loading cost worked out to about sixteen cents a ton against roughly five dollars and eighty-six cents a ton for conventional break-bulk loading on a comparable ship.

Two things had to happen before that saving became a system. First, standardization: through the 1960s the International Organization for Standardization settled on the corner fittings and the twenty and forty foot lengths that let any box ride any ship, train, or chassis anywhere. Second, complementary investment: new deepwater terminals with gantry cranes, which is why the piers of Manhattan and the docks of central London died while Elizabeth, New Jersey and Felixstowe grew. Ports are not interchangeable, and containerization redrew the map of which cities mattered.

How big was the effect? A study by Daniel Bernhofen, Zouheir El-Beyrouty, and Richard Kneller compared country pairs before and after each adopted container facilities and found increases in bilateral trade over subsequent years that were very large, and in their estimates larger than the trade effects of the free trade agreements signed over the same period. You should read that result as suggestive rather than final, since separating a technology's effect from everything else changing is genuinely hard. But the direction is not in doubt. When you cut the cost and the uncertainty of moving a physical thing, you make it possible to split production into stages located wherever each stage is cheapest, which is the precondition for everything in Module 6.

Jet air freight, satellite communication, and eventually the internet did the same for high-value and information-intensive goods. The economist Richard Baldwin describes two unbundlings: cheap transport separated production from consumption, and then cheap communication separated the stages of production from each other. The container is the hinge between them.

Key idea: Containerization collapsed the cost and uncertainty of moving goods, and by making it cheap to split production into stages located in different countries it enabled global value chains as decisively as any trade agreement.

Try it

Rank these four events by how much they plausibly increased world trade, and defend your ordering in three sentences: (a) the Suez Canal opening in 1869; (b) the 1947 GATT; (c) the standardization of the shipping container in the 1960s; (d) the 1925 British return to gold at prewar parity.

Answer: A defensible ranking is (c), (b), (a), then (d), with (d) negative. Containerization plus standardization cut the physical cost of trade by an order of magnitude and made multi-country production possible at all, an effect estimated to exceed that of the trade agreements of the same era. The GATT is second because cutting industrial tariffs from about twenty percent to under five percent removed a very large policy barrier, though it left agriculture and services out. Suez was a genuine but geographically narrower gain, shortening one major route. The 1925 gold decision belongs at the bottom because it forced deflation on Britain and helped set up the monetary rigidity that made the 1930s trade collapse worse. Reasonable people put (b) first; the point is that policy and technology both bind, and neither alone explains the history.

Common misconceptions

• "Globalization is new." By trade and capital measures, the world of 1913 was comparably integrated. Trade as a share of output did not regain its 1913 level until roughly the 1970s.
• "Globalization only moves forward." The 1914 to 1945 reversal was near-total. Integration is a policy choice that can be and has been undone.
• "Smoot-Hawley caused the Great Depression." Most economic historians disagree. US imports were too small a share of output. It deepened and internationalized the crisis rather than causing it.
• "Most-favored-nation status is a special favor." It is the opposite: a promise not to discriminate, extending to all members whatever you grant to one.
• "The GATT was a temporary agreement, so it did not matter much." The provisional stopgap governed world trade for forty-seven years and cut industrial tariffs by roughly three quarters.
• "The first wave was a straightforwardly good era of free exchange." Much of it ran through colonial rule, indentured labor, and gunboat-opened ports. The economics and the coercion are the same story.

Recap

• From about 1870 to 1914, steamships, railways, the telegraph, refrigeration, and the gold standard produced measurable price convergence across oceans, along with mass migration and enormous capital flows.
• That integration rested substantially on empire, indenture, and coerced market opening, and it generated a tariff and immigration backlash before 1914.
• War, the golden fetters of the interwar gold standard, Smoot-Hawley and the retaliation it triggered, and 1930s autarky cut world trade by roughly two thirds in nominal value.
• Bretton Woods created the IMF and the World Bank; the ITO failed in the US Senate, and the improvised GATT ran trade for forty-seven years on non-discrimination, bindings, and rounds.
• The Bretton Woods peg collapsed in 1971 to 1973 as the dollar's convertibility became untenable, leaving today's floating system.
• Containerization, standardized in the 1960s, cut cargo handling costs by an order of magnitude and made the multi-country production chain economically possible.

Sources

1. World Trade Organization. (n.d.). The GATT years: From Havana to Marrakesh. WTO. wto.org
2. International Monetary Fund. (n.d.). The IMF and the World Bank. IMF. imf.org
3. Encyclopaedia Britannica. (2024). Bretton Woods Conference. britannica.com
4. Encyclopaedia Britannica. (2024). Smoot-Hawley Tariff Act. britannica.com
5. Federal Reserve History. (2013). Nixon ends convertibility of US dollars to gold. Federal Reserve Bank of St. Louis. federalreservehistory.org
6. Wikipedia contributors. (2025). Containerization. Wikipedia. en.wikipedia.org

The big picture

If you started a career in international business in 1995 and retired in 2015, you could have gone the whole way believing that the direction of travel was fixed: more agreements, lower tariffs, longer supply chains, more countries joining in. Every year the map got more connected. Trade grew faster than output for most of two decades. A generation of managers built careers on the assumption that the cheapest place to make something was the right place to make it.

Anyone who started in 2016 has had a different education. Since then: a British vote to leave the European Union, an American withdrawal from a Pacific trade pact its own negotiators had written, tariffs on hundreds of billions of dollars of Chinese goods that then survived a change of administration, an appellate court at the WTO that stopped functioning, a pandemic that emptied shelves, a war that cut Europe off from Russian gas, export controls on advanced semiconductors, and enormous subsidy programs on both sides of the Atlantic.

The temptation is to pick a story: either globalization is ending, or nothing has really changed. Both are wrong, and the accurate account is more interesting. Trade has not shrunk; it has been re-routed and re-priced, and the criteria firms use to choose locations have shifted from cost alone toward cost plus resilience plus political acceptability. This lesson gives you the history and the vocabulary to say what has actually changed.

What the WTO added

The Uruguay Round ran from 1986 to 1994 and was the most ambitious negotiation the trading system had attempted. It concluded at Marrakesh in April 1994, and on 1 January 1995 the World Trade Organization replaced the provisional GATT with a permanent institution. Four things were genuinely new:

• Services came in under the General Agreement on Trade in Services, covering banking, telecommunications, transport, and professional services, sectors the GATT had never touched.
• Intellectual property came in under TRIPS, which required members to protect patents, copyrights, and trademarks to specified minimum standards. This was the most contested addition, and it produced a serious fight over access to medicines that led to the Doha Declaration on TRIPS and Public Health in 2001.
• Agriculture and textiles were brought inside the rules. The textile and clothing quota regime was phased out over ten years, ending in 2004, a change that reshuffled global apparel production toward China and Bangladesh almost immediately.
• Binding dispute settlement, with panels and a standing Appellate Body, and crucially with the ability of a winning member to impose retaliatory tariffs if the loser did not comply. Under the GATT, a losing country could block adoption of the ruling against it. Under the WTO it could not.

That last change is the one to understand. It converted trade rules from diplomacy into something much closer to law, which is exactly why it later became the target. Membership grew to over one hundred sixty economies covering the overwhelming majority of world trade.

Key idea: The WTO added services, intellectual property, agriculture, and textiles to the rulebook and, most consequentially, replaced blockable GATT panels with binding dispute settlement backed by authorized retaliation.

China joins, and the hyperglobalization decade

China acceded to the WTO on 11 December 2001 after fifteen years of negotiation. It is tempting to explain what followed by tariff cuts, but that is not quite the mechanism. The United States had already been granting China normal tariff treatment, but only through an annual renewal that Congress debated and could refuse. In 2000 the US granted permanent normal trade relations, and accession locked it in.

The economists Justin Pierce and Peter Schott argued that this removal of policy uncertainty is the key. An American retailer or manufacturer considering a large, irreversible investment in Chinese sourcing had previously faced the risk that tariffs might snap back to the punitive rates on the books. Once that risk disappeared, the investment made sense. Their evidence shows US manufacturing employment falling sharply after 2001 precisely in the industries where the gap between the threatened rate and the applied rate had been largest. It is a beautiful piece of empirical work and a useful lesson: what a policy does to uncertainty can matter more than what it does to prices.

The years from roughly 1990 to 2008 are sometimes called hyperglobalization. World trade as a share of output climbed from around forty percent to a peak above sixty percent. Production fragmented into global value chains: Richard Baldwin describes a first unbundling, in which cheap transport separated production from consumption, and a second unbundling, in which cheap communication and coordination separated the stages of production from each other. A shirt could be designed in Italy, cut from Indian cotton spun in Pakistan, sewn in Vietnam, and sold in Ohio, with each stage coordinated in near real time.

Then it flattened. After the 2008 financial crisis, trade as a share of world output stopped rising and has moved sideways since, a pattern journalists labeled slowbalization. The reasons are mixed and mostly not about protectionism: Chinese producers substituted domestic inputs for imported ones as their capabilities improved, which mechanically shortens chains; commodity prices fell from their peak; and the easy gains from fragmenting production had largely been taken.

Key idea: China's 2001 accession mattered largely by removing the annual risk that US tariff treatment would be revoked, and the hyperglobalization it capped gave way after 2008 to a long plateau driven more by exhausted gains and Chinese input substitution than by policy.

Doha, and the dispute settlement crisis

The WTO's negotiating function has largely stopped working. The Doha Development Agenda, launched in November 2001, was supposed to deliver gains especially for developing countries in agriculture. It broke down at Cancun in 2003 and again in 2008, and has never been revived. The reasons are structural: the WTO negotiates by consensus among more than one hundred sixty members with genuinely opposed interests on agricultural subsidies, and the single-undertaking approach means nothing is agreed until everything is.

That is not the same as saying the WTO does nothing. The Trade Facilitation Agreement, concluded at Bali in 2013 and in force from 2017, streamlines customs procedures and is estimated to have produced real cost savings, especially for smaller exporters. An expanded Information Technology Agreement eliminated tariffs on a large list of tech products. In 2022 members agreed a fisheries subsidies agreement, the organization's first with an environmental sustainability core.

The bigger institutional wound is in dispute settlement. Under the WTO system, a member could appeal a panel report to a standing seven-member Appellate Body. Beginning in 2017 the United States blocked appointments to fill vacancies, and in December 2019 the Appellate Body fell below the three members needed to hear a case. It has not functioned since.

You should be able to state both sides here, because this is the kind of question you will be asked to have a view on. The American case, made by administrations of both parties, is that the Appellate Body exceeded its mandate: it treated its own past reports as binding precedent when the agreements do not say so, routinely blew through the ninety-day deadline, allowed departing members to keep working on pending appeals, and above all wrote rules on anti-dumping and subsidy remedies that the United States says it never agreed to in negotiation. The defenders' case is that binding, independent adjudication was the central achievement of the WTO, that a body interpreting ambiguous text will inevitably fill gaps, and that a member unhappy with rulings should negotiate a fix rather than dismantle the court. In the meantime a group of members including the EU and China set up an interim appeal arbitration arrangement among themselves, and other losing parties simply appeal into a void, which suspends the case indefinitely.

Key idea: The WTO's negotiating arm stalled with Doha while its judicial arm was disabled after 2019 by a US blockade of Appellate Body appointments, a dispute in which the complaint about judicial overreach and the defense of binding adjudication are both serious.

The turn after 2016, and what the evidence shows

The policy shift is easiest to follow as a sequence. In June 2016 the United Kingdom voted to leave the European Union. In January 2017 the United States withdrew from the Trans-Pacific Partnership, whose remaining members proceeded without it as the CPTPP. In 2018 the United States imposed tariffs on steel and aluminum under a national security provision, and then tariffs on a widening list of Chinese goods reaching roughly three hundred fifty billion dollars of imports, with Chinese retaliation aimed heavily at American agriculture. NAFTA was renegotiated into the USMCA, in force from July 2020, with tougher automotive rules of origin and a labor value content requirement.

What is notable is what happened next. A change of administration in 2021 did not reverse the tariffs. Instead the toolkit expanded: sweeping export controls on advanced semiconductors and the equipment to make them from October 2022, the CHIPS and Science Act with roughly fifty-two billion dollars for domestic semiconductor manufacturing, and the Inflation Reduction Act with domestic-content-linked clean energy credits that annoyed allies enough to prompt European countermeasures. The European Union built its own program of a Green Deal Industrial Plan, a Critical Raw Materials Act, a foreign subsidies regulation, and a carbon border adjustment mechanism. In 2025 the United States imposed a further round of broad tariffs under emergency economic powers, and the legal authority for them was challenged in litigation. Because rates in this area now change on a timescale of months, treat any specific number you read, including in this course, as a starting point to verify against the current schedule.

Learn the vocabulary, because it is used loosely:

Term	Means	Origin and note
Reshoring	Bringing production back to the home country	Often announced, less often completed at scale
Nearshoring	Moving production to a nearby country	Mexico and Eastern Europe are the main beneficiaries
Friend-shoring	Concentrating supply among trusted allies	Popularized by US Treasury Secretary Janet Yellen in 2022
De-risking, not decoupling	Reducing dependence in critical sectors while keeping broad trade	European Commission framing from 2023, adopted by the G7
Small yard, high fence	Tight controls on a narrow set of strategic technologies	US national security framing; the argument is over how small the yard stays

Is it working? The honest answer is partly, and not in the way the headlines suggest. The share of US goods imports coming directly from China has fallen noticeably, and Mexico overtook China as the largest single source in 2023. But researchers examining what happened underneath find a great deal of re-routing rather than separation: exports to the United States from Vietnam and Mexico grew fastest precisely in the product categories where Chinese exports to the US fell, Chinese investment in and inputs to those countries rose, and goods from these connector economies with higher Chinese content became relatively more expensive. In other words, the supply chain grew a new link rather than being cut. Meanwhile the IMF has estimated that a severe fragmentation of the world economy into blocs could cost several percent of global output over the long run, with the largest losses falling on low-income countries that would lose access to technology and markets.

Key idea: Since 2016 policy has shifted from liberalization toward tariffs, export controls, and industrial subsidies, but the measured result so far looks more like re-routing through connector economies than genuine decoupling, at real cost.

Try it

A US electronics firm sees its imports from China fall from 70 percent of its component spend to 35 percent over four years, with Vietnam rising from 5 percent to 38 percent. The CFO announces that the company has halved its China exposure. What three questions would you ask before believing that, and what would each reveal?

Answer: First: what are the Vietnamese suppliers made of? If the Vietnamese plants are Chinese-owned or import most of their inputs from China, the country of origin changed and the dependency did not. Second: what happened to unit cost? Research on connector economies finds that goods re-routed through third countries with high Chinese content carry higher prices, so a shift that looks like risk reduction may partly be a cost increase. Third: what is the exposure at tiers two and three? A firm typically knows its direct suppliers and not its suppliers' suppliers, which is exactly where a single-source chokepoint hides. The general lesson is that country-of-origin statistics measure the last border crossed, not where the value or the risk lives.

Common misconceptions

• "The WTO is just the GATT with a new name." It added services, intellectual property, agriculture, and textiles, and replaced blockable panels with binding dispute settlement.
• "China's accession worked mainly by cutting tariffs." US tariff treatment was already normal but renewable annually; making it permanent removed the uncertainty that had deterred irreversible sourcing investment.
• "The Appellate Body was shut down by one country acting on a whim." The US complaints about precedent, deadlines, and remedy jurisprudence were specific and long-standing across administrations, even if you conclude the remedy was disproportionate.
• "Trade has collapsed since 2016." World trade has grown in value; what stopped rising, back in 2008, was trade as a share of output.
• "Falling imports from China prove decoupling." A large share of the shift shows up as re-routing through Vietnam and Mexico with Chinese ownership and inputs behind it.
• "De-risking and decoupling mean the same thing." De-risking is explicitly a narrower claim: reduce dependence in critical sectors, keep trading in everything else. Whether policy stays inside that line is the actual argument.

Recap

• The WTO, from 1995, extended the rules to services, intellectual property, agriculture, and textiles, and made dispute rulings binding with authorized retaliation.
• China's 2001 accession, combined with permanent normal trade relations from the US, mattered largely by removing policy uncertainty that had deterred sourcing investment.
• Trade as a share of world output rose steeply until 2008 and has been flat since, for reasons more structural than political.
• The Doha Round failed, though the Trade Facilitation Agreement and the fisheries subsidies agreement show the institution can still deliver.
• The Appellate Body has been non-functional since December 2019, with serious arguments on both sides and an interim arbitration workaround in use.
• Since 2016 the policy mix has shifted to tariffs, export controls, and industrial subsidies, with evidence pointing more toward re-routing through connector economies than true decoupling.

Sources

1. World Trade Organization. (n.d.). What is the WTO? WTO. wto.org
2. World Trade Organization. (n.d.). Dispute settlement. WTO. wto.org
3. Office of the United States Trade Representative. (2024). Section 301 investigations. USTR. ustr.gov
4. International Monetary Fund. (2023). Geoeconomic fragmentation and the future of multilateralism. IMF Staff Discussion Note. imf.org
5. Encyclopaedia Britannica. (2024). World Trade Organization. britannica.com

The big picture

Two complaints about trade are made constantly, usually by different people, and they cannot both be true. The first: we cannot compete with countries where workers earn three dollars an hour. The second, heard in those same low-wage countries: we cannot compete with countries whose factories are ten times more productive. If cheap labor were decisive, all production would move to the poorest country on earth. If high productivity were decisive, all production would move to the richest. Neither has happened, and it is not because of tariffs.

The resolution is two hundred years old and still the single most useful idea in this course. In 1817 David Ricardo showed that what determines who exports what is not how good you are at making something in absolute terms, but what you give up to make it. The mathematician Stanislaw Ulam once challenged Paul Samuelson to name one proposition in the social sciences that was both true and non-obvious. Samuelson said comparative advantage. Nobody has produced a better answer since.

This lesson works the idea numerically, all the way from productivity data to the wages that make it happen. Do the arithmetic yourself as we go; the idea does not really land until you have watched a country that is worse at everything end up exporting something and getting richer.

Absolute advantage is not the question

Two countries, Ambia and Borland. Two goods, wheat and cloth. Each country has 100 worker-days of labor available. Here is what one worker-day produces:

One worker-day produces	Wheat (tons)	Cloth (bolts)
Ambia	6	3
Borland	1	2

Ambia has absolute advantage in both goods: six tons of wheat a day against Borland's one, three bolts of cloth against Borland's two. Ambian workers are simply more productive. Adam Smith's version of the argument would stop here and conclude that Ambia should make everything and Borland nothing, which is obviously wrong as a description of the world and would leave Borland with nothing to sell.

Ricardo's move was to ask a different question: what does each country give up? To make one bolt of cloth, an Ambian worker-day is diverted from producing six tons of wheat to producing three bolts. So three bolts cost six tons of wheat, which means one bolt costs two tons of wheat. In Borland, one worker-day makes either one ton of wheat or two bolts of cloth, so two bolts cost one ton of wheat, and one bolt costs half a ton of wheat.

Opportunity cost	Cost of 1 bolt of cloth	Cost of 1 ton of wheat
Ambia	2 tons of wheat	0.5 bolts of cloth
Borland	0.5 tons of wheat	2 bolts of cloth

Now the picture inverts. Cloth is expensive in Ambia, costing two tons of wheat, and cheap in Borland, costing half a ton. Wheat is cheap in Ambia and expensive in Borland. Comparative advantage belongs to whoever gives up less: Ambia in wheat, Borland in cloth. Notice that comparative advantage is unavoidably relative. Nobody can have a comparative advantage in everything, because if you give up less to make wheat you must by definition give up more to make cloth. This is why a country cannot be competed out of every market, an assertion you will hear regularly.

Key idea: Absolute advantage compares productivity levels; comparative advantage compares opportunity costs, and because opportunity costs are relative, every country has a comparative advantage in something.

Watching the world gain

Here is the cleanest possible demonstration, and you can check every step on paper. Start anywhere and move labor at the margin. Shift one worker-day in Ambia from cloth to wheat: Ambia gains 6 tons of wheat and loses 3 bolts of cloth. Now shift two worker-days in Borland from wheat to cloth: Borland loses 2 tons of wheat and gains 4 bolts of cloth.

Add it up. World wheat: plus 6, minus 2, equals plus 4 tons. World cloth: minus 3, plus 4, equals plus 1 bolt. The world has more of both goods, using exactly the same total labor, with no new technology and nobody working harder. That surplus is the gains from trade, and it comes from nothing more mysterious than moving each task to whoever gives up less to do it.

Now scale it up. Suppose in isolation each country splits its labor evenly:

No trade (autarky)	Wheat days	Cloth days	Wheat produced	Cloth produced
Ambia	50	50	300	150
Borland	50	50	50	100
World			350	250

Now let Borland specialize completely in cloth, and let Ambia move 25 days from cloth into wheat, keeping 25 days in cloth:

With specialization	Wheat days	Cloth days	Wheat produced	Cloth produced
Ambia	75	25	450	75
Borland	0	100	0	200
World			450	275

World wheat is up 100 tons and world cloth is up 25 bolts, from the same 200 worker-days. Nothing has been invented. Labor simply stopped doing work it was comparatively bad at.

Now divide the surplus. Trade needs a price, called the terms of trade. It has to sit between the two countries' opportunity costs, that is, between half a ton of wheat per bolt and two tons of wheat per bolt, or neither side would bother. Suppose the world price settles at one ton of wheat per bolt of cloth. Ambia exports 75 tons of wheat and imports 75 bolts of cloth:

Consumption	Wheat, no trade	Cloth, no trade	Wheat, with trade	Cloth, with trade
Ambia	300	150	375	150
Borland	50	100	75	125

Check the arithmetic. Ambia produced 450 wheat and 75 cloth, exported 75 wheat, imported 75 cloth: 375 wheat and 150 bolts. It consumes 75 more tons of wheat than before and exactly as much cloth. Borland produced 200 bolts, exported 75, received 75 tons of wheat: 75 wheat and 125 bolts. It consumes 25 more tons of wheat and 25 more bolts. Both countries are strictly better off, and no one was made worse off at the national level. That is the theorem.

Key idea: Reallocating labor toward each country's comparative advantage raises world output of both goods, and any terms of trade between the two countries' opportunity costs splits the surplus so that both gain.

Where the wages come in

Students almost always ask the same excellent question at this point. Ambia is better at everything. Why does an Ambian firm not simply undercut Borland in cloth too? The answer is wages, and working it out is what makes the model click.

Put money on the goods. Suppose the world price of wheat is ten dollars a ton and, at our one-to-one terms of trade, cloth is ten dollars a bolt. An Ambian wheat worker produces 6 tons a day, so their output is worth sixty dollars, and competition among Ambian employers for wheat workers will push the Ambian wage toward sixty dollars a day. A Borlandish cloth worker produces 2 bolts a day, worth twenty dollars, so the Borlandish wage tends toward twenty dollars a day.

Now check whether either country can invade the other's industry. An Ambian cloth worker makes 3 bolts a day, worth thirty dollars, but must be paid the Ambian wage of sixty dollars. The firm loses thirty dollars a day. Ambia cannot profitably make cloth despite being better at it than Borland. Meanwhile a Borlandish wheat worker makes 1 ton a day, worth ten dollars, against a wage of twenty. That fails too. Each country is priced into the industry where its relative advantage lies.

Sit with the implication for a moment, because it dissolves both of the complaints we started with. Borland's low wage is not an unfair trick; it is the market's way of making a less productive country competitive in something. And Ambia's high wage is not a handicap; it is the reward for higher productivity, and Ambian workers earn three times what Borlandish workers earn. The country with low wages has low wages because its productivity is low. Wages track productivity over the long run, which is why high-wage countries like Germany and the Netherlands remain enormous exporters and why low-wage countries do not sweep every market.

Key idea: Wages adjust to productivity, so a less productive country becomes competitive through lower wages while a more productive country pays higher wages, and each ends up exporting where its comparative advantage lies.

What the model assumes, and what it does not claim

A model you cannot criticize is a model you do not understand. The Ricardian setup makes strong assumptions: two countries, two goods, one factor of production, constant returns to scale, no transport costs, full employment, and labor that moves freely between industries at home but not across borders. Relax those and things change. Transport costs and tariffs can wipe out a small comparative advantage. Increasing returns can lock in whoever got there first, which is the subject of Lesson 5. And in the real world, labor does not slide effortlessly from a closing textile mill into an expanding software firm, which is the subject of Lesson 6 and the most important qualification in this course.

Notice carefully what the theorem does and does not say. It says the gains to the winners in a country exceed the losses to the losers, so that compensation is theoretically possible. It does not say compensation happens. A country can be better off in aggregate while a large number of specific people are permanently worse off, and pointing at the aggregate gain does not answer them. Anyone who cites comparative advantage as if it settled trade policy has skipped the distributional question, which is a separate and much harder one.

Two further honest caveats. First, comparative advantage is not fixed. South Korea's comparative advantage in 1965 was in wigs, plywood, and cheap textiles; today it is in semiconductors, ships, and batteries. Opportunity costs shift as countries accumulate capital, skills, and institutions, which is the opening for the industrial policy argument in Lesson 7. Second, even Paul Samuelson, the model's greatest expositor, published a paper in 2004 showing that technical progress by a trading partner in the good you export can reduce your gains from trade. The model is robust, not sacred.

Is there evidence? Yes, and it is decent. Studies from Donald MacDougall in 1951 through modern work by Arnaud Costinot and Dave Donaldson, who tested Ricardian predictions against agronomic data on what crops each piece of land could actually grow, find that countries do tend to export goods in which their relative productivity is high. The relationship is not tight enough to predict any single trade flow, but the direction is real.

Key idea: The Ricardian model assumes away transport costs, increasing returns, unemployment, and painful adjustment, and it establishes that gains exceed losses in aggregate without establishing that anyone actually compensates the losers.

Try it

Two countries, Northland and Southaven, each with 200 worker-days. One worker-day in Northland yields 10 units of software or 5 units of furniture. One worker-day in Southaven yields 2 units of software or 4 units of furniture. Compute each country's opportunity costs, identify comparative advantage, and state the range of terms of trade at which both would trade.

Answer: In Northland, 1 unit of software costs 5 divided by 10, or 0.5 units of furniture, and 1 unit of furniture costs 2 units of software. In Southaven, 1 unit of software costs 4 divided by 2, or 2 units of furniture, and 1 unit of furniture costs 0.5 units of software. Northland gives up less furniture to make software, so it has comparative advantage in software; Southaven has comparative advantage in furniture. Trade is mutually beneficial at any price of software between 0.5 and 2 units of furniture. Note that Northland has absolute advantage in both goods, exactly as Ambia did, and it still pays both sides to trade. To confirm, take an even autarky split as the baseline: Northland makes 1,000 software and 500 furniture, Southaven 200 software and 400 furniture. Now let Southaven specialize fully in furniture, producing 800, while Northland puts 140 days into software and 60 into furniture, producing 1,400 software and 300 furniture. World output rises from 1,200 software and 900 furniture to 1,400 and 1,100. At a price of 1 furniture per software, Northland exports 200 software for 200 furniture and consumes 1,200 software and 500 furniture, gaining 200 software; Southaven consumes 200 software and 600 furniture, gaining 200 furniture. Both are strictly better off from the same total labor.

Common misconceptions

• "A country can lose its comparative advantage in everything." Impossible by construction. Giving up less in one good means giving up more in another.
• "Comparative advantage means competing on low wages." Wages follow productivity. High-wage Germany and the Netherlands are among the world's largest exporters per capita.
• "Trade only helps if you are the more productive country." In the worked example, the less productive country gained more in proportional terms.
• "The theory proves free trade is always the right policy." It proves aggregate gains exceed aggregate losses, leaving distribution, adjustment, security, and dynamic effects entirely open.
• "Comparative advantage is permanent." Korea moved from wigs and plywood to semiconductors in a generation. Opportunity costs change with capital, skills, and institutions.
• "Countries fully specialize in the real world." They rarely do, because of transport costs, decreasing returns, product variety, and the fact that most trade is within industries rather than between them.

Recap

• Absolute advantage compares productivity; comparative advantage compares opportunity cost, which is what actually determines trade patterns.
• Moving one worker-day in each country toward its comparative advantage raised world output of both goods, which is the gain from trade in its simplest form.
• In the worked case, specialization raised world wheat from 350 to 450 tons and world cloth from 250 to 275 bolts using the same labor.
• Any terms of trade strictly between the two opportunity costs, here between 0.5 and 2 tons of wheat per bolt, leaves both countries better off.
• Wages adjust to productivity, which is why a less productive country stays competitive and a more productive country pays more without losing its export industries.
• The model assumes away transport costs, increasing returns, unemployment, and adjustment friction, and it proves that compensation is possible, not that it occurs.

Sources

1. Greenlaw, S. A., & Shapiro, D. (2023). International trade. In Principles of economics (3rd ed.). OpenStax, Rice University. openstax.org
2. Encyclopaedia Britannica. (2024). Comparative advantage. britannica.com
3. World Trade Organization. (n.d.). The case for open trade. WTO. wto.org
4. Wikipedia contributors. (2025). Comparative advantage. Wikipedia. en.wikipedia.org

The big picture

Ricardo tells you why trade happens. He does not tell you what actually gets traded, and the moment you look at the data his model starts to strain. The largest trade flows in the world are not between rich countries and poor ones swapping aircraft for coffee. They are between rich countries swapping remarkably similar things: Germany sells cars to France and buys cars from France, the United States and Canada trade auto parts across the same bridge in both directions all day, and Japan and South Korea exchange chemicals and machinery that a non-specialist could not tell apart.

Ricardo also assumed one factor of production, so his model has nothing to say about who inside a country wins or loses. That question, which dominates trade politics, needs a different tool. And Ricardo treated countries as the actors, when in reality trade is done by firms, and it turns out that a startlingly small number of firms do almost all of it.

This lesson supplies the three theories that fill those gaps, and then puts a number on what all of it is worth. The number will probably strike you as small. Understanding why it is reported that way, and what it does and does not include, is part of the lesson.

Factor endowments, and the first honest bad news

Eli Heckscher and Bertil Ohlin, writing in Sweden in the early twentieth century, asked where comparative advantage comes from. Their answer: from what a country has a lot of. A country abundant in low-skilled labor will find labor cheap and will therefore have a comparative advantage in labor-intensive goods. A country abundant in capital, skills, or arable land will export goods that use those intensively. This is the Heckscher-Ohlin model, and its predictions are intuitive: Bangladesh exports garments, Saudi Arabia exports oil, the United States exports aircraft, software, pharmaceuticals, soybeans, and corn.

Now for the result that matters most in this course. Wolfgang Stolper and Paul Samuelson worked out, in 1941, what opening to trade does to factor incomes inside a country. The Stolper-Samuelson theorem says that trade raises the real return to a country's abundant factor and lowers the real return to its scarce factor. Read that again with a rich country in mind. If the United States is abundant in capital and skills and scarce in low-skilled labor relative to its trading partners, then trade should be expected to raise returns to capital and skilled labor and to reduce the real wages of less-skilled workers.

Notice where that finding comes from. It is not a complaint by trade's critics. It is a theorem proved by one of the most distinguished defenders of free trade, sitting inside the standard model. Economists have known for eighty years that trade produces losers within countries and that the losses are systematic rather than random. Anyone who is surprised by the political salience of trade has not read their own textbook carefully.

Two qualifications keep this honest. First, the theorem's magnitude in practice is contested; through the 1990s many economists concluded that technology, not trade, was the main driver of rising skill premiums in rich countries, and the balance of the evidence shifted only later. Second, the pure Heckscher-Ohlin model assumes factors move freely between industries at home. Relax that and you get the specific factors model, in which capital and skills stuck in one industry bear the shock. That version explains actual trade politics much better: steelworkers and steel company owners lobby together, against each other's supposed class interests, because in the short run both are stuck in steel.

One more thing to know, because it shows the field correcting itself: in 1953 Wassily Leontief tested the model on US data and found that American exports were, contrary to expectation, less capital-intensive than the goods the US imported. The Leontief paradox provoked decades of refinement, including treating skilled labor as human capital and allowing technology to differ across countries.

Key idea: Heckscher-Ohlin ties comparative advantage to factor abundance, and the Stolper-Samuelson theorem inside that model predicts that trade systematically lowers the real return to a country's scarce factor, which in rich countries means less-skilled labor.

Why similar countries trade similar things

Heckscher-Ohlin predicts that trade should be largest between countries that are most different. In reality, the biggest flows are between countries that are most alike. The explanation, developed by Paul Krugman in the late 1970s and early 1980s and central to his Nobel citation, drops two of Ricardo's assumptions.

The first is constant returns to scale. Real manufacturing has large fixed costs: a car platform costs billions to develop, and the cost per unit falls the more you build. That gives an advantage to producing a lot of a few models rather than a few of many. The second is that goods are identical. Consumers value variety, and a French buyer who wants a Volkswagen is not satisfied by being told a Renault is available.

Put those together and you get intra-industry trade. Each country's firms specialize in particular models, produce them at scale for the whole market, and export them to each other. France and Germany both end up with more car models available at lower cost per unit than either could manage alone. The gains here have nothing to do with comparative advantage; they come from scale and variety. This also implies something Ricardo would not have predicted: which country ends up making which model can be partly accidental, a matter of who got there first, since increasing returns lock in an early lead.

You can measure how much of a country's trade in a category is intra-industry with the Grubel-Lloyd index: take one minus the absolute difference between exports and imports, divided by their sum. Work two cases. Suppose Germany exports thirty billion euros of cars to France and imports twenty-two billion. The absolute difference is eight, the sum is fifty-two, so the index is one minus eight over fifty-two, about 0.85. Very high: this trade is overwhelmingly two-way. Now suppose the United States exports about a tenth of a billion dollars of apparel to Bangladesh and imports about seven billion. The difference is 6.9 and the sum is 7.1, giving an index of about 0.03. That trade is essentially one-way, exactly as Heckscher-Ohlin predicts for two countries with very different endowments.

The index tells you which theory is doing the work, and it also predicts political heat. One-way trade concentrates losses in an identifiable domestic industry. Two-way trade shuffles market share among firms in the same industry on both sides, which hurts less and organizes less.

Key idea: Increasing returns and consumer demand for variety generate large two-way trade in similar goods between similar countries, with gains from scale and variety rather than comparative advantage, measurable by the Grubel-Lloyd index.

It is firms that trade, and only a few of them

The third correction is the most useful for a business student. Countries do not trade; firms do. And the firm-level data, assembled by Andrew Bernard, Bradford Jensen, Stephen Redding, and Peter Schott among others, show a lopsided picture. Only a small minority of firms export at all. Among those that do, a small number of very large exporters account for the overwhelming majority of export value. Exporters are systematically bigger, more productive, more capital-intensive, and pay higher wages than non-exporters in the same industry, and this is true before they start exporting, not only after.

Marc Melitz built the theory that explains it. Exporting carries fixed costs: market research, distribution agreements, regulatory compliance, adapting the product, learning the customs paperwork. Only firms productive enough to earn back those fixed costs will pay them. When a country opens to trade, three things happen at once: the most productive domestic firms expand into exporting, the least productive shrink or exit as imports arrive, and the average productivity of the surviving industry rises because market share moved toward better firms.

That last effect is a gain from trade that Ricardo's model cannot see, and empirically it appears to be large. It is also brutal at the level of the individual firm, and it explains something managers observe directly: an import shock does not reduce every domestic firm proportionally. It kills the weakest and grows the strongest, which is why industry-level statistics can look mild while particular towns lose their only plant.

Layer on top of this the fact that production itself is split across borders. Gene Grossman and Esteban Rossi-Hansberg call it trade in tasks: it is no longer goods that are traded so much as the stages of making them. The value along that chain is famously unequal. Studies of early iPhone generations found that although each unit added its full wholesale value to China's export statistics, the assembly operations in China captured only a few dollars per phone, while design, chip fabrication, displays, software, and brand captured the rest. The Taiwanese entrepreneur Stan Shih described the pattern as a smile curve: value added is high at the research and design end and high again at the branding and service end, and lowest in the middle, where physical assembly happens. Countries and firms that want to move up the curve are trying to escape the middle of the smile.

Key idea: A small minority of unusually productive firms do nearly all exporting, opening to trade reallocates market share toward them and raises average productivity, and along fragmented value chains the assembly stage typically captures the least value.

So what is trade actually worth?

Now the number. Economists have tried to estimate what a country would lose if it stopped trading entirely, which is the cleanest way to define the gains. The answers cluster in a range that surprises people.

Christian Broda and David Weinstein estimated that the growth in the variety of imported goods available to Americans between 1972 and 2001 was worth roughly 2.6 percent of GDP, an effect that conventional price indexes miss entirely because they do not count new goods properly. Arnaud Costinot and Andres Rodriguez-Clare, surveying the quantitative models, put total US gains from trade at somewhere around two to eight percent of GDP depending on the model's assumptions, with the higher figures coming from models that include intermediate inputs and firm selection.

Two things to understand about those numbers. First, they look small because the United States is enormous and diversified; the same exercise for Belgium, Singapore, or Ireland yields far larger figures, because a small economy that cannot trade cannot make most of what it uses. Second, they are static estimates. The dynamic effects, technology diffusion, competitive pressure on domestic monopolies, and access to inputs that make new industries possible, are plausibly larger and much harder to measure credibly, so responsible economists leave them out of the headline number rather than guess.

There is a further finding that deserves to be better known. Pablo Fajgelbaum and Amit Khandelwal examined how the gains from trade are distributed across income levels within countries. Because poorer households spend a larger share of their budgets on traded goods, especially food and manufactured basics, while richer households spend more on locally produced services, trade lowers the cost of living more for the poor. In their estimates, moving a typical country from full autarky to observed trade levels raises the real income of the poorest tenth by far more in percentage terms than that of the richest tenth. Note the direction carefully, and note the qualification: this is about consumer prices, not about wages or jobs. The very same population can gain as consumers and lose as workers, which is precisely the tension in the next lesson.

Key idea: Credible estimates put the total gains from trade for a large diversified economy at roughly two to eight percent of GDP, far more for small economies, with consumer price gains falling disproportionately to lower-income households.

Try it

Country A exports 4.0 billion dollars of pharmaceuticals to Country B and imports 3.4 billion. It also exports 0.2 billion dollars of footwear to Country C and imports 5.8 billion. Compute the Grubel-Lloyd index for each and say which trade relationship you would expect to generate more political conflict, and why.

Answer: For pharmaceuticals, the difference is 0.6 and the sum is 7.4, so the index is one minus 0.6 divided by 7.4, about 0.92: almost entirely intra-industry trade explained by scale and variety, with firms on both sides both exporting and importing. For footwear, the difference is 5.6 and the sum is 6.0, so the index is one minus 5.6 divided by 6.0, about 0.07: essentially one-way trade explained by factor endowments. Expect the footwear relationship to generate far more political conflict. One-way trade produces concentrated, geographically identifiable losses in a domestic industry whose workers and owners share an interest in protection, exactly the specific factors story. Two-way pharmaceutical trade shuffles share among firms that are themselves exporters, so the domestic industry is internally split on trade policy and lobbies less coherently.

Common misconceptions

• "Most trade is rich countries buying from poor countries." The largest flows are between rich countries, much of it two-way trade in similar goods.
• "The idea that trade hurts some workers is a critique from outside economics." The Stolper-Samuelson theorem, proved in 1941 inside the standard model, predicts exactly that.
• "Class explains trade politics." The specific factors model fits better: in the short run, owners and workers in the same threatened industry lobby together.
• "Any firm can export if it wants to." Exporting has substantial fixed costs, and firms that export were already larger and more productive before they started.
• "Gains of two to eight percent of GDP are trivial." That is the estimate for a very large diversified economy, permanently, every year. For small open economies the figure is a multiple of it.
• "If the poor gain from cheaper imports, they cannot be trade's losers." The same household can gain as a consumer and lose as a worker. Both findings are well supported and they do not cancel.

Recap

• Heckscher-Ohlin explains comparative advantage by factor abundance, and Stolper-Samuelson predicts trade lowers the real return to the scarce factor within each country.
• The specific factors model explains why trade politics is organized by industry rather than by class.
• Increasing returns and demand for variety generate large intra-industry trade between similar countries, measured by the Grubel-Lloyd index.
• A small minority of unusually productive firms do nearly all exporting, and opening to trade reallocates share toward them, raising average productivity.
• Production is split into tasks across countries, and along the smile curve assembly captures the least value while design and branding capture the most.
• Gains from trade for a large economy are estimated at roughly two to eight percent of GDP, far more for small ones, with consumer gains skewed toward lower-income households.

Sources

1. World Trade Organization. (2023). World trade report. WTO. wto.org
2. Organisation for Economic Co-operation and Development. (2023). Global value chains and trade. OECD. oecd.org
3. Greenlaw, S. A., & Shapiro, D. (2023). The gains from trade. In Principles of economics (3rd ed.). OpenStax, Rice University. openstax.org
4. Wikipedia contributors. (2025). Heckscher-Ohlin model. Wikipedia. en.wikipedia.org
5. Wikipedia contributors. (2025). New trade theory. Wikipedia. en.wikipedia.org

The big picture

Last lesson ended with a number: the gains from trade for a large economy, roughly two to eight percent of GDP, spread thinly across three hundred million people and showing up mostly as prices that are lower than they would otherwise have been. Nobody experiences that as an event. You do not walk into a store and feel a two percent gain.

Now consider the other side of the ledger. In Hickory, North Carolina, furniture plants that had employed thousands closed within a few years of each other. In Galax, Virginia, in Dalton, Georgia, in dozens of small manufacturing towns across the Midwest and the South, a single dominant employer or an entire local industry disappeared inside a decade. That is not thin and it is not abstract. Everyone in town experiences it, and they experience it at once.

Both of those things are true. That sentence is the entire lesson, and most public argument about trade consists of insisting on one half of it. What follows is the research on the second half, presented at full strength, because a course that gave you the gains without the losses would be propaganda, and one that gave you the losses without the gains would be the same thing pointed the other way.

The question, and how it was answered

US manufacturing employment held roughly steady near seventeen to nineteen million workers from the late 1970s through 2000, and then fell by about a third in a decade, to around eleven and a half million by 2010. Two explanations compete: automation, since manufacturing output kept rising while employment fell, and import competition, since Chinese exports to the United States exploded over exactly that period. Both are real. The research question is the proportions.

The difficulty is that correlation is nearly useless here. If US imports from China rose while US manufacturing employment fell, maybe imports caused the job losses, or maybe a US demand shift caused both, or maybe declining US industries were losing anyway and imports filled the gap. David Autor, David Dorn, and Gordon Hanson designed a way around this, and the design is worth understanding because it teaches you how modern evidence is built.

Their approach has two parts. First, they used the fact that American local labor markets, defined as commuting zones, started with very different industry mixes. A commuting zone specialized in furniture, textiles, or toys was mechanically far more exposed to rising Chinese import competition than one specialized in aerospace or pharmaceuticals, purely because of what it happened to be making in 1990. Second, and crucially, they instrumented US imports from China with Chinese import growth in eight other high-income countries. The logic: the part of Chinese export growth that also shows up in Australia, Japan, and Western Europe is driven by China's own reforms and productivity growth, not by anything happening in American demand. That isolates the supply-side push.

Do not skip past that. The instrument is what converts a suggestive correlation into a credible causal estimate, and it is why these findings have held up under scrutiny from economists who would have been happy to knock them down.

Key idea: The China Shock studies compare US local labor markets with different initial industry mixes and use Chinese export growth to other rich countries as an instrument, isolating China's supply-side push from US demand conditions.

What they found

The results, published in 2013 and extended since, are consistent and uncomfortable:

• Manufacturing employment fell more where exposure was higher. Import competition from China explains roughly a quarter of the aggregate decline in US manufacturing employment between 1990 and 2007. Counting supplier industries through input-output linkages, the later review put total job losses through 2011 at roughly two to two and a half million.
• The losses did not stay in manufacturing. Exposed commuting zones saw declines in overall employment and in labor force participation, not just reshuffling between sectors. Wages fell in non-manufacturing work in those places too, because displaced workers competed for the remaining jobs.
• Transfers rose, and did not come close to covering the losses. Exposed areas saw higher take-up of unemployment insurance, disability insurance, food assistance, and medical transfers. The increase in government payments per capita was real but a fraction of the fall in earnings per capita.
• Individual workers, tracked over sixteen years, did not recover. Following people rather than places, workers in exposed industries accumulated substantially lower earnings, with the damage concentrated among those who had been lower paid to begin with. Higher-earning workers moved out of manufacturing more successfully; lower-earning ones churned between employers, spent more time out of work, and drew more disability.
• People did not move. The textbook remedy for a local shock is migration. It barely happened. Population responses in exposed commuting zones were small and slow, which is a serious challenge to the standard adjustment story.

A 2021 follow-up asked whether time healed it. Partly. By 2019, overall employment rates in exposed places had substantially recovered, helped by a long expansion and by slower in-migration rather than by manufacturing returning. Manufacturing employment itself did not come back, the new jobs paid less on average, and other indicators of local wellbeing remained worse. A separate line of work by Justin Pierce and Peter Schott found that counties more exposed to the removal of tariff uncertainty after China's accession experienced relative increases in deaths from suicide and drug overdose. Treat that specific result with appropriate caution, as its authors do, but it points in the same direction as everything else.

Then the politics. Autor, Dorn, Hanson, and Kaveh Majlesi examined congressional elections and found that trade-exposed districts became more likely to elect ideologically extreme representatives, with the direction depending on the district's demographics, and that exposure correlated with vote shifts in 2016. Italo Colantone and Piero Stanig found parallel patterns in Europe, including in the Brexit referendum and support for nationalist parties. These are harder to establish than the labor market results and you should hold them more loosely, but the body of evidence is substantial and it points the same way. Concentrated economic losses have political consequences, which is the mechanism by which the backlash of Lesson 3 was generated.

Key idea: Trade-exposed US local labor markets suffered lasting declines in employment, participation, and wages that transfers only partly offset, workers did not migrate away, and the political effects were measurable.

The counterarguments, taken seriously

A one-sided presentation would stop there. Do not stop there. Several strong objections exist, and you should be able to state them well.

Automation did more. US manufacturing output continued rising through the period even as employment collapsed, which is the signature of productivity growth, not of demand loss. Estimates attributing most of the long-run employment decline to automation are defensible. The counter is timing: productivity growth was reasonably smooth, while the employment collapse was concentrated in 1999 to 2011, which fits the trade shock better than a gradual technology trend. The most defensible position is that both mattered and that trade concentrated its damage geographically in a way automation did not.

The studies count losses without counting gains. Robert Feenstra, Hong Ma, and Yuan Xu argued that if you also count the jobs created by US export growth to China and to the rest of the world over the same period, the net national employment effect is roughly a wash. This is a genuinely important correction, and it is compatible with everything above, because the export gains landed in different places from the import losses. National neutrality with regional devastation is exactly the pattern trade theory predicts.

The shock was a transition, not a permanent state. Chinese import growth decelerated sharply after about 2011 as wages rose and the catch-up completed. The China Shock was a one-time integration of an enormous labor force into the world economy, not an ongoing process, and policy built to fight the last shock may misfire.

Consumers gained, and the poorest gained most. The findings from Lesson 5 remain in force. Research on price effects finds meaningful reductions in the cost of goods American households buy, with larger proportional benefits for lower-income households. A family that lost a plant job and pays less for clothing has experienced both, and no honest accounting nets them into a single number.

The counterfactual is not obvious. The alternative to trade with China was never a world in which those jobs stayed. It might have been Mexico, Vietnam, or automation arriving sooner. And the more interesting counterfactual is not less trade but different domestic policy, which brings us to the actual failure.

Key idea: Automation, export-side job creation, the transitional nature of the shock, and large consumer gains are all serious qualifications, and none of them eliminates the finding that losses were concentrated, local, and durable.

The adjustment that did not happen

Standard trade theory has an answer to concentrated losses: the winners gain more than the losers lose, so compensate the losers and everyone is better off. The United States had a program for exactly this. Trade Adjustment Assistance, created in 1962, offered extended income support and retraining to workers certified as having lost jobs to imports. In practice it was small, hard to qualify for, and slow. Establishing that a specific layoff was caused by imports rather than by anything else is difficult by design. Evaluations were unflattering: the major federal evaluation found participants earning no more than comparison workers years later, a result its defenders contest on the grounds that participants were more damaged to begin with. The program's authorization lapsed in 2022.

Zoom out and the picture is starker. The OECD tracks spending on active labor market policies, meaning training, job search assistance, and employment services. Denmark, Sweden, and Germany spend multiples of what the United States spends as a share of GDP. That is not a claim that the Nordic approach would have worked identically in Ohio, but it does mean the United States conducted an unusually large trade liberalization while running an unusually small adjustment apparatus, and then was surprised when adjustment did not happen.

The proposed fixes are worth knowing because you will hear them: wage insurance, which tops up the pay of a displaced worker who takes a lower-paying job, avoiding the retraining question entirely; relocation vouchers, which run into the finding that people mostly will not move; place-based policy aimed at the region rather than the worker, which economists long disliked and have warmed to as evidence of immobility accumulated; and simply strengthening universal supports so that losing a job does not also mean losing health insurance.

So where does that leave you? With two well-supported bodies of evidence and a genuine disagreement about what follows from them. Someone can accept every finding in this lesson and still favor open trade, on the grounds that the aggregate gains are real, that the alternative to trading with China was not stable employment, and that protection is a poor instrument for helping workers because it fails the ones already displaced. Someone else can accept every finding in Lesson 5 and still favor restriction, on the grounds that compensation has repeatedly been promised and not delivered, that a policy whose benefits require transfers that never arrive should be judged on its actual outcomes, and that communities have a claim that a GDP figure does not capture. The dispute is about weighting, counterfactuals, and political feasibility, not about the facts. Being able to state both positions in their strongest form is the mark of understanding this material.

Key idea: The theoretical case for open trade depends on compensating the losers, the United States liberalized on a very large scale with unusually small adjustment programs, and the resulting disagreement is about weighting and feasibility rather than facts.

Try it

A commuting zone had 40,000 workers in 1990, of whom 12,000 were in furniture and textiles. Between 1990 and 2007 it lost 5,000 manufacturing jobs. Employment in the rest of the local economy fell by 1,200 rather than rising, labor force participation fell by three percentage points, and government transfers per capita rose by 250 dollars a year while earnings per capita fell by 1,900 dollars. Which specific findings from this lesson does each of those four numbers illustrate, and which single number most challenges the standard adjustment story?

Answer: The 5,000 manufacturing losses illustrate the direct import-competition effect in a commuting zone whose 1990 industry mix made it highly exposed. The 1,200 decline in non-manufacturing employment shows the losses spilling beyond the affected sector, contradicting the prediction that displaced workers are absorbed elsewhere locally. The three-point fall in participation shows people leaving the labor force rather than being counted as unemployed, which is why unemployment rates understate the damage. The transfer-versus-earnings comparison shows compensation covering only about thirteen percent of the income loss. The number that most challenges the standard adjustment story is the fall in non-manufacturing employment, because the theory predicts that resources released from a shrinking sector are absorbed by growing ones, and here the rest of the local economy shrank too. A close second is the participation decline, since a worker outside the labor force is not adjusting at all.

Common misconceptions

• "The China Shock research shows trade was a mistake." Its authors say no such thing. The finding is that adjustment failed, that losses were concentrated, and that policy assumed a mobility that did not exist.
• "Automation explains everything, so trade did not matter." Automation matters enormously for the long trend, but it does not explain the geographic concentration or the timing of the 1999 to 2011 collapse.
• "If national employment was roughly unaffected, nobody was hurt." The export gains and the import losses landed in different places and on different people. Aggregation hides exactly what matters here.
• "Displaced workers just need to move where the jobs are." They largely did not move, and the research consistently finds migration responses far too small to do the adjusting.
• "Trade Adjustment Assistance handled this." It was small, hard to qualify for, evaluated poorly, and its authorization lapsed in 2022.
• "One side of this argument is simply ignoring the evidence." Both the aggregate gains and the concentrated losses are well established. Informed people disagree about weighting, counterfactuals, and what is politically achievable.

Recap

• The China Shock research used differences in local industry mix plus an instrument based on Chinese exports to other rich countries to identify causal effects.
• Import competition explains roughly a quarter of the 1990 to 2007 fall in US manufacturing employment, or about two to two and a half million jobs including supplier effects through 2011.
• Exposed local labor markets lost employment and participation beyond manufacturing, saw lower wages elsewhere, and received transfers covering only a fraction of lost earnings.
• Workers tracked individually did not recover, and geographic migration was far too small to serve as the adjustment mechanism theory assumed.
• Serious counterarguments exist on automation, export-side gains, the transitional nature of the shock, and consumer price benefits, and none of them erases the concentrated losses.
• Compensation is the hinge of the theoretical case for open trade; the United States liberalized heavily while spending little on adjustment, which is why the policy argument remains live.

Sources

1. Autor, D. H., Dorn, D., & Hanson, G. H. (2016). The China shock: Learning from labor market adjustment to large changes in trade. Annual Review of Economics, 8, 205-240. NBER Working Paper 21906. nber.org
2. Autor, D. H., Dorn, D., & Hanson, G. H. (2013). The China syndrome: Local labor market effects of import competition in the United States. American Economic Review, 103(6), 2121-2168. NBER Working Paper 18054. nber.org
3. U.S. Bureau of Labor Statistics. (2025). Employment, hours, and earnings: Manufacturing. BLS. bls.gov
4. Organisation for Economic Co-operation and Development. (2024). Public spending on labour markets. OECD Data. oecd.org
5. Wikipedia contributors. (2025). China shock. Wikipedia. en.wikipedia.org

The big picture

A government announces a twenty-five percent tariff on imported widgets. Within a day you will hear four confident claims: that foreign exporters will pay it, that domestic manufacturers will boom, that consumers will barely notice, and that it will save jobs. Every one of those is an empirical question with an answer, and by the end of this lesson you will be able to compute most of them on paper.

This is the most mechanical lesson in the course, and the most useful. Trade policy arguments are usually conducted with adjectives. The people who win them arrive with numbers. We will build the standard tariff diagram arithmetically, without drawing anything, so you can reproduce it in a spreadsheet or on the back of an envelope in a meeting.

One framing note before the arithmetic. Showing that a tariff has a net cost is not the same as showing a tariff is always wrong. There are recognized reasons a country might accept a net economic cost, including national security, resilience, and retaliation leverage. What the calculation does is tell you the size of the bill, so the argument can be about whether the objective is worth it rather than about whether there is a cost at all.

The instruments

Trade barriers come in more forms than most people realize, and the form matters:

• Ad valorem tariff: a percentage of the customs value, such as 2.5 percent on imported cars. Most tariffs work this way.
• Specific tariff: a fixed amount per unit, such as a set number of cents per kilogram. These bite hardest on cheap goods, which makes them quietly regressive.
• Compound tariff: both together, common in agriculture and apparel.
• Quota: a quantity limit. The price effect is similar to a tariff, but the money goes somewhere different, which is the crux of the last section.
• Tariff-rate quota: a low rate up to a threshold quantity and a punishing rate beyond it, the standard instrument in agricultural trade.
• Antidumping and countervailing duties: duties imposed after an investigation finds imports sold below normal value, or subsidized, and causing injury. These are by far the most used instruments worldwide, and the methods used to determine normal value are a permanent source of dispute.
• Safeguards: temporary protection when a surge of fairly traded imports causes serious injury, as with the US steel safeguard of 2002.
• Non-tariff barriers: product standards, conformity assessment, sanitary and phytosanitary rules, licensing, customs delay, local content requirements, and buy-national procurement rules. Now often larger obstacles than tariffs. Research on customs friction finds each additional day of delay in clearance acts like a meaningful additional tariff, which is why the Trade Facilitation Agreement mattered.

Key idea: Tariffs come in ad valorem, specific, and compound forms, quotas and tariff-rate quotas restrict quantity, antidumping duties are the most heavily used instrument, and non-tariff barriers now often exceed tariffs in effect.

Working the tariff, step by step

Take a small country, meaning one too small to move the world price. It imports widgets. The world price is 10 dollars. Domestic demand and supply, in thousands of units, are:

• Quantity demanded equals 100 minus 2 times the price.
• Quantity supplied equals 3 times the price minus 10.

Under free trade the domestic price equals the world price of 10. Demand is 100 minus 20, or 80 thousand units. Domestic supply is 30 minus 10, or 20 thousand units. Imports fill the gap: 60 thousand units.

Now impose a tariff of 5 dollars. Because the country is small, the world price does not move, so the domestic price rises by the full tariff, to 15. Recompute:

	Free trade (P = 10)	With 5 dollar tariff (P = 15)
Quantity demanded	80	70
Domestic supply	20	35
Imports	60	35

Four effects follow, and each is the area of a simple shape. All figures are thousands of dollars because quantities are in thousands.

Consumers lose. They pay 5 dollars more on everything they still buy, and they buy less. The loss is a trapezoid: average the two quantities, 80 and 70, giving 75, and multiply by the 5 dollar price rise. Consumer surplus falls by 375.

Domestic producers gain. They receive 5 dollars more on every unit they sell, and they sell more. Average the two supply quantities, 20 and 35, giving 27.5, times 5. Producer surplus rises by 137.5.

The government collects revenue. The tariff of 5 dollars on the 35 thousand units still imported gives 175.

What is left over is pure loss. Consumers lost 375. Producers gained 137.5 and the government gained 175, a total of 312.5 recovered. The difference, 62.5, is deadweight loss: value destroyed rather than transferred.

You can decompose that 62.5 into two triangles, and the decomposition tells you what actually went wrong:

• Production distortion: domestic output rose from 20 to 35 thousand units, and those extra 15 thousand units cost more to make at home than they cost to import. The triangle is one half times 15 times 5, or 37.5.
• Consumption distortion: consumers cut purchases from 80 to 70 thousand units, giving up 10 thousand units they valued above the true world cost of 10 dollars. One half times 10 times 5 is 25.

37.5 plus 25 equals 62.5. The arithmetic closes, which is a good habit to check.

Now look at the political shape of those numbers. Consumers lost 375 thousand dollars, spread across everyone who buys a widget, so an individual buyer barely notices. Producers gained 137.5 thousand dollars, concentrated on a handful of firms and their workers who notice very much indeed and will organize accordingly. Mancur Olson's logic of collective action explains trade politics better than any economic theory does: concentrated benefits mobilize, diffuse costs do not, and the outcome tilts toward protection even when the totals point the other way.

Key idea: A 5 dollar tariff in this market costs consumers 375, gives producers 137.5 and the treasury 175, and destroys 62.5 outright, split between production and consumption distortions.

Who actually pays, and what does a saved job cost?

The model assumed the domestic price rises by the full tariff. Is that what happens? The 2018 US tariffs provided an unusually clean natural experiment, and several teams studied it independently. Mary Amiti, Stephen Redding, and David Weinstein, and separately Pablo Fajgelbaum, Pinelopi Goldberg, Patrick Kennedy, and Amit Khandelwal, found near-complete pass-through into US import prices: the pre-tariff prices charged by foreign exporters barely fell, so American importers and ultimately American buyers bore essentially the whole burden. Retail pass-through was slower and less complete in the short run, since retailers absorbed part of it in margins, but the direction was unambiguous.

That result matters because the political claim was the opposite. It is worth stating precisely why economists expected it: pass-through is incomplete only when the importing country is large enough to push down the world price, and even then only partly. Which brings up the one genuine theoretical exception to the case against tariffs. A large country that reduces its demand enough to lower the world price captures a terms-of-trade gain at foreigners' expense, and there is an optimal tariff at which its own welfare peaks. Two problems: the gain comes entirely out of trading partners, so it invites retaliation that erases it, and in the 2018 episode the measured terms-of-trade gains were small.

Now the question a policymaker actually cares about. What does a saved job cost? The 2018 washing machine tariffs were studied by Aaron Flaaen, Ali Hortacsu, and Felix Tintelnot. Washer prices rose by roughly twelve percent. Remarkably, dryer prices rose by a similar amount even though dryers were not tariffed, because retailers price the pair together. Total annual cost to consumers came to roughly 1.5 billion dollars, against something like 1,800 manufacturing jobs added. That works out to well over 800,000 dollars per job per year, for jobs paying a small fraction of that. Estimates for the steel and aluminum tariffs from the Peterson Institute landed in the same order of magnitude.

Be careful how you use that figure. It is a fair answer to the claim that a tariff is a cheap way to create employment. It is not a refutation of a national security argument, which is not about jobs at all, and it does not tell you what the displaced workers of Lesson 6 were owed. Use the right number for the right claim.

Key idea: Evidence from the 2018 tariffs shows near-complete pass-through to domestic prices, and the washing machine case implies a cost of well over 800,000 dollars per manufacturing job created.

Quotas: the same price, a different recipient

Suppose that instead of a 5 dollar tariff the government simply caps imports at 35 thousand units. Work through it and you get exactly the same domestic price of 15, the same domestic production of 35, the same consumption of 70, and the same two deadweight loss triangles. One thing changes: the 175 thousand dollars that had been tariff revenue is now quota rent, the gap between the world price and the domestic price on the units that do get in. Somebody pockets it. Who depends entirely on how the licenses are allocated. If the government auctions them, it collects the same as a tariff. If it hands them out administratively, importers with licenses collect it, and the allocation process becomes worth lobbying for. If the restriction takes the form of a voluntary export restraint, in which the exporting country agrees to limit shipments, the foreign exporters keep the rent.

That last case is not hypothetical. From 1981 the Japanese government restrained auto exports to the United States under American pressure. Japanese producers, limited on quantity, did the rational thing: they shipped fewer, larger, better-equipped, higher-margin cars. Prices of both Japanese and American cars rose, estimated in the hundreds of dollars to over a thousand dollars per vehicle by the mid-1980s. The rent flowed to Japanese manufacturers rather than to the US Treasury. Cost per US job saved was estimated in the hundreds of thousands of dollars. And the restraint helped push Japanese producers upmarket into the luxury segment and to build assembly plants inside the United States, which is not what the policy was designed to do. Voluntary export restraints were prohibited under the WTO agreements for exactly these reasons.

The general rule to carry: tariffs and quotas can produce identical prices and quantities while distributing the money completely differently, so always ask who captures the rent. And note the asymmetry under growth. If domestic demand rises, a tariff still permits imports to expand, while a quota does not, so a quota becomes progressively more restrictive over time without anyone voting for it.

Key idea: A quota can replicate a tariff's price and quantity effects exactly while transferring the revenue to license holders or foreign exporters, and it tightens automatically as demand grows.

Try it

Same market: world price 10, demand equals 100 minus 2P, supply equals 3P minus 10. Now impose a 3 dollar tariff instead of 5. Compute the new price, quantities, consumer loss, producer gain, revenue, and deadweight loss, then compare the deadweight loss with the 5 dollar case.

Answer: The price rises to 13. Demand is 100 minus 26, or 74. Domestic supply is 39 minus 10, or 29. Imports are 45. Consumer surplus loss is the average of 80 and 74, which is 77, times 3, giving 231. Producer surplus gain is the average of 20 and 29, which is 24.5, times 3, giving 73.5. Revenue is 3 times 45, or 135. Deadweight loss is 231 minus 73.5 minus 135, which is 22.5. Check with triangles: production distortion is one half times 9 times 3, or 13.5; consumption distortion is one half times 6 times 3, or 9; total 22.5. Now the comparison that matters: cutting the tariff from 5 to 3, a forty percent reduction, cut the deadweight loss from 62.5 to 22.5, a sixty-four percent reduction. Deadweight loss grows roughly with the square of the tariff, because both triangles have a base and a height that each scale with the tariff. That is why very high tariffs are disproportionately destructive and why the last few percentage points of liberalization deliver much less than the first.

Common misconceptions

• "Foreign exporters pay the tariff." Studies of the 2018 tariffs found near-complete pass-through into US import prices, meaning domestic buyers bore essentially the whole burden.
• "A tariff is just a transfer, so nothing is lost." Most of it is a transfer, but the two distortion triangles are pure loss, and they grow with the square of the tariff.
• "Tariffs and quotas are basically the same thing." They can produce identical prices and quantities while sending the revenue to the treasury, to license holders, or to foreign exporters.
• "Protection is a cheap way to create jobs." The washing machine tariffs implied over 800,000 dollars of annual consumer cost per job created.
• "Tariffs are the main barrier to trade today." In most rich-country manufacturing, standards, conformity assessment, licensing, and customs friction now matter more.
• "Economics proves tariffs are always wrong." It prices them. Security, resilience, and leverage arguments are about objectives the calculation does not measure, and the large-country terms-of-trade case is a genuine theoretical exception.

Recap

• Trade instruments include ad valorem, specific, and compound tariffs, quotas and tariff-rate quotas, antidumping and countervailing duties, safeguards, and a large family of non-tariff barriers.
• In a small open economy the domestic price rises by the full tariff, and the welfare effects are four computable areas.
• In the worked example a 5 dollar tariff cost consumers 375, gave producers 137.5 and the treasury 175, and destroyed 62.5 in production and consumption distortions.
• Deadweight loss grows roughly with the square of the tariff, so high tariffs are disproportionately costly.
• Evidence from 2018 shows near-complete pass-through to domestic prices, and cost per job saved in the washing machine case exceeded 800,000 dollars a year.
• Quotas can match a tariff exactly on price and quantity while transferring the rent to license holders or, under a voluntary export restraint, to foreign exporters.

Sources

1. Greenlaw, S. A., & Shapiro, D. (2023). Restrictions on international trade. In Principles of economics (3rd ed.). OpenStax, Rice University. openstax.org
2. U.S. International Trade Commission. (2025). Harmonized tariff schedule of the United States. USITC. usitc.gov
3. Peterson Institute for International Economics. (2024). Trade and tariffs research. PIIE. piie.com
4. U.S. Customs and Border Protection. (2025). Trade. CBP. cbp.gov
5. Encyclopaedia Britannica. (2024). Tariff. britannica.com

The big picture

A client calls with a practical question. She makes electric bicycle motors in Ohio, sells into Canada and Mexico, and wants to know whether the USMCA lets her ship duty free. You would think the answer is a yes or a no. It is neither. The answer is a calculation about where her magnets, windings, and controller boards came from, run against a legal test, documented in a certification she signs under penalty of law, and the calculation may come out differently for two motors that look identical.

That is the first thing to understand about trade agreements. The headline is tariff elimination. The substance is rules about which goods qualify, and those rules are where policy actually lives. The second thing is that the agreements themselves have become the main venue for trade liberalization, because the multilateral track we followed in Lesson 3 stopped delivering. Over three hundred fifty regional agreements are in force and notified to the WTO, so nearly every member is in several.

Then, in the second half of this lesson, the argument that has moved from the fringe to the center of policy in less than a decade: whether governments should deliberately build industries rather than letting comparative advantage decide. Economists spent forty years mostly saying no. The profession's position has shifted, unevenly and with real disagreement, and you should understand why.

Depths of integration, and Viner's warning

Regional agreements come in escalating depths:

Form	Internal tariffs	Common external tariff	Factor mobility	Example
Free trade area	Eliminated	No	No	USMCA, CPTPP
Customs union	Eliminated	Yes	No	Mercosur, EU customs union
Common market	Eliminated	Yes	Labor and capital move freely	EU single market
Economic union	Eliminated	Yes	Yes, plus policy coordination	The EU with the euro area inside it

Now the analytical point that separates a serious reader from a headline reader. In 1950 Jacob Viner showed that a preferential agreement is not automatically beneficial, because it does two opposite things at once. Trade creation occurs when the agreement lets a lower-cost partner displace higher-cost domestic production, which is a genuine gain. Trade diversion occurs when the agreement lets a partner displace an even lower-cost outside supplier purely because the outsider still faces the tariff. That is a loss: the same good now costs the world more to produce, and the importing country's treasury loses the duty it used to collect.

Make it concrete. Suppose your country imports shirts, and before any agreement Vietnam supplies them at 10 dollars and a partner country at 12 dollars, both facing a 30 percent tariff, so consumers pay 13 and 15.60 respectively and buy Vietnamese. Now sign an agreement with the partner. Vietnamese shirts still cost 13 with duty; partner shirts now cost 12 duty free. Buyers switch to the partner. Consumers save one dollar per shirt, the treasury loses three dollars of duty per shirt, and the world is now making shirts in the more expensive place. That is trade diversion, and whether an agreement is net positive is an empirical question about which effect dominates.

Key idea: Preferential agreements create trade by displacing higher-cost domestic production and divert trade by displacing lower-cost outsiders, so their net benefit is an empirical question rather than a given.

Rules of origin: where the policy actually lives

Because a free trade area has no common external tariff, it needs a way to stop goods entering through whichever member has the lowest outside tariff and then moving inside duty free. That is what rules of origin do: they define when a good counts as originating in the region. The tests come in three families, often combined:

• Tariff shift: the good must be classified under a different tariff heading than its non-originating inputs, meaning real transformation happened.
• Regional value content: a minimum percentage of the value must originate in the region, computed by a specified formula.
• Specified process: certain operations must occur in the region, common in textiles, where the yarn-forward rule requires the yarn itself to be regional.

The USMCA automotive rules are the clearest illustration of rules of origin used as industrial policy. To qualify duty free, a passenger vehicle must meet a regional value content threshold that was raised to 75 percent from NAFTA's 62.5, must use a high share of North American steel and aluminum, and must satisfy a labor value content requirement under which a substantial percentage of the vehicle's value comes from plants paying at least a specified hourly wage. That last provision has no precedent in earlier agreements and was designed explicitly to reduce the incentive to relocate work to low-wage Mexican plants.

The practical consequence, and the thing to tell your bicycle motor client, is that a preference has a compliance cost. Documenting origin through multiple supplier tiers takes staff, systems, and audit-ready records. Utilization rates for preferences are well below one hundred percent, and small firms in particular sometimes pay the tariff because it is cheaper than proving they do not have to. An agreement with generous coverage and impossible paperwork can deliver less than a modest agreement with simple rules.

Know the current map. The USMCA replaced NAFTA in July 2020 with tougher automotive rules, a labor dispute mechanism aimed at specific Mexican facilities, and a sunset clause requiring joint review. The CPTPP continued after the US withdrawal and later added the United Kingdom. RCEP, in force from 2022, links fifteen Asia-Pacific economies including China, Japan, and South Korea, and is the largest agreement by population and output; its tariff cuts are shallower than CPTPP's but its single set of regional rules of origin across a huge production zone is its real contribution. The African Continental Free Trade Area began trading in 2021 and aims to reduce the striking historical weakness of intra-African trade. And the European Union remains the deepest integration project anywhere, which is exactly why leaving it, as the United Kingdom did, proved so complicated: a common market involves standards, conformity assessment, and mutual recognition, not just tariffs.

Key idea: Rules of origin decide who actually benefits from an agreement, they are used deliberately as industrial policy in the USMCA automotive provisions, and compliance costs mean preferences are often unused.

Industrial policy: the argument that came back

From roughly the 1980s until recently, the mainstream economic position on industrial policy, meaning deliberate government action to build particular industries, was skeptical bordering on dismissive. The objections were serious and remain so: governments do not know which industries will succeed, subsidies attract the firms best at lobbying rather than the best at producing, protection tends to outlast its justification because beneficiaries organize to keep it, and when everyone subsidizes the same industry the result is global overcapacity and a subsidy race nobody wins. Evidence supported the skepticism. A well-known study of Japanese industrial policy by Richard Beason and David Weinstein found that Japanese support flowed disproportionately to declining sectors rather than to the electronics and autos that succeeded, which is roughly the opposite of the legend.

The case for has always had a respectable theoretical core, and it rests on market failures rather than on optimism. Learning by doing means today's costs fall with cumulative experience, so an industry that would eventually be competitive may never start. Coordination failures mean an assembly plant needs suppliers and suppliers need an assembly plant, and neither moves first. Research spillovers mean the firm that develops a technology captures only part of its value, so private investment is below the social optimum. And there are objectives outside the market entirely: a country may want domestic capacity in vaccines, semiconductors, or ammunition for reasons that have nothing to do with cost.

What changed recently is partly politics and partly evidence. A newer empirical literature, associated with Reka Juhasz, Nathan Lane, and Dani Rodrik among others, has gone back with modern methods and found cases where targeted intervention appears to have worked, including studies of South Korea's heavy and chemical industry drive in the 1970s and historical episodes of protection producing durable capability. Their own summary is measured: the evidence base is thinner than the policy enthusiasm, the successes are real but conditional, and design and accountability matter more than the decision to intervene.

The practical era arrived regardless. The CHIPS and Science Act put roughly fifty-two billion dollars behind domestic semiconductor manufacturing. The Inflation Reduction Act tied clean energy credits to domestic content and assembly, which allies experienced as discriminatory and answered with their own programs. The EU built a Chips Act, a Net-Zero Industry Act, and a Critical Raw Materials Act. China's long-running programs, of which Made in China 2025 was the most publicized, are the case that motivated much of the response. Layered on top are export controls on advanced semiconductors and manufacturing equipment, which are not industrial policy in the subsidy sense but denial policy, aimed at capability rather than at cost.

How should you evaluate a specific proposal? Ask four questions. What market failure or non-market objective is being addressed, stated precisely rather than as a vibe? Is the instrument matched to it, since a research spillover argues for research funding while a security argument argues for stockpiles or capacity requirements? Is there a sunset and a measurable test of success, or does the support continue by default? And what is the retaliation and overcapacity risk if every major economy does the same thing, since the answer is frequently a global glut and a fiscal bill. Those four questions will serve you better than a general position for or against.

Key idea: The case for industrial policy rests on learning by doing, coordination failures, research spillovers, and security objectives, the case against rests on information problems, capture, and subsidy races, and a specific proposal should be judged on the market failure named, the instrument match, the sunset test, and the retaliation risk.

Try it

Your country imports 200,000 industrial pumps a year. Before any agreement, Country X supplies them at 800 dollars and Country Y at 900 dollars, and both face a 25 percent tariff. You sign a free trade agreement with Country Y only. Compute the delivered price from each supplier before and after, say which supplier wins, and state whether this is trade creation or trade diversion and what it costs the treasury.

Answer: Before the agreement, Country X delivers at 800 plus 25 percent, or 1,000 dollars, and Country Y at 900 plus 25 percent, or 1,125. Buyers choose Country X at 1,000 dollars, and the treasury collects 200 dollars per pump, or 40 million dollars a year. After the agreement, Country X still delivers at 1,000 while Country Y delivers duty free at 900. Buyers switch to Country Y. This is trade diversion, not trade creation, because the displaced supplier was the lower-cost producer and not domestic industry. Consumers save 100 dollars per pump, or 20 million dollars a year. The treasury loses all 40 million dollars of duty. The country is worse off by roughly 20 million dollars a year, and the world now makes pumps in a place that uses 100 dollars more of real resources per unit. If instead domestic production at 950 dollars had been displaced, that would have been trade creation and a genuine gain.

Common misconceptions

• "A free trade agreement is automatically good economics." Viner showed it can divert trade from a lower-cost outsider to a higher-cost partner, which is a net loss.
• "Rules of origin are technical trivia." They determine which goods qualify, and the USMCA automotive rules were designed explicitly to shape where production happens.
• "If a preference exists, exporters use it." Utilization is well below full, because documenting origin through supplier tiers costs more than the duty for many small firms.
• "A free trade area and a customs union are the same." A customs union has a common external tariff, which is precisely why it does not need rules of origin internally.
• "Economists have always opposed industrial policy." The theoretical case from learning, coordination, and spillovers is old; what changed is the political appetite and a newer empirical literature.
• "The new empirical work proves industrial policy works." Its authors say the evidence is thinner than the enthusiasm and that design and accountability determine outcomes.

Recap

• Agreements deepen from free trade areas through customs unions and common markets to economic unions, and over three hundred fifty are in force.
• Trade creation is a gain and trade diversion is a loss, so an agreement's net effect must be established, not assumed.
• Rules of origin, using tariff shift, regional value content, and specified process tests, decide who actually benefits and impose real compliance costs.
• USMCA raised automotive regional value content to 75 percent and added a labor value content requirement, using origin rules as industrial policy.
• The main agreements to know are USMCA, CPTPP, RCEP, AfCFTA, and the EU single market, which differ in depth as much as in coverage.
• Industrial policy rests on market failures and security objectives, faces information and capture problems, and should be judged on the failure named, the instrument match, the sunset, and the retaliation risk.

Sources

1. World Trade Organization. (n.d.). Regional trade agreements. WTO. wto.org
2. Office of the United States Trade Representative. (n.d.). United States-Mexico-Canada Agreement. USTR. ustr.gov
3. Organisation for Economic Co-operation and Development. (2023). Industrial policy. OECD. oecd.org
4. International Monetary Fund. (2024). Industrial policy coverage in IMF surveillance. IMF. imf.org
5. Wikipedia contributors. (2025). Rules of origin. Wikipedia. en.wikipedia.org

The big picture

In 2022, within weeks of the invasion of Ukraine, BP announced it was giving up its roughly twenty percent stake in Rosneft and took a write-down in the region of twenty-five billion dollars. Shell exited its Russian ventures and wrote off billions more. ExxonMobil walked away from the Sakhalin project it had operated for decades. More than a thousand Western companies curtailed or ended Russian operations. Not one of those firms lacked lawyers, and not one of those investments had been made carelessly. They were, in ordinary commercial terms, good assets. What destroyed them had nothing to do with markets.

That is political risk: the possibility that a government action, a political event, or a social upheaval changes the value of your investment in ways no operating plan anticipates. It is the risk category that most reliably separates domestic business from international business, and it is the one where managers are most likely to substitute optimism for analysis.

This lesson gives you a taxonomy of political risk, the history of the worst version of it, the treaty machinery that partially protects investors and the serious criticism of that machinery, and finally the skill of reading a country's economic indicators without being misled by them. That last part sounds dry. It is the difference between a market entry memo that survives contact with reality and one that does not.

What political risk actually looks like

Sort political risk two ways at once: by mechanism and by whom it hits.

Type	What happens	Typical warning signs
Expropriation or nationalization	The state takes ownership, with or without compensation	Resource nationalism, election rhetoric, contract renegotiation demands
Creeping expropriation	Value is taken gradually through taxes, price controls, local content mandates, or license conditions	Sudden royalty increases, export restrictions, forced local partners
Currency and transfer risk	Profits cannot be converted or repatriated	Reserve depletion, parallel exchange rates, new capital controls
Contract repudiation	A state entity stops honoring an agreement	Change of government, corruption investigations into the prior deal
Political violence and war	Assets are damaged, staff endangered, operations halted	Insurgency, border disputes, sanctions escalation
Regulatory and policy risk	New rules destroy the business model without touching ownership	Sector reviews, data localization rules, sudden licensing changes
Sanctions and secondary sanctions	Your home country forbids you to operate, or penalizes you for dealing with someone who does	Escalating geopolitical conflict, entity list additions

The second cut matters as much. Macro political risk hits every foreign investor in the country, as with a coup or a currency crisis. Micro political risk targets your firm, your sector, or firms from your home country, as when a government renegotiates only mining contracts, or only with companies from a state it has fallen out with. Micro risk is the one managers systematically underestimate, because the country-level indicators look fine right up until your sector becomes the issue.

A structural point worth internalizing: your bargaining power against a host government is usually highest before you invest and falls afterward. Raymond Vernon called this the obsolescing bargain. Before the capital is committed, you can walk away and the government wants the investment. Once the mine is dug, the refinery built, and the capital sunk, you cannot move it, and the terms that looked generous at signing start to look, from the government's side, like something a previous administration gave away too cheaply. This is why extractive industries with enormous sunk costs and long payback periods have historically been the most expropriated, and why firms in those industries fight so hard for stabilization clauses, international arbitration, and financing structures that give powerful third parties a stake in the outcome.

Key idea: Political risk ranges from outright expropriation through creeping value transfer to sanctions, it can be country-wide or targeted at your firm, and the obsolescing bargain means your leverage peaks before the capital is sunk.

Expropriation, and what replaced it

The history is instructive because it shows both how bad this can get and how the rules changed. Mexico nationalized foreign oil holdings in 1938, creating Pemex. Iran nationalized the Anglo-Iranian Oil Company in 1951, an act followed by a British boycott and a 1953 coup. Chile nationalized its copper industry in 1971. A wave of nationalizations swept newly independent states through the 1960s and 1970s. In 2007 Venezuela required foreign operators in the Orinoco heavy oil belt to convert to minority stakes in state-controlled joint ventures; ConocoPhillips and ExxonMobil refused and went to international arbitration, where they eventually won awards running into the billions, collection of which proved to be another matter entirely.

Outright nationalization is rarer today, partly because it is expensive: it frightens off future investment and triggers arbitration. What replaced it is subtler. Creeping expropriation achieves the same result through instruments each of which looks defensible in isolation: a windfall profits tax, a mandated local partner, a price cap on your output, an export licensing requirement, a refusal to renew a permit, a currency rule that traps your earnings. No headline says a company was expropriated. The asset simply stops being worth what it was.

Investors have three main defenses, and none is complete. The first is contractual: stabilization clauses freezing the fiscal terms, arbitration clauses selecting a neutral forum, and structuring the investment through a jurisdiction covered by a favorable treaty. The second is political risk insurance, offered by the World Bank Group's Multilateral Investment Guarantee Agency, by national agencies such as the US International Development Finance Corporation, and by private underwriters, covering expropriation, transfer restriction, political violence, and breach of contract. The third is structural: bringing in local partners, development finance institutions, and lenders from several countries so that harming the project harms people the government would rather not harm.

Key idea: Outright nationalization has largely given way to creeping expropriation through taxes, mandates, and licensing, and investors defend themselves with stabilization and arbitration clauses, political risk insurance, and partners the host government would rather not antagonize.

Investment treaties and the argument about them

Beginning in the late 1950s, countries began signing bilateral investment treaties, and there are now on the order of twenty-five hundred to three thousand such agreements plus investment chapters in trade deals. A typical treaty promises investors from the other country fair and equitable treatment, national treatment, protection against expropriation without prompt and adequate compensation, and free transfer of funds. The enforcement mechanism is the controversial part: investor-state dispute settlement, which lets a foreign investor bring a claim directly against the host state before an international arbitral tribunal, most often under the rules of the World Bank Group's International Centre for Settlement of Investment Disputes.

The case for ISDS is straightforward. A foreign investor suing a host government in that government's own courts faces an obvious problem. Neutral arbitration substitutes for a domestic judiciary the investor cannot trust, and awards are enforceable across borders. Proponents argue this lowers the risk premium on investment in countries that most need capital.

The case against has grown strong enough to change policy. Critics argue that tribunals of private arbitrators, without appeal and with the same lawyers rotating between arbitrating and advocating, have interpreted vague standards like fair and equitable treatment expansively enough to capture ordinary regulation. They point to a regulatory chill effect in which governments soften health or environmental measures rather than risk a claim. The examples cited most often include tobacco packaging rules challenged by a cigarette manufacturer against Uruguay, where the state ultimately won, and energy sector claims against European governments over nuclear and coal phase-outs. Defenders answer that states win a substantial share of cases, that treaty texts have been rewritten to preserve regulatory space, and that chill is difficult to demonstrate empirically.

What is not in dispute is the direction of policy. The European Union terminated intra-EU bilateral investment treaties, several countries have exited the Energy Charter Treaty, some states have withdrawn from ICSID or denounced treaties outright, and newer agreements increasingly narrow the standards or replace ad hoc arbitration with standing courts. For a manager the practical implication is that the treaty protection you rely on may be less durable than the investment it protects, so verify which instrument actually covers your structure and what its termination and survival clauses say.

Key idea: Thousands of investment treaties give foreign investors direct arbitration rights against host states, the mechanism is defended as a substitute for untrustworthy courts and criticized for capturing ordinary regulation, and states are now actively narrowing or exiting these instruments.

Reading a country's numbers properly

Country assessment usually starts with economic indicators, and this is where careless readers go wrong. Six rules will keep you honest.

First, know which income measure you are looking at. Gross domestic product measures output produced inside the borders. Gross national income measures income accruing to residents, wherever earned, and the two diverge sharply where foreign-owned production or large remittance flows exist. Ireland is the standard cautionary case: multinational accounting has inflated its measured GDP so far above the income actually available to Irish residents that its own statistical office publishes an alternative aggregate.

Second, decide whether you want market exchange rates or purchasing power parity. Converting local income at market rates tells you what a country can buy on world markets, which is what you want for import demand or debt service. Converting at PPP adjusts for the fact that haircuts and housing are cheaper in poorer countries, which is what you want for living standards and for local wage comparisons. The two can differ by a factor of three or more for the same country, so a report that does not say which it uses is unusable.

Third, per capita and median are not the mean. A country with a large resource sector and a small population can look rich per capita while most residents are not. Pair any per capita figure with an inequality measure such as the Gini coefficient and, if available, a median income figure.

Fourth, income is not development. The UN Development Programme's Human Development Index combines life expectancy, schooling, and income precisely because they diverge. Countries at similar income levels can differ substantially in health and education outcomes, and for a firm deciding where to place skilled operations, the schooling component may matter more than the income figure.

Fifth, a large share of activity may not be counted at all. The International Labour Organization estimates that around six in ten workers worldwide are in informal employment, with far higher shares in many low-income countries. Informality means your market is bigger than official retail data suggests, your competitors may not be paying tax or observing labor law, and your distribution partners may keep records that no auditor would accept.

Sixth, the numbers themselves get revised, sometimes enormously. When Nigeria rebased its national accounts in 2014 to a more recent base year and better sector coverage, measured GDP rose by roughly eighty-nine percent overnight, making it Africa's largest economy without a single additional good being produced. Ghana's 2010 rebasing had a similar effect. And institutions make mistakes: the World Bank discontinued its long-running Doing Business rankings in 2021 after an investigation into data irregularities, replacing it with a redesigned Business Ready assessment. If you cited Doing Business rankings in a strategy memo written in 2019, you cited something that was later withdrawn. Check what an index measures, who produces it, and whether it still exists.

Key idea: Distinguish GDP from GNI, market rates from PPP, and averages from distributions, remember that informality and rebasing can move the picture by large margins, and treat any index as a fallible product with a methodology worth reading.

Try it

You are evaluating two countries for a consumer appliance plant. Country A has GDP per capita of 4,200 dollars at market rates, 11,800 at PPP, a Gini of 0.52, an HDI in the medium band, and estimated informal employment near 55 percent. Country B has GDP per capita of 6,800 at market rates, 9,900 at PPP, a Gini of 0.31, an HDI in the high band, and informal employment near 20 percent. Which country has the larger effective consumer market for a mid-priced appliance, and what does the gap between market and PPP figures tell you?

Answer: Country B is the better bet for a mid-priced appliance despite the smaller PPP figure, because a Gini of 0.31 with high HDI means income is spread widely enough to support a real middle market, while Country A's 0.52 implies its 11,800 PPP average sits above what most households actually have. The market-versus-PPP gap is itself informative: Country A's PPP figure is nearly three times its market figure, indicating a low domestic price level, which means local wages and local inputs are cheap in dollar terms, so Country A is the more attractive production location even if it is the weaker sales market. Country B's small gap indicates prices closer to world levels, so it will be a more expensive place to manufacture. High informality in Country A also means official retail data understates the market and that your distributors may operate partly outside the formal system, a compliance issue as much as a forecasting one. The professional answer is often to sell into B and consider producing in A, with the political risk assessment of A done carefully.

Common misconceptions

• "Expropriation is a historical problem." Outright nationalization is rarer, but creeping expropriation through taxes, mandates, and licensing is routine, and 2022 showed how fast an asset can become unrecoverable.
• "A stable country rating means my project is safe." Country ratings capture macro risk. Micro risk targets your sector or your home country's firms, and is what usually bites.
• "Our bargaining power will improve once we have invested." The obsolescing bargain says the opposite: leverage peaks before the capital is sunk.
• "Investment treaties guarantee protection." States are terminating and narrowing them, and coverage depends on your holding structure and the treaty's survival clause.
• "GDP per capita tells me the size of my market." It ignores distribution, the difference between output and resident income, the price level, and the informal economy.
• "Published indices are objective facts." The World Bank withdrew Doing Business in 2021 over data irregularities. Read the methodology and check the index still exists.

Recap

• Political risk includes expropriation, creeping expropriation, transfer restriction, contract repudiation, violence, regulatory change, and sanctions, and can be macro or targeted.
• The obsolescing bargain explains why sunk-cost industries face the most pressure to renegotiate after investing.
• Defenses include stabilization and arbitration clauses, political risk insurance from MIGA and national agencies, and partner structures that raise the political cost of interference.
• Thousands of investment treaties provide investor-state arbitration, a mechanism defended as neutral adjudication and criticized for capturing ordinary regulation, and states are now narrowing or exiting it.
• Read GDP against GNI, market rates against PPP, and averages against Gini and median, since each pair answers a different question.
• Informality covers a majority of workers worldwide, national accounts rebasing can move GDP by large margins, and indices can be withdrawn, as Doing Business was in 2021.

Sources

1. Multilateral Investment Guarantee Agency. (2025). Political risk insurance. World Bank Group. miga.org
2. International Centre for Settlement of Investment Disputes. (2025). About ICSID. World Bank Group. icsid.worldbank.org
3. World Bank. (2025). World development indicators. World Bank Group. data.worldbank.org
4. United Nations Development Programme. (2024). Human Development Index. UNDP. hdr.undp.org
5. World Bank. (2024). Business Ready. World Bank Group. worldbank.org

The big picture

A German machinery firm sells a production line to a buyer in Indonesia. The contract is four pages long, because that is normal in Germany where the civil code fills the gaps. The buyer's American lender insists on a forty-page agreement, because that is normal where courts read the document and not much else. The delivery term says FOB Hamburg, which the parties assume means the seller loads the container, though FOB is a sea freight term that does not fit container shipping and will not do what either expects if the box is damaged in the terminal. There is no arbitration clause, so if things go wrong the German firm can win a judgment in Hamburg and discover that Indonesian courts are not obliged to enforce it. Somewhere in the middle, the local agent mentions that clearing customs will require a facilitation payment, and asks whether the company would prefer not to know about it.

Every element of that paragraph is a real and common failure, and every one is avoidable with knowledge you can acquire in a single lesson: how legal traditions differ, the four instruments that make cross-border contracts work, and the anti-corruption law that has done more to change how multinationals behave than any other rule in this course. A caveat that matters: this is business education, not legal advice, and international transactions need real counsel. What you are building is the ability to spot the issue, ask the right question, and understand the answer.

Legal traditions and what they do to your contract

Common law, originating in England and spread through the British Empire to the United States, Canada outside Quebec, Australia, India, Singapore, Nigeria, and many others, builds law from accumulated judicial decisions. Precedent binds, and judges referee an adversarial contest between parties who develop the evidence themselves. Contracts tend to be long, because the parties anticipate every contingency in writing rather than rely on a background code.

Civil law, descended from Roman law through the Napoleonic and German codifications, governs most of continental Europe, Latin America, much of Africa and Asia, and Japan and South Korea in significant part. Comprehensive codes state the rules; judges apply them and take a more active role in developing the evidence, and prior decisions are persuasive rather than binding. Contracts are typically shorter because the code supplies default terms, and many civil codes impose a general duty of good faith extending even to the negotiation stage, a concept common law traditionally resisted.

Other traditions matter regionally. Islamic commercial law shapes Gulf finance, most visibly through the prohibition on interest, producing cost-plus sale and asset-backed structures that finance without an interest charge. Customary law governs land in many places, which matters to any project acquiring a site, and jurisdictions such as Quebec, Louisiana, Scotland, and South Africa are mixed. Two practical consequences: common law courts prefer damages while civil law systems more readily order actual performance, so if you need the goods rather than the money the forum matters; and US-style pretrial discovery, compelling broad document production, is close to unique and astonishes foreign counterparties.

Key idea: Common law builds from precedent and produces long, self-contained contracts, civil law applies comprehensive codes and produces shorter ones with good faith duties supplied by the code, and the differences show up concretely in remedies, discovery, and drafting length.

The four instruments that make cross-border deals work

Rather than resolving which law is better, international practice built tools to route around the question.

The CISG. The UN Convention on Contracts for the International Sale of Goods, in force since 1988 and adopted by most major trading nations, supplies a uniform sales law covering formation, obligations, risk, and remedies, applying automatically between businesses in two contracting states unless the parties opt out. Two things to know: the United Kingdom is not a party, and many American lawyers exclude it reflexively without evaluating whether it would have served their client better. Excluding it is a choice; make it deliberately.

Choice of law and choice of forum. Nearly every cross-border contract specifies which country's law governs and where disputes are resolved. These are separate choices that can point to different places, and a neutral governing law such as English or Swiss law is common where neither party will accept the other's.

Incoterms. The International Chamber of Commerce publishes standardized three-letter delivery terms, currently the 2020 edition, answering two questions precisely: who arranges and pays for each leg of transport and insurance, and at exactly what point risk of loss passes from seller to buyer. Note the distinction that catches people out: EXW puts almost everything on the buyer, DDP puts almost everything on the seller including import duties, and the sea-only terms FOB, CFR, and CIF should not be used for containerized cargo, where FCA, CPT, and CIP are correct. Risk under FOB passes when goods are on board, but a container is handed over at the terminal days earlier, leaving a gap in which neither party's insurance clearly responds.

Arbitration and the New York Convention. This is the most important item on the list and the least understood by non-lawyers. A court judgment from one country is often difficult or impossible to enforce in another, because recognition depends on patchy treaties and local doctrine. An arbitral award is different: under the 1958 New York Convention, to which over one hundred seventy states are party, courts in member states must recognize and enforce foreign arbitral awards subject only to narrow exceptions. That single treaty is why the overwhelming majority of significant cross-border contracts choose arbitration. The drafting decisions are the seat, which determines the supervising courts, the institution and rules such as the ICC or the Singapore International Arbitration Centre, the number of arbitrators, and the language.

Key idea: The CISG supplies default sales law, choice of law and forum clauses select the rules and the referee, Incoterms fix cost and risk transfer precisely, and arbitration is chosen above all because the New York Convention makes awards enforceable in over one hundred seventy countries when judgments are not.

The FCPA and what it actually forbids

The Foreign Corrupt Practices Act was passed in 1977 after a post-Watergate investigation revealed that hundreds of American companies had made questionable payments abroad, including payments by an aircraft manufacturer large enough to help bring down a Japanese prime minister. Congress treated it as both a foreign policy problem and a corrosion of securities disclosure. The statute has two distinct halves, and confusing them is the most common error.

The anti-bribery provisions prohibit corruptly offering, promising, or giving anything of value to a foreign official, party, or candidate to obtain or retain business or secure an improper advantage. Note the breadth. Anything of value is not limited to cash and has covered internships, travel, directed charitable donations, and lavish hospitality. Foreign official includes employees of state-owned enterprises, which in many countries means the hospital purchasing manager and the telecom procurement officer. And an offer counts; the payment need not be made or succeed.

The accounting provisions apply to companies with securities registered in the United States and require accurate books and records and adequate internal controls. They contain no bribery element at all, which matters enormously, because proving a payment was corrupt is hard while proving it was recorded as consulting fees when it was not is comparatively easy. A large share of enforcement actions rest wholly or partly on them.

Coverage is wide: issuers listed in the United States, domestic concerns meaning US persons and companies, and foreign persons or firms taking an act in furtherance of a corrupt scheme while in US territory, which has been read to include routing an email or a dollar payment through the country. That is how many non-American companies have ended up settling with US authorities.

There is a narrow exception for facilitating payments, small sums to expedite routine non-discretionary government action such as processing a permit the applicant already qualifies for. It is far narrower than people assume, it does not exist under many other countries' laws, and most large companies prohibit such payments outright rather than ask employees to police a shifting line. Affirmative defenses exist for payments lawful under the host country's written laws, which almost never applies, and for reasonable, bona fide promotional expenses, which is real but easily abused.

The practical risk sits with third parties. The great majority of enforcement actions involve agents, distributors, customs brokers, consultants, or joint venture partners rather than employees handing over envelopes, and a company is exposed when it knew, or consciously avoided knowing, what its intermediary was doing. Successor liability applies in acquisitions, which is why anti-corruption due diligence is now standard in cross-border deals.

Key idea: The FCPA forbids corrupt payments of anything of value to foreign officials including state enterprise employees, and separately requires accurate books and internal controls, a provision with no bribery element that carries much of the enforcement load.

Enforcement, the global regime, and building a program

The scale of resolutions is what changed corporate behavior. A German engineering conglomerate settled in 2008 for roughly eight hundred million dollars in the US plus a comparable amount in Germany. A Brazilian construction group and its petrochemical affiliate resolved a case in 2016 with global penalties in the billions. A major US investment bank resolved a Malaysian sovereign fund matter in 2020 with penalties near two point nine billion dollars, the largest FCPA resolution to date, and a European aircraft manufacturer settled coordinated US, French, and UK actions the same year for close to four billion. A telecommunications maker, a commodities trader, and a large retailer reached substantial resolutions more recently. The pattern in almost every case: local intermediaries, off-book funds, and accounting entries that disguised the payments.

The United States is no longer alone. The OECD Anti-Bribery Convention, in force since 1999 and covering more than forty countries, obliges parties to criminalize bribery of foreign public officials and created a peer review system that has driven real change in national law. The UK Bribery Act 2010 is broader than the FCPA in several respects: it covers commercial bribery between private parties, has no facilitating payments exception, and creates a corporate offence of failing to prevent bribery by an associated person, with a defense only if the organization had adequate procedures. That last provision made having a compliance program a legal requirement rather than a mitigating factor, and it reaches any organization carrying on business in the UK. France's Sapin II law of 2016 took a similar approach.

One current complication you should handle carefully. In 2025 the US executive branch directed a pause on new FCPA investigations pending revised enforcement guidelines, which were subsequently issued with narrowed priorities. The statute was not repealed, the limitations period continues to run so conduct today can be charged later under a different administration, UK, French, and Brazilian authorities enforce independently, and multilateral development banks debar firms found to have engaged in corrupt practices. Because this area is genuinely in motion, check the Department of Justice FCPA page for the current position rather than relying on any textbook, including this one.

Does corruption at least speed things up? The grease the wheels hypothesis has not held up. Work by Daniel Kaufmann and Shang-Jin Wei found that firms paying more bribes reported spending more management time negotiating with officials, not less. A bureaucrat who can extract a payment has every reason to manufacture more delays. Bribery buys you a place in a queue that exists because you pay to leave it.

A workable compliance program has recognizable parts: a risk assessment focusing effort where exposure is greatest; third-party due diligence testing an intermediary's ownership, government connections, and compensation against what the service is worth; contractual rights to audit and terminate; training aimed at the people who face the situations; a gifts policy with clear limits; accurate books, since the accounting provisions catch what the bribery provisions might not; a reporting channel people believe is safe; and documenting the rationale for payments before they are made. And know the CPI for what it is: Transparency International's index measures perceptions of experts and executives, not counts of bribes, making it a useful directional signal and a poor precision instrument.

Key idea: Multi-billion dollar resolutions, the OECD Convention, and the UK Bribery Act's failure-to-prevent offence made compliance programs mandatory in practice, and evidence indicates bribery increases rather than reduces the time firms spend dealing with officials.

Try it

Your distributor in Country X sends two requests. First, a 40,000 dollar expediting fee to release a shipment that has been held at customs for three weeks, described as standard practice. Second, a proposal to retain the brother-in-law of the customs director as a consultant at 8,000 dollars a month to advise on regulatory matters. Analyze both under the FCPA and the UK Bribery Act, and say what you would do.

Answer: The 40,000 dollar payment is not a facilitating payment under any reasonable reading. Facilitating payments are small sums for routine non-discretionary acts; forty thousand dollars to release a held shipment is neither small nor routine, and the amount alone suggests discretion is being exercised. Under the FCPA this is a probable anti-bribery violation if any part of it reaches an official, and recording it as an expediting fee would separately violate the books and records provisions. Under the UK Bribery Act there is no facilitating payment exception at all, so the answer is simply no. The consultancy is the more dangerous of the two because it is designed to look legitimate. Apply four tests: does the person have genuine relevant qualifications; is 8,000 dollars a month commensurate with services actually rendered and documented; does the timing coincide with a pending decision by the relative; and can anyone articulate the business need without reference to the family relationship? An unsatisfactory answer to any of them means this is a bribe routed through a payroll. Practically: refuse both in writing, escalate to legal and compliance, ask the distributor for the legal basis and an official receipt for the customs charge, and treat the request as a signal about this distributor. Your contract should already contain anti-corruption representations, audit rights, and a termination right for exactly this.

Common misconceptions

• "The FCPA only applies to American companies." It reaches US-listed issuers, domestic concerns, and foreign parties taking any act in furtherance while in US territory, including routing an email or payment through the country.
• "Foreign official means a government minister." It includes employees of state-owned enterprises, covering hospital purchasers, utility managers, and telecom procurement staff.
• "If we did not pay, we are fine." Offering or promising is enough, and the accounting provisions can be violated with no bribe.
• "We used an agent, so it is the agent's problem." Most enforcement actions involve intermediaries, and conscious avoidance counts as knowledge.
• "Facilitating payments are legal, so they are safe." The exception is narrow under the FCPA and absent from UK law.
• "Bribery gets things done faster." Firms paying more bribes spend more management time with officials, because delay is the product being sold.

Recap

• Common law and civil law differ in the source of rules, the judge's role, contract length, good faith duties, remedies, and discovery.
• The CISG supplies default international sales law unless excluded; the UK is not a party and many US lawyers opt out reflexively.
• Incoterms 2020 fix who pays and where risk transfers, and FOB, CFR, and CIF should not be used for containers.
• Arbitration dominates cross-border contracting because the New York Convention makes awards enforceable in over 170 states, unlike court judgments.
• The FCPA covers anything of value given to foreign officials, plus separate accounting provisions on books and internal controls that carry much of the enforcement load.
• The OECD Convention and the UK Bribery Act's failure-to-prevent offence made compliance programs effectively mandatory worldwide.

Sources

1. U.S. Department of Justice. (2025). Foreign Corrupt Practices Act. DOJ Criminal Division. justice.gov
2. U.S. Securities and Exchange Commission. (2025). Foreign Corrupt Practices Act enforcement actions. SEC. sec.gov
3. United Nations Commission on International Trade Law. (n.d.). Convention on the Recognition and Enforcement of Foreign Arbitral Awards (New York, 1958). UNCITRAL. uncitral.un.org
4. United Nations Commission on International Trade Law. (n.d.). United Nations Convention on Contracts for the International Sale of Goods. UNCITRAL. uncitral.un.org
5. Transparency International. (2025). Corruption Perceptions Index. transparency.org

The big picture

An American project manager runs her first status meeting with a newly acquired team in Osaka. She asks whether the December deadline is achievable. Everyone nods. The senior engineer says it will be difficult. She hears difficult and thinks tight but doable. In December the project is four months from done, and she is furious, because nobody told her.

Somebody told her. She did not know that in that room, from that person, in front of his manager, difficult was the strongest word available and meant no. She had a communication problem, and a textbook would call it a culture problem.

Here is where this lesson becomes unusual. That story is true to life, and it is also exactly the kind of story that trains people to think badly. Told carelessly, it produces a manager who now believes Japanese people never say no, applies that to every Japanese colleague, and is wrong most of the time. The frameworks in this lesson are the most useful and the most misused material in international business. You are going to learn them properly, which means learning simultaneously what they show and what they cannot possibly show.

Start with a definition. Culture is the set of shared values, assumptions, and practices a group learns and transmits, most of which its members cannot see because they have never encountered the alternative. Two consequences follow immediately. Culture is learned, not inherited, so it is not about ethnicity. And nations are not cultures: India, Nigeria, Belgium, and the United States each contain many, and a Mumbai investment banker may have more in common professionally with a London counterpart than with a farmer four hundred kilometers inland.

Hofstede: what the study was and what it found

Geert Hofstede was a psychologist working for IBM when the company ran employee attitude surveys across its worldwide operations between roughly 1967 and 1973. He obtained access to more than one hundred thousand questionnaires from dozens of countries and noticed that answers clustered by country in ways that persisted after controlling for job type. From that he extracted dimensions of national culture, the most influential framework in the field by an enormous margin.

Dimension	High score means	Management implication if the hypothesis holds
Power distance	Unequal power is expected and accepted	Decisions expected from the top; subordinates may not volunteer disagreement
Individualism	Ties between individuals are loose; self and immediate family first	Individual incentives and recognition land well; group-based rewards may not
Motivation toward achievement, originally labeled masculinity	Competition, achievement, and material reward emphasized	Overt competition motivates in some settings and alienates in others
Uncertainty avoidance	Ambiguity is uncomfortable; rules and structure preferred	Detailed plans and clear procedures reduce anxiety; unstructured brainstorming may stall
Long-term orientation	Perseverance and thrift over immediate results	Payback horizons and patience with slow-building investments differ
Indulgence	Gratification of desires is broadly accepted	Attitudes to leisure, spending, and workplace informality vary

The framework earned its influence honestly. It gave managers a vocabulary, it generated thousands of studies, and its predictions are sometimes borne out. Now the problems, which are severe enough that you should never cite Hofstede without them.

The sample. The data come from employees of one American multinational with a strong internal culture, in a small number of white-collar job categories, half a century ago. Whether IBM employees in Peru represent Peru is a real question, not a rhetorical one, and Hofstede's answer was that the sample was matched, so differences must reflect national culture. Brendan McSweeney's widely cited 2002 critique argued this reasoning does not survive scrutiny: matched samples control for job, not for the self-selection into a foreign multinational.

The unit. The framework equates nation with culture. Some national scores were extrapolated rather than measured. And for countries with deep internal diversity, a single score is close to meaningless.

The age. The core data predate the personal computer. Cultures change, and several dimensions have been re-measured by others with different results.

The method. Anthropologists have objected that culture is not the kind of thing that survey averages capture, and that inferring stable values from questionnaire responses is circular.

Key idea: Hofstede's six dimensions came from IBM employee surveys taken half a century ago, and the framework's usefulness as a vocabulary sits alongside serious criticisms of its sample, its equation of nation with culture, its age, and its method.

The ecological fallacy, stated plainly

This section is the most important in the lesson, and it applies to every framework here, not only Hofstede's.

A country score is an average. The ecological fallacy, named after a 1950 paper by the sociologist William Robinson, is the error of inferring something about an individual from a statistic about the group. It is not a subtle bias. It is a logical mistake, and it is the mistake almost everyone makes with these frameworks.

Make it concrete with numbers. Suppose Country A scores 90 on individualism and Country B scores 40. That looks like a chasm. But those are means of distributions that are wide and heavily overlapping. Individual variation within any country dwarfs the difference between country means on most of these dimensions. If you pick one person at random from each country, the person from the higher-scoring country is more individualist somewhat more often than not, but far from always. You have a weak probabilistic tilt, not a prediction. Now consider that you are not meeting a random person. You are meeting a specific engineer selected by her employer, educated abroad perhaps, working in an industry with its own strong norms. Every one of those filters is likely to matter more than her passport.

Hofstede said this himself, repeatedly and clearly: the dimensions describe societies, not individuals. Nearly every misuse of his work ignores the warning printed on the box.

There is a second, subtler error worth naming: attributing to culture what is better explained by structure. When a junior local employee does not contradict a visiting senior executive from headquarters, is that high power distance, or is it a junior person declining to contradict someone who influences their pay, in a language that is not their first, in front of their own boss? Most people in most countries would behave the same way in that structural position. Before reaching for a cultural explanation, check whether role, hierarchy, incentives, language, or simple self-interest explains the behavior. Cultural explanations are satisfying, unfalsifiable in the moment, and frequently wrong.

Key idea: Country scores are averages of widely overlapping distributions, so using them to predict an individual is the ecological fallacy, and behavior attributed to culture is often better explained by role, hierarchy, incentives, or language.

GLOBE, and a finding worth carrying

The GLOBE project, led by Robert House and published in 2004, was designed partly to address Hofstede's limitations. It surveyed roughly seventeen thousand middle managers in about a thousand organizations across sixty-two societies, in three industries, and used nine cultural dimensions including power distance, uncertainty avoidance, humane orientation, institutional and in-group collectivism, assertiveness, gender egalitarianism, future orientation, and performance orientation.

Its most valuable innovation is a distinction the older framework did not draw. GLOBE measured each dimension twice: once as practices, what people report their society is actually like, and once as values, what they think it should be like. And on several dimensions the two turned out to be negatively correlated. People living in high power distance societies tended to report wanting less of it. People in societies with weak future orientation reported wanting more.

Sit with that, because it overturns the assumption most cultural training rests on. You cannot infer what people want from what they are used to. A manager who concludes that employees in a hierarchical society prefer to be told what to do, and therefore never invites their input, may be delivering exactly the opposite of what those employees say they want. The safest reading is that practices tell you what the environment currently rewards, and values tell you what people would move toward if the environment changed. A newly arrived manager can sometimes change the environment.

GLOBE also studied leadership, asking which leader attributes are seen as effective across societies. Two clusters came out close to universally endorsed: charismatic or value-based leadership, meaning vision, integrity, and inspiration, and team-oriented leadership. Others varied enormously, particularly participative leadership and autonomous leadership. That combination gives a useful practical rule: integrity and a clear vision travel, while the specific mechanics of how much you consult and how independently you act do not.

You should also know that Hofstede and the GLOBE team conducted a sharp public dispute in the journals about method and interpretation. That is not a scandal; it is a reminder that these are contested academic products under active argument, not measurements of physical constants.

Key idea: GLOBE surveyed managers in sixty-two societies and measured practices separately from values, finding they often diverge, so what people are accustomed to does not tell you what they prefer, while integrity and vision are endorsed as leadership qualities almost everywhere.

Context, time, and how to use any of this

The anthropologist Edward Hall proposed a distinction that survives largely because it is practically useful. In low-context communication, meaning is carried by the explicit words: say what you mean, put it in writing, and the document governs. In high-context communication, a great deal of meaning is carried by the relationship, the setting, shared history, what is not said, and who says it. The Osaka engineer who said difficult was communicating precisely, in a register the American manager could not read.

Hall also distinguished monochronic time, in which schedules are commitments and one thing happens at a time, from polychronic time, in which relationships take priority over the clock and several conversations may run at once. Anyone who has waited alone in a meeting room for twenty minutes, or been interrupted three times mid-meeting by people who clearly had precedence, has felt the difference.

Be honest about the evidence: Hall's work was ethnographic and observational, not statistical, and the country lists attached to his categories in textbooks are far more confident than his data warrant. Treat high and low context as a lens on a specific conversation, not a label on a nation.

So how should a working manager actually use all of this? Four rules.

• Generate hypotheses, not conclusions. A framework should make you ask whether indirectness might be carrying meaning here, not conclude that it is.
• Update fast on individual evidence. The moment you have observed the actual person, the country average is obsolete. Individual data beats group data always.
• Check structural explanations first. Role, power, incentives, language fluency, and organizational culture explain more workplace behavior than nationality does.
• Ask. The single most underused technique is to say plainly that you want to make sure you are reading the situation correctly, and to ask how disagreement is usually raised on this team. Most people will tell you, and asking signals more respect than any amount of country research.

A useful practitioner adaptation is Erin Meyer's culture map, which plots eight behavioral scales such as communicating, evaluating, and deciding. Its virtue is that it insists on relative positioning: the question is never whether a culture is direct in the abstract, but whether it is more or less direct than yours. A French manager may find Americans blunt about praise and evasive about criticism at the same time. Relative framing prevents you from mistaking your own norms for neutrality, which is the most common failure of all.

Key idea: High and low context and monochronic and polychronic time are useful lenses on specific interactions rather than national labels, and frameworks are best used to generate hypotheses that individual observation immediately overrides.

Try it

You manage a software team split between Stockholm and Bengaluru. In video meetings the Bengaluru engineers rarely disagree with your proposals, while the Stockholm engineers argue freely. Your head of HR says this is a power distance difference and suggests cultural training. Before accepting that, list four alternative explanations you would test, and say how you would test one of them.

Answer: First, structural position: you are based in Stockholm, and the Bengaluru team may read itself as the junior or vendor-status site whose work is reviewed by headquarters. Second, language and channel: disagreeing precisely in a second language on a laggy video call, in front of a group, is far harder than doing it in your first language in a room. Third, meeting design: if the Stockholm engineers have already discussed the proposal informally before the call, they arrive with formed positions while the Bengaluru team is hearing it for the first time. Fourth, employment context: contract terms, visa dependency, or a recent restructuring can suppress dissent anywhere on earth. Only after all four is national culture a live hypothesis, and even then it is an average that says little about these specific engineers. To test the second and third: circulate the proposal in writing forty-eight hours ahead, invite written comments before the meeting, and ask one Bengaluru engineer privately to give you the strongest objection they have heard. If written and private disagreement flows freely, the constraint was channel and meeting design, not values.

Common misconceptions

• "Hofstede's scores tell me how a colleague will behave." That is the ecological fallacy. The scores describe societies, as Hofstede himself insisted, and within-country variation swamps between-country differences.
• "Nation equals culture." India, Nigeria, Belgium, and the United States each contain many cultures, and professional and organizational cultures often matter more.
• "People in hierarchical cultures prefer hierarchy." GLOBE's practices and values measures often diverge, with respondents in high power distance societies reporting they want less of it.
• "High-context cultures are vague." They are precise in a register that requires shared context to read. The Osaka engineer communicated exactly; the listener lacked the decoder.
• "If a colleague behaves unexpectedly, culture explains it." Role, power, incentives, language, and organizational norms usually explain more, and should be checked first.
• "These frameworks are settled science." Hofstede and the GLOBE team argued publicly about method and interpretation, and both remain contested academic products.

Recap

• Culture is learned and largely invisible to insiders, and nations contain many cultures rather than being one.
• Hofstede's six dimensions came from IBM surveys of the late 1960s and early 1970s and remain the field's common vocabulary despite serious sampling, unit, age, and method criticisms.
• The ecological fallacy makes country averages useless for predicting individuals, a limitation Hofstede stated explicitly and most users ignore.
• Behavior attributed to culture is frequently better explained by role, hierarchy, incentives, language, or organizational norms.
• GLOBE separated practices from values across sixty-two societies and found them often diverging, so accustomed practice does not reveal preference.
• High and low context communication and monochronic and polychronic time are lenses for specific interactions, and the working rules are to hypothesize, update on individual evidence, check structure first, and ask.

Sources

1. Hofstede Insights. (2025). National culture. hofstede-insights.com
2. GLOBE Project. (2024). Culture and leadership across the world. globeproject.com
3. Encyclopaedia Britannica. (2024). Culture. britannica.com
4. Wikipedia contributors. (2025). Hofstede's cultural dimensions theory. Wikipedia. en.wikipedia.org
5. Wikipedia contributors. (2025). High-context and low-context cultures. Wikipedia. en.wikipedia.org

The big picture

Three days into negotiations in Seoul, the American team is frustrated. They have covered the entire agenda twice, been to two long dinners, answered questions about their families, and are no closer to a signed term sheet than on Monday. Their counterparts seem pleasant and unhurried. On the flight home the deal lead concludes the other side was not serious.

The other side thought the negotiation went well. They had spent three days doing what they considered the substantive work: establishing whether these people could be relied on over a ten-year supply relationship. The term sheet was the easy part, to be dealt with once the important question was settled.

Neither team was irrational. They disagreed about what a negotiation is for. This lesson covers three practical situations where that kind of mismatch does real damage: negotiating a deal, sending someone to live and work abroad, and running a team that spans countries. Everything from the previous lesson still applies, especially the warning against reading a person off a country average. What follows are tendencies reported in research and practice, all of which an individual counterpart may not share, and each of which you should hold as a hypothesis to test in the room.

Negotiation: the fundamentals, then the variation

Get the universals right first, because they carry more weight than any cultural adjustment. Every negotiation has a BATNA, your best alternative to a negotiated agreement, which is the only real source of leverage anyone has. Your reservation price is the worst deal you would still accept, derived from your BATNA. The zone of possible agreement is the overlap between what you would accept and what they would accept, and if there is no overlap no amount of skill produces a deal. Distributive bargaining divides a fixed amount; integrative bargaining looks for differences in priorities that let both sides gain, which requires exchanging information about interests rather than positions. These hold in every country. A negotiator with a weak BATNA and a strong grasp of local etiquette will still lose.

Now the variation. Jeswald Salacuse surveyed negotiators across many countries and found systematic differences in what they took a negotiation to be. The dimensions worth carrying:

Dimension	One pole	Other pole
Goal	Sign a contract	Build a relationship, of which the contract is one artifact
Agreement form	Detailed and exhaustive	General principles, details worked out as circumstances develop
Building process	Bottom up: agree clause by clause	Top down: agree the principle, then derive the terms
Communication	Direct and explicit	Indirect, with meaning carried by context and by what is not said
Team organization	One leader with authority to decide	Consensus, with authority distributed and decisions ratified elsewhere
Time sensitivity	High: schedule pressure is real leverage	Low: rushing signals that you need the deal more than they do

Three of these cause most of the real damage. The first is the goal difference in the Seoul story: if you think you are drafting a contract and they think they are assessing a partner, you will misread every signal. The second is agreement form. A party accustomed to detailed contracts reads a request for flexible language as evasiveness, while a party accustomed to general agreements reads a two-hundred-page draft as distrust. Both readings are wrong and both are natural. The third is authority. In some organizations the decision is genuinely made in the room; in others, extensive consensus building happens before any meeting, so the meeting ratifies rather than decides, and pressing for a commitment on the spot achieves nothing except signaling that you do not understand the process. The remedy is not to guess. Ask, early and plainly, who needs to approve this and what their process looks like. That question is welcome nearly everywhere.

Some smaller mechanics with outsized effects. Silence is not emptiness; in many settings it is thinking, and the negotiator least comfortable with it tends to fill it by conceding. Agreement noises can mean I am following you rather than I accept. Interpreters need briefing on your terminology and objectives before the meeting, short chunks of speech to work with, and monitoring for whether they are translating or summarizing, and you should never rely solely on the other side's interpreter. Budget roughly double the time for anything interpreted, and put numbers and defined terms in writing. Finally, gifts and hospitality: what is normal courtesy in a relationship-building culture can be a legal problem under the FCPA or the UK Bribery Act, particularly when the counterpart works for a state-owned enterprise. Have a written policy, keep values modest, document the business purpose, and never let politeness make the decision.

Key idea: BATNA and the zone of possible agreement determine outcomes everywhere, while goal, agreement form, and locus of authority vary enough to cause serious misreadings, and the fix is to ask directly about the approval process rather than to guess from a country profile.

Sending people abroad, and a myth worth dismantling

Companies post employees abroad for three reasons: to fill a skill gap the local labor market cannot, to transfer knowledge and exercise control from headquarters, and to develop the individual. Those three purposes call for different people, and a firm that has not decided which one it is pursuing will select badly.

Start with cost, because it disciplines the conversation. A full expatriate package, once you add housing, schooling for children, relocation, home leave, tax equalization, and any hardship or cost-of-living allowance, commonly runs two to three times the employee's home country salary. That is the bar the assignment has to clear.

Now a claim you will meet in almost every textbook and consulting deck: that expatriate failure rates, meaning early return, run somewhere between sixteen and forty percent. Anne-Wil Harzing went looking for the evidence behind those numbers and published the results in 1995 under a title that says it all: the persistent myth of high expatriate failure rates. Tracing the citation chains, she found the figures had been passed from paper to paper, each citing an earlier source, back to a small number of studies whose own data did not support the claim, and in some cases to sources that were speculating. The numbers had become established through repetition rather than measurement. Careful subsequent work suggests premature return is considerably less common than the folklore.

Two lessons from that. The narrow one is about expatriates: do not budget or design a program around a statistic that dissolves when you check it. The broad one is about you: a number cited everywhere without an original source is a number you should chase down before you use it in a decision. That habit will serve you across this entire field.

None of which means assignments go well by default. Early return is only the most visible failure; underperforming in place, damaging a relationship with the local team, or coming home and quitting are all failures that no statistic captures. The predictors that matter are not the ones firms weight most heavily. Technical competence is necessary and dramatically overweighted in selection, because it is easy to assess. What actually distinguishes successful assignees is tolerance for ambiguity, relationship-building skill, willingness to be a beginner again, and, repeatedly across the research, family adjustment. The accompanying partner's career disruption and the family's ability to settle are among the most consistently cited reasons for early return, and they are the factors most often left out of the selection process entirely.

Then there is the phase almost everyone neglects. Repatriation is where companies waste the investment. The returning employee comes back with capabilities the firm paid a large premium to build, into a role chosen at short notice, reporting to people who did not follow their work, with their home network eroded and their local expertise now irrelevant. Surveys consistently find substantial voluntary turnover among returnees within a year or two of coming back, which means a competitor acquires the capability. The fixes are unglamorous and effective: begin the return-role conversation six to twelve months before the assignment ends, assign a home-country sponsor who stays in contact throughout, and build an explicit plan for using what the person learned rather than assuming it will surface on its own.

Finally, the alternatives, which are frequently better. Short-term assignments of a few months, international commuter arrangements, virtual assignments, hiring locally with a strong onboarding into company practice, and inpatriation, which brings host-country nationals to headquarters, all achieve some of the same purposes at a fraction of the cost. One more honest note: the familiar U-curve of culture shock, with its neat honeymoon, crisis, adjustment, and mastery phases, is far weaker in the evidence than its ubiquity suggests. Individual adjustment paths vary enormously, and a model that tells someone they should feel a crisis in month three is not doing them a favor.

Key idea: Expatriate assignments cost two to three times home salary, the widely quoted failure rates trace back to citation chains rather than data, selection overweights technical skill relative to adaptability and family adjustment, and repatriation is where firms most often lose the investment.

Teams that span countries

What does research say about whether multicultural teams perform better? The honest answer is that diversity reliably increases both the gains and the costs, and management determines which dominates. Meta-analytic work finds cultural diversity associated with greater creativity and a wider range of information brought to bear, and simultaneously with more communication difficulty, lower social integration, and more conflict. For routine execution work under time pressure, the costs can exceed the benefits. For novel problems where the range of perspectives is the point, the reverse. A manager who says diversity always helps and a manager who says it always costs are both quoting half the literature.

Two findings sharpen this. The first is faultlines: diversity is far more dangerous when several attributes line up together. A team split into engineers in one country and marketers in another has a fracture line along which it can cleanly break into two camps, whereas a team where nationality, function, and tenure crosscut has no natural fault. If you are staffing a global team, deliberately crosscutting the categories is one of the cheapest interventions available.

The second is language. Research by Tsedal Neeley on companies mandating English as a corporate language found that non-native speakers lose status, contribute less, and are judged as less competent than their actual expertise warrants, while native speakers unintentionally dominate by speaking faster, using idiom, and filling pauses. This is not a cultural values difference; it is a mechanical consequence of running a meeting in one group's first language. The remedies are concrete: slow down and drop idiom, circulate written material in advance so non-native speakers can prepare, take substantive decisions in writing where everyone competes on equal terms, rotate who facilitates, and explicitly invite specific people rather than waiting for them to break in.

For distributed teams add the mechanics. Rotate meeting times so the same region is not always taking calls at ten at night, and be seen to do it. Default to asynchronous written decisions with a decision log, because a team spanning three time zones cannot be in one room and pretending otherwise privileges whoever shares the headquarters clock. Overinvest in the first in-person meeting if you can afford one; the research on trust in virtual teams consistently finds early face-to-face contact pays for itself.

Above all, make norms explicit. A team drawn from one place runs on unspoken shared defaults about how disagreement is raised, how deadlines are treated, and what silence means. A team drawn from five places has no such defaults, and the ones it falls into by accident will be the headquarters ones. State the norms out loud: how we disagree, how we escalate, what a commitment means, what happens when someone will miss a date. That is not bureaucracy. It is the substitute for a shared context the team does not have.

Key idea: Cultural diversity raises both creative gains and coordination costs, faultlines and shared language mechanics drive much of the downside, and explicit norms plus asynchronous written decisions substitute for the unspoken context a cross-border team lacks.

Try it

You must staff a three-year assignment to launch operations in a country your firm has never entered. Two candidates: Priya, your best process engineer, technically outstanding, no international experience, married to a physician with an unportable license and two children in secondary school. Marco, a solid but not exceptional operations manager, previously spent two years in another country, speaks the language at working level, single, and has asked for the role. Who do you send, and what would change your answer?

Answer: On the evidence, Marco is the safer choice, and the reason is not that Priya is less capable. Prior successful international experience, working language ability, and volunteering are all associated with better adjustment, while Priya's situation contains the single most consistently cited driver of early return: an accompanying partner with a career that cannot move. A physician cannot practice on a license that does not transfer, and mid-secondary-school children are the hardest age to relocate. None of that is a judgment about Priya. What would change the answer: if the assignment's core purpose is transferring a specific technical process that only Priya commands, then the right move is often not to choose between them but to redesign the assignment, sending Priya for a series of short trips or a six-month posting to transfer the process while Marco holds the resident role. If Priya is the choice, the mitigations are concrete: fund a job search and licensing pathway for her spouse, pay for schooling continuity, provide language training for the whole family, and appoint a home sponsor and a return-role discussion date before departure. And ask Priya directly whether her family wants this, because a candidate who has not had that conversation at home is a risk regardless of qualifications.

Common misconceptions

• "Cultural fluency beats a strong BATNA." It does not. Etiquette improves execution; your alternatives determine your leverage.
• "A vague contract means they are not serious." Preference for general agreements that develop with circumstances is a legitimate approach, not evasion, and the reverse reading is equally common.
• "Between sixteen and forty percent of expatriates fail." Harzing traced those figures to citation chains rather than data. Chase any ubiquitous statistic to its source before budgeting on it.
• "Technical excellence is the main predictor of assignment success." It is necessary, easy to measure, and overweighted; adaptability and family adjustment predict better.
• "The assignment ends when they come home." Repatriation is where firms lose the capability they paid for, through turnover among returnees within a year or two.
• "Diverse teams always outperform." Diversity raises both creative gains and coordination costs; task type and management determine the net effect.

Recap

• BATNA, reservation price, and the zone of possible agreement govern outcomes everywhere and outrank cultural technique.
• Negotiating practice varies most damagingly in the goal pursued, the preferred form of agreement, and where decision authority actually sits.
• Ask directly about the approval process, brief interpreters in advance, budget double the time, and keep gifts inside a documented policy.
• Expatriate packages cost two to three times home salary, and the famous failure rates trace to citation chains, not measurement.
• Adaptability and family adjustment predict assignment success better than technical skill, and repatriation planning is where most firms fail.
• Cultural diversity raises creativity and coordination costs together; crosscut faultlines, manage the language penalty, and make norms explicit.

Sources

1. Bright, D. S., & Cortes, A. H. (2019). Conflict and negotiations. In Principles of management. OpenStax, Rice University. openstax.org
2. Society for Human Resource Management. (2025). Managing international assignments. SHRM. shrm.org
3. Organisation for Economic Co-operation and Development. (2024). International migration outlook. OECD. oecd.org
4. Wikipedia contributors. (2025). Expatriate. Wikipedia. en.wikipedia.org
5. Wikipedia contributors. (2025). Negotiation. Wikipedia. en.wikipedia.org

The big picture

Your company has decided that Vietnam is the market. Congratulations: you have answered the easy question. The hard one is how you get in, and the answer determines what kind of company you become there. Ship from home and you keep your factory busy and learn almost nothing about the customer. License your brand and you collect royalties while someone else builds the relationship, the reputation, and eventually the capability to replace you. Build your own plant and you control everything, including the losses.

These are not shades of one decision. They commit different capital, expose different assets, produce different rates of learning, and are reversible to very different degrees. A firm that picks its entry mode by asking what everyone else does is making a strategic choice by accident.

This lesson does two connected things. First, the modes and their trade-offs, with a test for choosing between them. Second, the question that follows immediately once you are in more than one country: how much of what you do should be the same everywhere, and how much should change.

The modes, and what each one costs you

Mode	Control	Capital at risk	Speed	Main danger
Indirect exporting through a trading company	Very low	Almost none	Immediate	You learn nothing about the market
Direct exporting via agent or distributor	Low	Low	Fast	Termination protection laws make partners hard to remove
Licensing	Contractual only	Minimal	Fast	You may be training your future competitor
Franchising	Moderate, through the system	Low	Moderate	Brand damage from operators you do not control directly
Contract manufacturing	Low over process	Low	Fast	Supplier learns to make your product without you
Joint venture	Shared	Moderate	Moderate	Partner conflict, knowledge leakage, deadlock
Wholly owned subsidiary, greenfield	Full	High	Slow	Expensive mistakes, slow learning curve
Wholly owned subsidiary, acquisition	Full	Highest	Fast	You inherit the culture, the liabilities, and the integration problem

Several deserve a closer look because the differences are not obvious from the labels.

Agent versus distributor is a distinction people blur and lawyers do not. An agent finds customers and earns commission without taking title; you remain the seller, set the price, and carry the credit risk. A distributor buys from you and resells at its own price; you get paid on shipment and lose control of pricing and often of positioning. In many jurisdictions, particularly in Europe, Latin America, and the Middle East, both enjoy statutory termination protections entitling them to compensation regardless of what your contract says, so choosing badly can cost you the market for years.

Licensing has the widest gap between how it looks on a spreadsheet and how it plays out. A royalty stream with near-zero capital employed is an outstanding return on investment. The catch is that you are transferring know-how to an organization whose interests diverge from yours the moment it has learned enough, and quality control exists only to the extent your contract creates it and local courts enforce it. Licensing is strongest for mature technology in markets you would not otherwise reach, and weakest for your core capability in a market that matters.

Joint ventures get chosen for three reasons: the partner has knowledge or relationships you cannot buy, the cost or risk is more than you want alone, or the law requires it. That third reason has receded in some places, notably in Chinese automotive manufacturing, where longstanding joint venture requirements were phased out in the early 2020s, but it persists in many sectors and countries. Joint ventures are also the mode with the highest instability. Figures in the range of thirty to fifty percent for dissolution or restructuring circulate widely, and you should treat them the way you now treat expatriate failure rates: directionally informative, weakly sourced. The mechanism of failure is well documented even where the rate is not. Partners who agreed on the market often disagree on reinvestment versus dividends, on transfer pricing between the venture and each parent, and on whose people run it, and each parent has a private incentive to learn faster than it teaches. The dispute between a French food company and its Chinese joint venture partner in the late 2000s, which spilled into litigation and arbitration on several continents, is the standard cautionary case: the venture agreement did not clearly control what the partner could do outside it.

Acquisition versus greenfield trades speed against integration risk. Acquisition buys you a market position, a customer base, licenses, and staff on day one, which is why it dominates entry into mature markets. It also buys you a company built by other people for other purposes, and the finance literature on cross-border acquisitions is sobering about how often the acquirer captures the value it paid for. Greenfield lets you build the culture and processes you want and is correspondingly slow.

Key idea: Entry modes trade control against capital and speed, with licensing risking your know-how, distributors carrying statutory termination protection, joint ventures failing through diverging parent incentives, and acquisitions buying speed at the price of integration risk.

A test for choosing: ownership, location, internalization

John Dunning's eclectic paradigm, usually called OLI, gives you a genuinely useful three-question test for whether foreign direct investment is the right answer at all.

• Ownership advantage. Do you have something valuable that local firms do not: a brand, a technology, a process, a scale advantage? If not, you have no business competing abroad against firms that know the market better than you do. This is the question most enthusiastic entry plans skip.
• Location advantage. Is there a reason to do the activity there rather than at home and export? Reasons include tariffs and freight, customer proximity, input costs, talent, and regulatory requirements for local presence.
• Internalization advantage. Is there a reason to do it yourself rather than contract with someone local? Reasons include protecting know-how that a contract cannot fence, maintaining quality that customers attribute to your brand, and avoiding the cost of writing and enforcing contracts over something hard to specify.

Read the test as a decision tree. Ownership advantage but no internalization advantage points to licensing: you have something worth selling and no reason to operate it yourself. Ownership and internalization but no location advantage points to exporting: keep production at home and sell abroad. All three present points to direct investment. This is a fifteen-minute conversation that regularly saves a company from an expensive answer to a question it never asked.

Two more ideas to carry. The Uppsala model, from Swedish researchers Johanson and Vahlne, describes what firms actually do: they enter psychologically and culturally close markets first, commit incrementally, and expand as uncertainty resolves. It is a good description of most manufacturers and a poor description of the born global firms, common in software and digital services, that sell into a dozen countries within two years of founding because their distribution cost per additional country is close to zero.

The case record is instructive and rarely flattering. A large American retailer entered Germany and exited in 2006, and sold control of its Japanese business in 2020, while entering India through a large acquisition of a domestic e-commerce firm in 2018, three different answers to the same strategic question in three markets. An American coffee chain bought out its East China joint venture partner in 2017 to take full ownership of a market it had decided to run itself. A ride-hailing company sold its China operations to a domestic rival in 2016 after concluding it could not win. A global fast food chain sold majority control of its China and Hong Kong business to a local consortium in 2017, shifting from operating to licensing. Notice that these firms did not have one entry mode; they had a portfolio, revised as they learned.

Key idea: Dunning's test asks whether you have an ownership advantage, whether the activity belongs abroad, and whether you must do it yourself, and the answers point respectively toward licensing, exporting, or direct investment.

Standardize or adapt

In 1983 Theodore Levitt argued in a famous article that technology and communication were converging tastes worldwide, and that the winners would be firms selling standardized products at low prices everywhere. The argument was influential, partly right, and wrong in its strong form. Four decades later the world is neither one market nor a set of unrelated ones.

The framework that organizes this properly comes from Prahalad and Doz and from Bartlett and Ghoshal. Two independent pressures act on any multinational. Pressure for global integration comes from scale economies, cost competition, universal product needs, customers who buy globally, and rivals who operate globally. Pressure for local responsiveness comes from differences in tastes, distribution structures, regulation, and host government demands. Score any business high or low on each and you get four postures:

Posture	Integration pressure	Responsiveness pressure	Typical example
International	Low	Low	Exporting a specialist industrial product with little adaptation
Multidomestic	Low	High	Packaged food brands run largely country by country
Global standardization	High	Low	Semiconductors, aircraft, luxury goods
Transnational	High	High	Consumer electronics and fast food chains, standard system with local menus

The transnational cell is where most large consumer businesses now try to live, and it is genuinely hard, because it requires being cheap through scale and different by market at the same time. The practical resolution is usually to standardize the expensive, invisible parts, meaning platforms, components, systems, back office, and supply chain, and to adapt the cheap, visible parts, meaning menu, packaging, sizing, positioning, and service. A furniture retailer entering China standardized its global product platform and adapted almost everything customer-facing: showroom layouts built around small apartments, store restaurants designed as social destinations, and a decision not to fight the customers who came to nap on the display beds.

Pankaj Ghemawat's contribution is the best corrective to both extremes. His CAGE framework says distance between two countries has four components: cultural, administrative and political, geographic, and economic. Distance is not symmetric across industries, which is the useful part. Cultural distance matters enormously for media and food and hardly at all for cement. Administrative distance matters enormously in regulated sectors. Ghemawat's broader claim, which he calls semi-globalization, is that most economic activity remains far more domestic than the rhetoric suggests, and that firms consistently overestimate how borderless their market is.

Across the marketing mix, the recurring lessons are these. On product, decide which attributes are core and which are peripheral, and adapt only the peripheral. On price, remember that differences between country prices create gray markets, in which distributors buy in the cheap country and resell in the expensive one, so plan a price corridor rather than pricing each market in isolation. On place, expect distribution structures to differ more than anything else: a market dominated by small independent retailers requires a completely different route to market from one dominated by a handful of chains. On promotion, adapt meaning rather than translating words.

A word on the famous translation disasters, because most are false. The story that a Chevrolet model failed in Latin America because its name reads as does not go in Spanish is apocryphal, and the tale of a soft drink slogan promising to raise your ancestors from the dead has never been substantiated. A real and documented case is the large international bank that in 2009 abandoned and replaced a global tagline, at a reported cost near ten million dollars, because it did not survive rendering into other languages. Repeating the myths teaches that the risk is silly wordplay, when the actual risk is expensive and prosaic.

Key idea: The integration-responsiveness framework gives four postures, most consumer multinationals aim for the difficult transnational cell by standardizing invisible costs and adapting visible experience, and CAGE distance explains why the right answer differs by industry.

Try it

A mid-sized US maker of industrial water filtration systems wants to enter Brazil. Its products embody patented membrane technology, require installation and annual servicing, and Brazil imposes high tariffs on imported industrial equipment plus local content preferences in public procurement. Its Brazilian competitors are weaker technically but have long-standing municipal relationships. Apply OLI and recommend an entry mode.

Answer: Ownership advantage: yes, clearly, in the patented membrane technology and systems engineering, which local competitors lack. Location advantage: yes, and strongly, because high tariffs plus local content preferences in public procurement make exporting from the United States expensive and in some tenders disqualifying, while installation and annual servicing require people on the ground regardless. Internalization advantage: yes, because the core asset is process know-how that a licensing contract cannot fence effectively, and because service quality is attributed to the brand. All three present points to direct investment. Between greenfield and acquisition, the specific facts favor acquisition or a joint venture with a local firm, because the binding constraint is not manufacturing capability but municipal relationships and tender access, which are exactly what an incumbent has and a greenfield plant does not. The sensible structure is a majority-owned venture or acquisition that keeps the local commercial organization intact while retaining membrane production at home or in a tightly controlled local line, with contractual and practical protection of the technology. If the firm cannot fund that, the fallback is direct exporting through a distributor with service capability, accepting the tariff and the exclusion from some public tenders, while it builds the case for investment.

Common misconceptions

• "Licensing is free money." It is a royalty stream purchased by transferring know-how to a party whose interests diverge once it has learned enough.
• "An agent and a distributor are the same thing." One takes commission without title; the other buys and resells at its own price, and both often enjoy statutory termination protection.
• "Acquisition is the safe way in because the business already works." You inherit a company built by other people for other purposes, and cross-border acquirers frequently fail to capture the value they paid for.
• "Joint ventures fail because partners pick badly." They more often fail through structurally diverging incentives on reinvestment, transfer pricing, staffing, and learning.
• "Global brands succeed by standardizing everything." The workable pattern standardizes invisible cost drivers and adapts visible customer experience.
• "Famous translation blunders show the real risk." Most are apocryphal; the documented failures are expensive and dull.

Recap

• Entry modes run from indirect exporting to wholly owned subsidiaries, trading control against capital at risk, speed, learning, and intellectual property exposure.
• Agent and distributor arrangements differ in title, pricing control, and credit risk, and statutory termination protections make partner choice hard to reverse.
• Dunning's OLI test points to licensing without internalization advantage, exporting without location advantage, and direct investment when all three hold.
• The Uppsala model describes incremental commitment to nearby markets, while born globals contradict it in digital sectors.
• The integration-responsiveness framework yields international, multidomestic, global standardization, and transnational postures, with most consumer multinationals pursuing the last.
• CAGE distance varies by industry, gray markets punish inconsistent country pricing, and adaptation should target meaning and visible experience rather than words alone.

Sources

1. U.S. International Trade Administration. (2025). Market entry strategies and country commercial guides. Trade.gov. trade.gov
2. United Nations Conference on Trade and Development. (2024). World investment report. UNCTAD. unctad.org
3. Encyclopaedia Britannica. (2024). Multinational corporation. britannica.com
4. Wikipedia contributors. (2025). Eclectic paradigm. Wikipedia. en.wikipedia.org
5. Wikipedia contributors. (2025). Uppsala model. Wikipedia. en.wikipedia.org

The big picture

You have just signed a purchase order for five hundred thousand euros of components, payable ninety days after shipment. Your revenue is in dollars. As of this morning that invoice costs about five hundred forty thousand dollars, but you will not pay it this morning. Between now and then, the amount of your own money that this fixed euro obligation consumes will move, possibly by several percent in either direction, for reasons that have nothing to do with your business.

You have four choices: accept the risk, lock the rate today with a forward, replicate the forward using the money markets, or buy the right but not the obligation to convert at a set rate. Each has a computable cost, and by the end of this lesson you will price all four on the same invoice and see why the answers come out so close together.

Currency is where international business becomes arithmetic, and it is where students most often bluff. The good news is that the arithmetic is simple, and that almost all confusion comes from two avoidable errors: not knowing which currency a quote is in, and getting a percentage backwards. We will kill both.

Reading a quote and converting money

The foreign exchange market is the largest financial market in existence. The Bank for International Settlements, which surveys it every three years, put average daily turnover at roughly seven and a half trillion dollars in April 2022, with the most recent round higher still. It is decentralized and over the counter, runs continuously from the Monday open in Asia to the Friday close in New York, and London remains its largest single center. The US dollar is on one side of the overwhelming majority of trades, far out of proportion to the American share of world trade, which will matter later.

Quotes name a currency pair, and the first currency is the base. EUR/USD at 1.0850 means one euro costs 1.0850 dollars; USD/JPY at 150.00 means one dollar costs 150 yen. Get this backwards and every number you produce afterward is wrong, so form the habit of saying the quote aloud as a sentence: one of the first currency costs this many of the second.

Banks quote two prices: the bid, at which the bank buys the base currency from you, and the ask, at which it sells to you. A quote of EUR/USD 1.0848 / 1.0852 means the bank buys euros at 1.0848 and sells at 1.0852, and you always transact on the side worse for you. Buying five hundred thousand euros at the ask costs 542,600 dollars. Interbank spreads on major pairs are tiny; what you actually pay is not. If a bank charges roughly one and a half percent over interbank, your effective rate becomes about 1.1015 and the same five hundred thousand euros costs about 550,750 dollars, roughly 8,150 dollars for a conversion in which nothing looks different. Corporate treasurers negotiate spreads for exactly this reason, and so should you: retail card and cash-exchange markups of two to three percent are routine and usually the largest fee anyone pays while traveling.

Cross rates handle pairs that are not quoted directly. If EUR/USD is 1.0850 and USD/JPY is 150.00, then one euro buys 1.0850 times 150.00, or 162.75 yen. Multiply when the shared currency cancels; divide when it does not. Write out the units the way you would in a physics problem and you will never get it wrong.

Now the percentage trap, which catches nearly everyone. Suppose EUR/USD moves from 1.0000 to 1.2500. The euro appreciated by twenty-five percent against the dollar. By how much did the dollar depreciate against the euro? Not twenty-five percent. Invert: a dollar used to buy one euro and now buys 1 divided by 1.25, or 0.80 euros, a fall of twenty percent. Appreciation and depreciation between the same two currencies are never equal percentages, because they are computed on different bases. If a report gives you one number, compute the other yourself before quoting it.

Key idea: Read the base currency first, transact at the worse side of the bid-ask spread, build cross rates by cancelling units, and never assume that a currency's appreciation equals the other's depreciation in percentage terms.

Forwards, and why interest rates set the price

A forward contract obliges you to exchange a set amount at a set rate on a set future date. The rate is not anyone's forecast. It is arithmetic, forced by the possibility of arbitrage, and the relationship is called covered interest parity.

The logic: you can turn dollars into euros in ninety days two ways. Convert now and hold euros in a euro deposit, or hold dollars in a dollar deposit and buy euros forward. If those two routes gave different results, a bank would borrow in one and lend in the other until they did not. So the forward rate must satisfy: the forward, in quote currency per base currency, equals the spot times one plus the quote currency interest rate, divided by one plus the base currency interest rate.

Work it. Spot EUR/USD is 1.0800. The one-year dollar interest rate is 5.0 percent, the one-year euro rate 3.0 percent. Dollars are the quote currency, euros the base. So the one-year forward is 1.0800 times 1.05 divided by 1.03, which is 1.0800 times 1.01942, or about 1.1010.

Read what that means. The euro trades at a forward premium of about 1.9 percent, and the dollar at a forward discount, despite the dollar paying the higher interest rate. That is the whole point. The currency with the higher interest rate always trades at a forward discount, by exactly enough to cancel the interest advantage. If it did not, you could borrow in the low-rate currency, lend in the high-rate one, sell the proceeds forward, and pocket a riskless profit. There is no free lunch in the forward market, and a salesperson who tells you a forward is cheap because a currency pays more interest is telling you they do not understand the instrument.

One warning: the forward rate is a terrible forecast of the future spot rate, and is not trying to be one. The forward premium has historically pointed the wrong way on average, an anomaly known as the forward premium puzzle and the basis of the carry trade.

Key idea: Forward rates are set by covered interest parity rather than by forecasts, the higher-interest currency always trades at a forward discount that cancels its yield advantage, and the forward is not a prediction of the future spot rate.

Three exposures, four hedges, one invoice

Before hedging, identify what you are exposed to. There are three kinds, and firms usually manage the least important most carefully.

• Transaction exposure: contracted cash flows in a foreign currency, like our euro invoice. Easy to see, easy to hedge.
• Translation exposure: the accounting effect of restating a foreign subsidiary into the parent's currency. It moves reported equity without moving cash, so hedging spends real money on a paper number, which is why many firms deliberately do not.
• Economic or operating exposure: the effect on the present value of future cash flows, including competitive position. The largest and hardest. A US manufacturer with all costs and revenues in dollars is still badly exposed if a strong dollar lets a Japanese rival undercut it, and no forward contract fixes that.

Now price the invoice four ways. Five hundred thousand euros due in ninety days. Spot 1.0800, so 540,000 dollars at today's rate. Ninety-day rates: dollars 5 percent a year, which is 1.25 percent for the quarter; euros 3 percent a year, or 0.75 percent for the quarter. The bank offers a ninety-day forward of 1.0850.

Approach	Cost if spot ends 1.1500	Cost if spot ends 1.0800	Cost if spot ends 1.0200
Unhedged	575,000	540,000	510,000
Forward at 1.0850	542,500	542,500	542,500
Money market hedge	542,680	542,680	542,680
Call option, strike 1.0900, premium 7,500	552,500	547,500	517,500

Check the two you cannot read off directly. The money market hedge replicates the forward using deposits: you need 500,000 euros in ninety days, so deposit 500,000 divided by 1.0075, which is 496,278 euros today. Buying those euros at spot costs 535,980 dollars. Borrow that in dollars at 1.25 percent for the quarter and you repay 542,680 dollars. Notice how close that is to the forward at 542,500. It is close because covered interest parity says it must be; the small gap is the bank's margin, and comparing the two is a legitimate way to check whether your forward quote is fair.

The option caps your cost without capping your benefit. A call on euros struck at 1.0900 costing 0.015 dollars per euro means 7,500 dollars paid up front. If the euro rises to 1.15 you exercise: 545,000 dollars plus the premium equals 552,500. If the euro falls to 1.02 you let it expire and buy spot: 510,000 plus premium equals 517,500. The option is worse than the forward in the bad case, by the premium, and better in the good case. That is what insurance is, and treasurers who describe an unexercised option as wasted money are describing every insurance policy that did not pay out.

The cheapest hedges are usually not financial at all. Natural hedging means arranging the business so exposures cancel: earn revenue in the currency you incur costs in, borrow in the currency you earn, source locally where you sell, and net offsetting exposures across group companies before going to a bank. A Japanese automaker building assembly plants in North America gains a natural hedge alongside everything else. Also know when not to hedge: it costs spread, margin, and management attention, and a firm with genuinely offsetting exposures may rationally leave much of it alone. The unforgivable position is not being unhedged; it is not knowing what your exposure is.

Key idea: Transaction exposure is easy to hedge, translation exposure usually is not worth hedging, economic exposure is the largest and needs operating solutions, and the forward, money market hedge, and option price out close together with the option buying optionality for a premium.

Regimes, the impossible trinity, and getting paid

Not every currency floats. The IMF classifies actual behavior rather than official claims, and the spectrum runs from hard pegs through soft pegs to floating. At the rigid end sit countries using another country's currency outright, as Panama and Ecuador use the dollar, and currency boards, of which Hong Kong's is the best known, holding the local dollar in a narrow band against the US dollar for decades. In the middle sit conventional pegs, crawling pegs, and managed arrangements: China operates a managed float around a daily reference rate within a permitted band. At the free end sit the dollar, euro, yen, and sterling.

Why not simply choose stability? Because of the impossible trinity, the most useful idea in international macroeconomics. A country can have at most two of these three: a fixed exchange rate, free movement of capital, and an independent monetary policy. Hong Kong takes the fixed rate and free capital, and therefore imports US interest rate policy whether or not it suits Hong Kong. The United States, the euro area, and Japan take free capital and monetary independence and accept a floating rate. China for years took a managed rate and monetary independence, maintaining capital controls to make that possible.

The trinity is not a theory that might be wrong; it is a constraint that has broken governments repeatedly. Britain was forced out of the European Exchange Rate Mechanism in September 1992 when defending the peg required interest rates the domestic economy could not bear. Thailand abandoned its dollar peg in July 1997, triggering the Asian financial crisis. Argentina's currency board collapsed in 2001 and 2002. And in January 2015 the Swiss National Bank abandoned its floor against the euro without warning; the franc jumped roughly thirty percent intraday, ruining hedges and brokerages that had assumed the floor would hold. Pegs hold until they do not, and they break fast.

Over long horizons, exchange rates are loosely disciplined by purchasing power parity, which fails constantly in the short run because most of what people buy is not traded. One modern refinement: because the dollar invoices a large majority of world trade, exporters in many countries set dollar prices and hold them sticky, so exports do not cheapen for foreign buyers as fast as textbook models predict when a currency falls. That is part of why devaluations disappoint.

Finally, getting paid. Cross-border sales sit on a risk ladder. Cash in advance is safest for the exporter and hardest to sell. A letter of credit substitutes a bank's promise for the buyer's, paying against conforming documents, which shifts risk from an unknown foreign buyer to a bank at a fee, with the discipline that documents must match exactly. Documentary collection is cheaper and weaker. Open account, where you ship and invoice and hope, is riskiest for the seller and carries the majority of world trade by value, because established relationships do not need letters of credit. Export credit agencies insure or finance what the private market will not. On the tax side, know two terms: transfer pricing, governed by the OECD arm's length standard, which determines how much profit a multinational books in each country, and the global minimum tax agreed by well over one hundred jurisdictions and taking effect across many from 2024.

Key idea: Regimes run from dollarization and currency boards through managed floats to free floats, the impossible trinity forces every country to give up one of fixed rates, free capital, or monetary independence, and payment risk is managed along a ladder from cash in advance through letters of credit to open account.

Try it

Your firm will receive 2,000,000 Canadian dollars in six months. Spot USD/CAD is 1.3600, meaning one US dollar costs 1.36 Canadian dollars. Six-month rates are 4.0 percent a year in the US and 3.0 percent a year in Canada. Compute the six-month forward rate implied by covered interest parity, the US dollar amount you would lock in, and state whether you are hedging against the Canadian dollar rising or falling.

Answer: The base currency is the US dollar and the quote currency is the Canadian dollar, so the parity formula puts the Canadian rate on top: forward equals 1.3600 times 1.015 divided by 1.020, since six months is half of each annual rate. That is 1.3600 times 0.99510, about 1.3533 Canadian dollars per US dollar, meaning the Canadian dollar is at a small forward premium, consistent with its lower interest rate. To convert your receipt, divide: 2,000,000 divided by 1.3533 locks in about 1,477,900 US dollars, against about 1,470,600 at today's spot, so the forward is slightly better here. As a receiver of Canadian dollars you are exposed to USD/CAD rising, because more Canadian dollars per US dollar means your fixed receipt converts into fewer US dollars. The forward protects against exactly that and gives up the gain if the Canadian dollar strengthens.

Common misconceptions

• "A forward rate is the market's forecast." It is arithmetic from interest rates, forced by arbitrage, and historically a poor predictor of the future spot rate.
• "You should hedge into the currency paying higher interest." That currency trades at a forward discount that exactly cancels the yield advantage. There is no free lunch.
• "A twenty-five percent appreciation means a twenty-five percent depreciation the other way." Inverting 1.25 gives 0.80, a twenty percent fall.
• "No foreign currency receivables means no currency risk." Economic exposure through competitive position is usually a firm's largest exposure.
• "An option that expired unexercised was wasted money." That describes every insurance policy that did not pay out.
• "A country can peg its currency, allow free capital flows, and run its own monetary policy." The impossible trinity says pick two, and 1992, 1997, 2002, and 2015 are what happens when a government forgets.

Recap

• Quotes name the base currency first, you transact at the worse side of the spread, and retail markups of two to three percent dwarf interbank spreads.
• Cross rates come from cancelling units, and appreciation and depreciation between two currencies are never equal in percentage terms.
• Covered interest parity sets the forward rate, so the higher-interest currency trades at a forward discount and the forward is not a forecast.
• Transaction, translation, and economic exposure differ in visibility and importance, with economic exposure the largest and least hedgeable financially.
• Forward, money market hedge, and option priced our five hundred thousand euro invoice at about 542,500, 542,680, and a capped 552,500 respectively, and natural hedging is often cheapest of all.
• The impossible trinity constrains every exchange rate regime, and payment risk runs from cash in advance through letters of credit to open account, which carries most world trade by value.

Sources

1. Bank for International Settlements. (2022). Triennial central bank survey of foreign exchange and OTC derivatives markets. BIS. bis.org
2. Board of Governors of the Federal Reserve System. (2025). Foreign exchange rates, H.10. Federal Reserve. federalreserve.gov
3. International Monetary Fund. (2024). Annual report on exchange arrangements and exchange restrictions. IMF. imf.org
4. U.S. International Trade Administration. (2025). Trade finance guide. Trade.gov. trade.gov
5. Encyclopaedia Britannica. (2024). Foreign exchange. britannica.com

The big picture

On 11 March 2011 an earthquake and tsunami struck northeastern Japan. Within days, automakers on three continents were cutting production, and the reason was not damaged assembly plants. A single semiconductor facility in Naka made a large share of the world's automotive microcontrollers, the chips that run engine management and braking systems. It went down. Separately, a plant producing a specialty pigment used in metallic automotive paint was knocked out, and for months some manufacturers could not supply certain colors. Neither the chip plant nor the pigment plant appeared in any carmaker's list of major suppliers, because neither sold to carmakers. They sold to companies that sold to companies that sold to carmakers.

That is the central fact of modern supply chains. The efficiency that makes them cheap comes from specialization and scale, and specialization and scale mean concentration, and concentration means that a local event becomes a global event. Every optimization that removed cost also removed slack, and slack is what absorbs shocks. This is not a design flaw that better managers would have avoided. It is the direct consequence of decades of correctly following the incentives.

This lesson does three things: it establishes how the fragility works using cases you can verify, it teaches you to build a total landed cost so sourcing decisions rest on the real number, and it prices resilience honestly, because resilience is not free and pretending otherwise is how initiatives die in the second budget cycle.

How supply chains actually break

The disruption record of the last fifteen years is long enough to draw patterns from.

The 2011 Thailand floods submerged industrial estates producing a large share of the world's hard disk drives. Prices of drives roughly doubled and took about a year to normalize, and computer manufacturers rationed. The lesson is geographic concentration: an entire industry had clustered in one floodplain because clustering is efficient.

The semiconductor shortage of 2020 to 2022 is the best-documented case of a demand-side mistake amplifying through a chain. When the pandemic hit, automakers forecast a collapse in car sales and cancelled chip orders. Foundries reallocated that capacity to consumer electronics, where demand was surging as the world moved indoors. Car sales recovered far faster than expected, the automakers went back, and the capacity was gone, with lead times on some parts stretching beyond a year. Industry analysts estimated the resulting lost automotive production in the millions of vehicles and lost revenue in the range of two hundred billion dollars for 2021 alone. Note the mechanism: nobody's factory burned down. A forecast error propagated because the chain had no buffer and reallocated capacity does not come back on demand.

The Ever Given grounding in the Suez Canal in March 2021 blocked a waterway carrying roughly a tenth of world trade for six days, with estimates of goods held up running to several billion dollars per day and schedule effects lasting months. From late 2023, attacks on shipping in the Red Sea pushed most carriers to route around the Cape of Good Hope instead, adding something like ten to fourteen days and thousands of nautical miles to Asia-Europe voyages. Around the same period, drought reduced water levels in the Panama Canal enough to force transit restrictions. Three of the world's critical maritime chokepoints came under pressure within a few years, from causes with nothing in common.

Underneath the headline events sit two structural problems.

The first is the bullwhip effect, identified in the 1960s and formalized by Hau Lee and colleagues in the 1990s. Small variations in end-customer demand amplify as they travel upstream, because each tier adds safety stock, batches its orders, reacts to price promotions, and rations during shortages. A five percent wobble at the retailer can become a forty percent swing at a component supplier. The chip shortage is the bullwhip with a global amplifier attached.

The second is visibility. Most firms know their direct suppliers well, know their suppliers' suppliers vaguely, and know nothing beyond that. Yet the chokepoints usually sit at tier two or three, precisely because those tiers are where extreme specialization lives. The Japanese chip plant, the pigment producer, a single manufacturer of a photoresist chemical, one company making the machines that make advanced chips: these are the nodes that stop the world, and they are invisible on a standard supplier list. Ask any executive who their single points of failure are and the honest answer is usually that they do not know.

One case cuts the other way and is worth carrying for balance. In February 1997 a fire destroyed the plant of a supplier making a brake valve used in most Toyota vehicles, a part sourced from essentially one location under just-in-time discipline. Toyota had days of inventory. Within roughly a week, more than two hundred companies in its supplier network, including firms with no experience of the part, had improvised production using shared drawings and borrowed equipment, and output recovered far faster than anyone expected. The lesson is not that just-in-time is safe. It is that a dense network with deep relationships and a shared interest in the system's survival is itself a form of resilience, and one that arms-length transactional sourcing does not produce.

Key idea: Supply chains break through geographic concentration, demand amplification through the bullwhip effect, chokepoints at invisible lower tiers, and physical constriction at a handful of maritime passages, and dense supplier relationships can substitute for inventory.

Building a total landed cost

Offshoring decisions are frequently made on the unit price. That number is close to meaningless. Total landed cost counts everything required to get a usable unit into your facility, and building one honestly changes decisions.

Work a case. You need 100,000 units a year of an aluminum housing. A domestic supplier quotes 18.00 dollars per unit with a two-week lead time. An overseas supplier quotes 11.20 dollars per unit at its port, a headline saving of thirty-eight percent. Now add everything else. A forty-foot container holds 6,000 units.

Cost element	Basis	Per unit
Unit price at origin port	Quoted	11.20
Ocean freight	3,600 per container, 6,000 units	0.60
Marine insurance	0.5 percent of goods value	0.06
Import duty	4.5 percent of 11.20	0.50
Customs brokerage and entry	250 per container	0.04
Drayage and inland freight	900 per container	0.15
Extra inventory carrying	8 additional weeks of stock at 20 percent annual carrying cost	0.38
Quality, travel, supplier management	Audits, two trips a year, defect premium	0.30
Total landed cost		13.23

Check the inventory line, since it is the one people omit. The overseas lead time is ten weeks against two domestically, so you carry roughly eight extra weeks of pipeline and safety stock. Eight weeks of 100,000 annual units is 100,000 times 8 divided by 52, or about 15,400 units, valued at roughly 12.40 each, which is about 191,000 dollars of extra working capital. At a twenty percent annual carrying cost covering capital, storage, insurance, and obsolescence, that is about 38,000 dollars a year, or 0.38 per unit.

So the true comparison is 13.23 against 18.00, a saving of 4.77 per unit or 477,000 dollars a year. Still a clear win, though notably smaller than the thirty-eight percent headline. Now stress it with two shocks that both actually happened.

Shock one: the duty rises from 4.5 percent to 25 percent. Duty per unit goes from 0.50 to 2.80. Landed cost becomes 15.53. The saving falls to 2.47 per unit, or 247,000 a year.

Shock two: container rates spike. In 2021 spot rates on some major lanes rose roughly fivefold. At 18,000 dollars a container, freight per unit goes from 0.60 to 3.00. Landed cost becomes 17.93. The saving is now 0.07 per unit, or 7,000 dollars a year on a hundred thousand units. The decision that looked like a thirty-eight percent cost advantage is a rounding error, and that is before recognizing that duty and carrying costs would rise with the higher landed value.

Draw the right conclusion. The point is not that offshoring is wrong. In the base case it saved nearly half a million dollars a year, and the domestic option carries its own risks. The point is that the decision was never as robust as the unit price suggested, and that a sourcing choice which flips under two plausible shocks is a choice you should make with a hedge attached: a qualified second source, a contract clause that shares freight and duty risk, or deliberately keeping some domestic capability alive at a premium you now know how to price.

Key idea: Total landed cost adds freight, duty, brokerage, inland transport, inventory carrying, and supplier management to the unit price, and a 38 percent headline saving shrank to 26 percent on full costing and to nearly nothing under a tariff and freight shock.

What resilience costs

Every resilience measure is an insurance premium, and treating them as free is why they get cut. Price them explicitly.

• Multi-sourcing. Qualify a second supplier, often in a different country. Costs: you lose volume discounts, pay qualification and tooling costs, and split engineering attention. Watch for the trap of dual sourcing from two firms that both depend on the same tier three input, which is not dual sourcing at all.
• Buffer inventory. The simplest and most honest. Costs are the carrying cost you just learned to compute plus obsolescence risk. Cheap for stable commodity parts, expensive for anything that dates.
• Regionalization. Serve each major region from within it. Costs: you give up scale economies and duplicate fixed assets. Buys shorter lead times, less exposure to chokepoints, and tariff insulation.
• Supplier mapping. Trace critical components to tier two and three. Cost is analyst time, and the obstacle is usually that direct suppliers regard their own sources as confidential. Cheapest high-value measure on this list.
• Stress testing. Pick your top components and ask what happens if this node disappears for twelve weeks, then compute time to recovery and the revenue at risk. A well-known approach concentrates on time to recover rather than on estimating probabilities, which is wise, since nobody forecast a container ship wedging across the Suez Canal.
• Contractual and financial tools. Capacity reservation agreements, index clauses that share freight and currency swings, and business interruption insurance covering named suppliers.

How much to buy depends on a calculation you should do explicitly: what does an outage cost per week, and how long would recovery take? A component whose absence stops a line generating two million dollars a week of contribution justifies a great deal of insurance. One where you can substitute in a fortnight justifies almost none. The failure mode in most companies is not too little resilience overall; it is uniform resilience policy applied to a portfolio where risk is wildly unequal.

A closing observation on the wider picture. Firms have been rebalancing toward nearer and politically safer sourcing since 2020, and as Lesson 3 showed, a good deal of what looks like relocation is re-routing that leaves the underlying dependency intact. Trace your inputs, not your invoices.

Key idea: Multi-sourcing, buffer inventory, regionalization, supplier mapping, stress testing, and contractual tools each carry a computable cost, and resilience should be bought unevenly in proportion to what an outage costs per week and how long recovery takes.

Try it

Using the base case above, suppose the overseas supplier offers a price cut to 10.40 dollars but requires you to order in 12,000-unit lots rather than 6,000, doubling your average cycle stock contribution to inventory. Assume that raises the extra inventory carrying charge from 0.38 to 0.62 per unit and leaves everything else unchanged, including the 4.5 percent duty computed on the new price. Does the deal improve the total landed cost?

Answer: Recompute the changed lines. The unit price falls by 0.80, from 11.20 to 10.40. Duty at 4.5 percent of 10.40 is 0.468, call it 0.47, down from 0.50, a saving of 0.03. Insurance at 0.5 percent of the lower value is about 0.05, saving 0.01. Inventory carrying rises by 0.24, from 0.38 to 0.62. Net change is minus 0.80 minus 0.03 minus 0.01 plus 0.24, which is minus 0.60. Total landed cost falls from 13.23 to 12.63, so yes, the deal improves the number, by about 60,000 dollars a year. Two cautions before accepting. First, the larger lot size lengthens your exposure to a demand or specification change, so if this part is at risk of engineering revision the obsolescence cost is understated. Second, and more important, check whether the price cut is being funded by the supplier cutting its own quality or inventory buffers, because a supplier that finances your discount by making itself more fragile has sold you a saving and a risk in the same transaction.

Common misconceptions

• "Supply chains broke because managers were careless." Fragility is the direct consequence of correctly pursuing efficiency, which removes the slack that absorbs shocks.
• "The chip shortage was caused by factory closures." The main mechanism was automakers cancelling orders on a bad forecast and finding the reallocated capacity unavailable when demand returned.
• "We know our supply chain because we know our suppliers." Chokepoints usually sit at tier two or three, where extreme specialization lives and visibility ends.
• "The unit price tells you which supplier is cheaper." Freight, duty, brokerage, inland transport, inventory carrying, and supplier management moved a 38 percent advantage to 26 percent before any shock.
• "Dual sourcing removes single-point risk." Not if both suppliers depend on the same tier three input, which is common and rarely checked.
• "Just-in-time is simply reckless." The 1997 valve plant fire showed a dense supplier network recovering in about a week, so relationship depth can substitute for inventory.

Recap

• Concentration is the price of efficiency, so local events such as the 2011 Japanese earthquake and Thai floods become global shortages.
• The bullwhip effect amplifies demand variation upstream, and the 2020 to 2022 chip shortage is its clearest large-scale example.
• Chokepoints sit at invisible lower tiers and at a handful of maritime passages, three of which came under pressure within a few years.
• Total landed cost adds freight, insurance, duty, brokerage, inland freight, inventory carrying, and supplier management to the quoted unit price.
• In the worked case a 38 percent headline saving became 26 percent fully costed and nearly zero under a 25 percent tariff plus a fivefold freight spike.
• Resilience measures each have a computable cost and should be purchased unevenly, in proportion to weekly outage cost and time to recover.

Sources

1. Organisation for Economic Co-operation and Development. (2023). Global value chains and supply chain resilience. OECD. oecd.org
2. World Trade Organization. (2023). World trade report: Re-globalization. WTO. wto.org
3. U.S. Census Bureau. (2025). Foreign trade: Imports by country and commodity. U.S. Department of Commerce. census.gov
4. Wikipedia contributors. (2025). 2021 Suez Canal obstruction. Wikipedia. en.wikipedia.org
5. Wikipedia contributors. (2025). 2020-2023 global chip shortage. Wikipedia. en.wikipedia.org

The big picture

On 24 April 2013, an eight-storey building called Rana Plaza in Savar, outside Dhaka, collapsed. It housed several garment factories. Cracks had appeared the previous day and the building had been evacuated; workers were ordered back the next morning. One thousand one hundred thirty-four people died and more than two thousand five hundred were injured. Labels from well-known brands were pulled from the rubble, and the building had been through social audits.

You now have to hold two things at once. The first is that this is not an argument about wages. A building collapsing on people told to go back inside is a failure of safety regulation, of enforcement, and of a purchasing system that made speed and price the only signals travelling up the chain. The second is that the Bangladeshi garment industry employs several million people, mostly women, and the research on it includes some of the strongest evidence anywhere that export manufacturing changes lives for the better.

Both are true. This lesson presents the strongest version of each position in the labor and environmental debates, shows a case where the design of a commitment mattered more than its sincerity, and turns to what you might do with this professionally.

The labor standards argument, both sides at full strength

Start with the economists' case, because it is frequently caricatured and it is not stupid. The argument, stated most famously by Paul Krugman in a 1997 essay, is about the counterfactual. When a Western consumer compares a garment job in Dhaka with a job in Ohio, they are comparing the wrong things. The relevant comparison for the woman taking that job is the set of options actually available to her: subsistence agriculture, informal domestic work, scavenging, or nothing. Export factory work generally pays more than those, which is why people queue for it. Low-wage export manufacturing preceded large income gains in South Korea, Taiwan, and China, and no country has yet developed by refusing to do it. On this view, a campaign that closes factories without changing the alternatives makes real people worse off in the name of a principle they were not asked about.

The evidence is stronger than the rhetoric suggests. Rachel Heath and Mushfiq Mobarak studied Bangladeshi villages by proximity to garment factories and found the industry's growth associated with substantially higher school enrollment among young girls and with delayed marriage and childbearing. The prospect of a wage job requiring literacy changes what a family does with a daughter. That is not a wage statistic; it is a finding about life trajectories, and the single most powerful item in this argument's favor.

Now the critique, which is also not stupid. First, the compared to what framing quietly treats the alternatives as fixed, when they are themselves the product of policy, land tenure, education spending, and bargaining power. Asking whether a worker prefers this job to starving is not the same as asking whether the job could be safe.

Second, and decisively, most of what critics object to is not the wage. It is unsafe buildings, blocked fire exits, forced and unpaid overtime, wage theft, harassment, and the suppression of organizing. Those are not the price of development; they are violations that can be reduced without reducing employment, and a defense of low wages is not a defense of any of them. Third, the boycott framing is largely a straw man: the demands after Rana Plaza were for binding, funded, independently inspected safety standards and freedom of association.

One study speaks to both sides at once. Ann Harrison and Jason Scorse examined the 1990s anti-sweatshop campaigns in Indonesia, which combined activist pressure on foreign-owned exporting plants with minimum wage increases. Real wages for unskilled workers in targeted plants rose substantially with only modest employment effects there, because large exporters had margins to absorb the increase, while smaller domestic plants closed at higher rates. The lesson is precise: pressure aimed at firms earning rents can raise wages without destroying jobs, and blunt across-the-board pressure falls hardest on marginal firms and their workers.

One caution in the spirit of this course. You will often read that child garment workers dismissed under pressure from proposed 1990s US legislation moved into worse work. The claim is repeated because it is rhetorically convenient, and its documentation is thin. Chase the source first.

Key idea: Export manufacturing jobs generally beat locally available alternatives and are associated with real gains including girls' schooling, while unsafe buildings and suppression of organizing are violations rather than the price of development, and targeted pressure on high-margin exporters can raise wages without eliminating jobs.

Governance: why the design of a promise mattered

For two decades the standard corporate answer to supply chain labor problems was the voluntary code of conduct backed by social audits. Rana Plaza is the clearest demonstration of that model's limits: the factories in the building had been audited. Audits check documents and announced conditions, and systematically miss structural building safety, unauthorized subcontracting to unlisted factories, and workers coached on what to say. A model in which the buyer pays for an inspection of its own supplier, with no obligation to fund fixes and no enforceable consequence, produces reports rather than remediation.

What happened afterward is close to a natural experiment in institutional design. One initiative, driven by European brands with global union federations, was a legally binding agreement: independent engineering inspections, published findings, funded remediation, worker participation, and binding arbitration enforceable in the brands' home courts. The other, formed by North American retailers, was brand-controlled and voluntary, without binding arbitration or an equivalent funding commitment. Both inspected large numbers of factories, and independent assessments generally conclude the binding arrangement produced deeper and more verifiable remediation. The lesson is not that European brands were sincerer. It is that an enforceable commitment produced different behavior, exactly as the design predicted. Carry this forward as a general test: when evaluating any responsibility commitment, ask who can enforce it, what happens if it is broken, and who pays for compliance.

The policy world reached the same conclusion, which is why the field has shifted from voluntary standards to legal duties:

• UN Guiding Principles on Business and Human Rights, endorsed 2011: the duty to protect, the responsibility to respect, access to remedy, and due diligence as the operative concept.
• OECD Guidelines for Multinational Enterprises, updated 2023, with National Contact Points hearing complaints.
• ILO core conventions and the 1998 Declaration on Fundamental Principles and Rights at Work, which added a safe and healthy working environment as a fifth category in 2022.
• Due diligence laws: France's 2017 duty of vigilance law and Germany's supply chain act in force from 2023, plus the EU directive adopted in 2024 and since subject to delay proposals.
• Import bans: the US Uyghur Forced Labor Prevention Act, effective June 2022, presumes goods with Xinjiang inputs are made with forced labor, making traceability a customs requirement rather than an ethics preference.

Key idea: Voluntary codes and audits failed at Rana Plaza, the binding agreement outperformed the voluntary one because enforceability changed behavior, and the field has since moved toward legal due diligence duties and import bans.

Environmental standards, leakage, and one counterintuitive fact

The parallel environmental question is whether firms relocate to escape regulation. The pollution haven hypothesis predicts they do, and the evidence says the effect is real but modest and concentrated in the dirtiest, most energy-intensive industries, because compliance is usually a small share of total cost next to labor, logistics, and market access. Do not overstate this in either direction: the effect exists and it is not the main driver of location.

Where it does bind is heavy industry, and that produces carbon leakage: if one region prices carbon and its partners do not, emissions-intensive production migrates and global emissions do not fall. The European Union's answer is the Carbon Border Adjustment Mechanism, which began with a transitional reporting phase in October 2023 and moves to a definitive regime requiring importers to buy certificates reflecting embedded emissions in covered goods, initially cement, iron and steel, aluminium, fertilisers, electricity, and hydrogen, phasing in as free allowances under the EU emissions trading system are withdrawn.

You should be able to argue both sides. Supporters say it is the only way to price carbon domestically without exporting the emissions, and that charging imports the price domestic producers already pay is a level playing field, not a tariff. Opponents, including India, Brazil, South Africa, and China, call it protectionism in climate clothing, argue it shifts the burden onto developing exporters least responsible for cumulative emissions, and question its WTO compatibility. Both are serious positions, turning on whether the charge genuinely equals the domestic carbon price and how credit is given for carbon already paid at origin.

Underneath sits an accounting question. Countries report production-based emissions, what is emitted within their borders, while consumption-based accounting attributes emissions to where goods are finally consumed. Roughly a quarter of global carbon dioxide emissions are embodied in traded goods, and for several wealthy countries consumption-based emissions exceed production-based ones, so part of a rich country's reported reduction can reflect importing what it used to make. The IMO's revised 2023 strategy targets net-zero emissions from international shipping by or around 2050.

Now the counterintuitive fact, well supported and almost universally misunderstood. For most foods, transport is a small share of the total carbon footprint. The large synthesis by Joseph Poore and Thomas Nemecek found that what you eat matters far more than how far it travelled, because production, particularly land use and methane from ruminants, dominates. Ocean shipping is extraordinarily carbon-efficient per tonne-kilometre; air freight is one to two orders of magnitude worse, which is why air-freighted perishables are the genuine exception. Intuitions about environmental impact based on distance are frequently wrong, so ask for the footprint breakdown rather than assuming local means low carbon.

Key idea: Pollution haven effects are real but modest outside heavy industry, carbon border adjustment addresses leakage while raising serious fairness objections, consumption-based accounting reassigns the quarter of global emissions embodied in trade, and transport is a small share of most food footprints.

Where the careers are

International business is rarely a job title. It is a specialization inside another function, and there are more such roles than students expect:

Path	What you do	Entry route
Trade compliance	Classify goods, determine origin, screen restricted parties, manage duties	Importers, brokers, law firms; customs broker exam
Global sourcing	Select and manage suppliers, negotiate, build landed cost models	Category analyst; supply chain certifications
Logistics and freight	Move goods, manage carriers, forwarders, customs flows	Operations roles at forwarders and carriers
Treasury	Manage currency exposure, hedging, cross-border cash	Corporate treasury or bank; finance credentials
Country management	Run a market, adapt strategy, manage local teams	After functional depth plus in-country experience
Development finance	Trade and development policy and lending	Graduate study, multilateral internships

On pay and prospects, use the source rather than a textbook number: the free BLS Occupational Outlook Handbook publishes median pay and projected growth by occupation, showing logisticians near eighty thousand dollars with growth well above average.

Three things genuinely help, and only one is a credential. A second language at working level, meaning a technical conversation rather than ordering dinner, opens doors no certificate does. Time actually living abroad is treated by employers as a proxy for adaptability. And quantitative fluency is the differentiator hiding in plain sight: most people in this field cannot build the landed cost model or the hedge comparison you built here, and the ones who can end up in the room where the decision is made. Credentials such as the customs broker examination supplement those three rather than substitute for them, and policy volatility has made trade compliance and supply chain risk among the fastest-growing corners of the profession. Keep using the free tools in this course's reference list: everything you learned to compute here runs on public data.

Key idea: International business careers sit inside functions such as trade compliance, sourcing, logistics, treasury, and country management, and the three things that matter most are a working second language, real time abroad, and the quantitative fluency to build the models in this course.

Try it

You source apparel from three factories in a low-income country. An investigative report alleges blocked fire exits at one and unauthorized subcontracting to an unlisted facility at another. Your head of sourcing proposes terminating all three and moving country. Evaluate that proposal.

Answer: Terminating all three is the worst option. It ends your leverage precisely when you have it, removes income from workers who did nothing wrong, and fixes no blocked fire exit. It is also legally hollow: under emerging due diligence regimes the expected response to an identified risk is to address it, not disengage. Instead: immediate safety triage at the factory with blocked exits, funded by you, since a supplier told to fix problems on its own margin usually cannot. Then treat the unauthorized subcontracting as the more serious systemic finding, because it means your visibility is broken; require disclosure of all production locations as a contract condition. Then change the purchasing behavior that generated it, since last-minute order changes and unrealistic lead times are the usual cause. Finally, join the binding industry safety arrangement rather than commissioning your own audit. Reserve exit for a supplier that refuses remediation, and if you exit, do it with notice and a plan for the workers.

Common misconceptions

• "The economists' argument is that sweatshop conditions are fine." It is a claim about the alternatives available to workers, and says nothing in defense of unsafe buildings or wage theft.
• "Labor advocates want boycotts." The demands after Rana Plaza were for binding, funded, independently inspected safety standards.
• "Social audits catch the serious problems." The Rana Plaza factories had been audited. Audits miss structural safety, hidden subcontracting, and coached workers.
• "Firms relocate mainly to escape environmental rules." Pollution haven effects are detectable but modest outside the most energy-intensive industries.
• "Buying local always lowers the carbon footprint." For most foods production dominates; ocean shipping is carbon-efficient and air freight is the real exception.
• "Carbon border adjustment is simply protectionist, or simply fair." Both are seriously argued, and the answer turns on whether the charge truly equals the domestic carbon price.

Recap

• Export manufacturing generally pays more than local alternatives, and research links Bangladeshi garment work to higher girls' schooling and delayed marriage.
• Unsafe buildings, wage theft, and suppression of organizing are violations rather than the price of development, and Harrison and Scorse found targeted pressure on high-margin exporters raised wages with modest job losses.
• Voluntary codes failed at Rana Plaza; the binding post-collapse agreement outperformed the voluntary one because enforceability changes behavior.
• Responsibility has shifted toward legal duties: the UN Guiding Principles, OECD Guidelines, due diligence laws, and forced labor import bans.
• Carbon border adjustment addresses leakage while raising real fairness objections, and roughly a quarter of global emissions are embodied in trade.
• Careers sit inside trade compliance, sourcing, logistics, treasury, and country management, where a second language, time abroad, and quantitative fluency matter more than any credential.

Sources

1. International Labour Organization. (2024). International labour standards and fundamental principles and rights at work. ILO. ilo.org
2. Organisation for Economic Co-operation and Development. (2023). OECD guidelines for multinational enterprises on responsible business conduct. OECD. oecd.org
3. European Commission. (2025). Carbon border adjustment mechanism. Taxation and Customs Union. europa.eu
4. Ritchie, H. (2020). You want to reduce the carbon footprint of your food? Focus on what you eat, not whether your food is local. Our World in Data. ourworldindata.org
5. U.S. Bureau of Labor Statistics. (2025). Logisticians. In Occupational Outlook Handbook. BLS. bls.gov