International Trade: Why Countries Trade and What Decides the Pattern

Countries trade because what a thing costs differs from place to place, and cost here means what has to be given up to make it. Being worse at everything is not the same as being equally worse at everything. A country can be worse at making everything and still gain. The gap between those costs decides the pattern, not the levels.

Two ideas do the work, and both are already settled. One is how a price forms out of supply and demand. The other is that the real cost of doing anything is the next best thing given up in order to do it, set out under opportunity cost and used here exactly as it stands. Trade adds one thing to both. Until now every buyer and every seller stood inside the same borders, and now a second place has its own workers, its own soil, its own machinery, and its own answer to what a thing costs. Comparing those two answers is the whole subject.

If Sankhya can build a machine itself, why would it ever buy one from Marut?

Start with the question in its rudest form. Almost nobody says it out loud, and almost everybody carries it. Sankhya has workers. Sankhya has workshops. Sankhya can build a machine. So why hand money to somebody in Marut for a machine Sankhya could have built at home? Put like that, buying from abroad looks like an admission of weakness, and a great deal of loose talk about trade rests on exactly that feeling.

Here is the everyday version first. A sweet maker can cook. Nobody disputes it. She can also, at a stretch, cut her own hair. On a Sunday when the wedding season is on and there are four orders of sweets to finish for paying customers, she does not cut her own hair, and the reason is not that the barber is a better cook than she is. The reason is that the hour spent on her hair is an hour of sweets not made. The barber charges less than that hour is worth to her, so she pays and gets on with the sweets. Nothing in that decision required the barber to be better at anything.

The answer to why a country buys from abroad is never that the other country is better at making the thing, and a reader who believes it is will not understand a single trade dispute they ever read about. The wrong answer forces a conclusion. If trade happens because the other side is better, then a country that is better at everything should buy nothing, and a country that is worse at everything should sell nothing and would have nothing to offer anybody. Neither of those describes any place on earth. The wrong answer breaks immediately, and it breaks in a way that matters. Most of the anger in trade politics comes from people applying it faithfully.

What does cost mean in this argument, and what does one machine cost in each place?

Cost here does not mean a price on a label. Cost means the other work that had to be stopped. Counting a cost that way comes from the earlier work on marginal costThe cost of producing one more unit of something, counted as the extra resources that unit swallows rather than as an average spread over everything already made. and what an hour or a worker gives up when it is put to one use rather than another. Nothing about it changes when the two uses sit in different countries. The only new thing is that the comparison now crosses a border.

So take the two places. In the Republic of Sankhya, one worker-yearOne worker employed for one full year. A unit of labour used to compare what different places get out of the same amount of work, and nothing more than that. of work produces either 100 quintalsA unit of weight used for farm produce in India and several other places. One quintal is one hundred kilograms. of onions or 2 machines. In Marut, one worker-year produces either 40 quintals of onions or 4 machines.

Now do the division. The division is the whole subject. If a Sankhya worker-year can be spent on 100 quintals or on 2 machines, then getting 2 machines means not getting 100 quintals, so one machine costs 50 quintals of onions never grown. In Marut, 4 machines means not getting 40 quintals, so one machine costs 10 quintals. The same machine costs 50 quintals in one place and 10 in the other. Marut makes machines more cheaply in the only sense that matters here, and grows onions worse in every sense that could be named.

Run the same division the other way round. A cost always has two faces. A quintal of onions in Sankhya means 2 divided by 100, or 0.02 machines forgone. In Marut a quintal means 4 divided by 40, or 0.10 machines forgone. Onions are five times dearer in Marut counted in machines, exactly as machines are five times dearer in Sankhya counted in onions. The two statements are one statement seen from two ends.

What is the difference between being better at a thing and giving up least to make it?

Skill and cost are two different measurements, and the difference between them is the heart of the matter. The two countries set beside each other on the raw numbers read honestly as follows. Sankhya gets 100 quintals of onions out of a worker-year and Marut gets 40, so Sankhya is straightforwardly better at growing onions. Sankhya gets 2 machines out of a worker-year and Marut gets 4, so Sankhya is straightforwardly worse at building machines. The phrase absolute advantage describes exactly that pair of statements. One place gets more out of the same resources than another does, and absolute advantage compares levels between the two places.

Now read the other row, the one nobody looks at. A machine costs Sankhya 50 quintals and costs Marut 10. The cost row is not a comparison between countries at all in the first instance. The cost row compares two things inside each country, both of them possible uses of the same worker-year, and only then are the two answers set side by side. The first row asks who produces more. The second row asks what each of them destroys in order to produce it.

A country that is worse at making everything still has something it gives up least to make, and that is arithmetic rather than encouragement. Nothing kind is being said about the weaker producer. The claim is mechanical. Of any two ratios that are not equal, one is smaller. One smaller ratio is all the argument needs. Sankhya could be twice as good as Marut at everything, or half as good at everything, and the same test would run: divide the two things a worker-year can do, in each place separately, and compare the answers. The argument is neither new nor anonymous, incidentally. The argument belongs to David Ricardo, and his name travels with it the way it does with any borrowed frame.

The everyday version keeps this honest. A senior surgeon may genuinely be faster at typing than the person hired to type for her. She still does not type her own notes, and the reason is not politeness. An hour of her typing is an hour of surgery not performed, and an hour of the typist's is an hour of nothing much else. She is better at both jobs and gives up far more to do one of them. The hospital gets more surgery and more typing done by ignoring who is better and looking at what each hour destroys.

Suppose each side follows the cost row. Give each country 100 worker-years, a round number that keeps the arithmetic clean. Under autarkyA state in which a country supplies everything it uses from inside its own borders and trades with nobody. Used here only as the baseline to compare against., suppose each splits its labour half and half. Sankhya then has 5,000 quintals and 100 machines. Marut has 2,000 quintals and 200 machines. Between them that is 7,000 quintals and 300 machines. Now let Sankhya put every worker-year into onions and Marut put every worker-year into machines. Sankhya has 10,000 quintals and Marut has 400 machines. The two of them together now hold 10,000 quintals and 400 machines, or 3,000 more quintals and 100 more machines out of exactly the same labour.

Which terms leave both sides better off, and where does that range come from?

Knowing who should make what is only half the work. The other half is the price. If Sankhya is going to grow onions and buy machines, somebody has to say how many quintals of onions a machine trades for, and that number is not handed down by anybody. The number gets settled between the two sides, and only some settlements are worth agreeing to.

Work out the refusals. The refusals define the range. Sankhya will not pay more than 50 quintals for a machine. Why would it? At 51 quintals it would put a worker-year into machine building at home, give up 50 quintals, and be a quintal ahead. So 50 is a ceiling, and it is a ceiling made of Sankhya's own cost rather than of anyone's goodwill. A machine Marut builds could have been 10 quintals of its own onions instead. Marut will not accept less than 10 quintals for one. So 10 is a floor, made in exactly the same way.

The range of terms that leaves both sides better off runs from 10 to 50 quintals per machine, and it exists because the two costs differ, so its width of 40 quintals is precisely the size of that difference. If the two costs had been equal there would be no range at all, only a single number at which neither side gains anything and there is nothing to divide. The width of the band is the boundary of the whole argument, and worth holding on to.

Put the terms at 25 quintals a machine, inside the band, and count the gain on both sides for one machine. Sankhya pays 25 quintals for a machine that would have cost it 50 quintals to build, so Sankhya is 25 quintals better off. Marut hands over a machine that cost it 10 quintals of onions forgone and receives 25 quintals, so Marut is 15 quintals better off. Add them: 40 quintals, exactly 50 less 10. The whole gain available on a machine is the difference between the two costs, and the terms of trade decide nothing except how that fixed 40 quintals gets split.

Now push the terms outside the band. A range whose edges have never been tested is just a claim. Offer 5 quintals a machine. Sankhya loves it: it pays 5 for something that would have cost it 50, a gain of 45 quintals. Marut walks away, and the arithmetic of the walking away is exact. A machine cost Marut 10 quintals of onions it did not grow; it is being handed 5 quintals; it is 5 quintals worse off than if it had simply grown onions with that worker-year and never built the machine at all.

Now offer 60. Marut is delighted, receiving 60 quintals for a machine that cost it 10, a gain of 50 quintals. Sankhya now refuses, and again the arithmetic is exact. Sankhya is paying 60 quintals for a machine it could have produced by giving up 50, and that leaves Sankhya 10 quintals worse off than staying home. At every price outside 10 to 50 one side is worse off than not trading at all. The band is a real boundary, not a convention somebody agreed to.

Where inside the range do the terms actually land?

A lot of teaching quietly overreaches at exactly this point. Everything above establishes one thing and one thing only: that between 10 and 50 quintals a machine there is a deal both sides prefer to staying home. The argument does not say the terms will be 25. Nor does it say 20 or 45. The argument produced a range and stopped.

Three things settle the number inside the range, and the argument holds none of them. Demand is the first, meaning how badly each side wants what the other has, and demand is set out under how a price forms. The second is bargaining powerThe ability of one side of a negotiation to hold out longer or walk away more cheaply than the other, which is what usually decides where inside a feasible range a price settles., meaning who can wait and who cannot. The third is a rival seller. Marut's ability to hold out for 45 collapses the moment somebody else will supply a machine for 20. None of that is arithmetic. All of it is circumstance.

The argument establishes that a gain is available and says almost nothing about who captures it, and treating a range as though it were a prediction is the most common way this material gets oversold. The wedding hall again. Knowing that a booking is worth somewhere between eighty thousand and two lakh rupees to the parties involved establishes that a deal exists. Knowing the band does not give the price, and anybody who claims it does has slipped from arithmetic into fortune telling. Two of the three things that decide the split are not economics at all, and their absence is not a flaw in the argument. The absence is the argument's actual size, and knowing the size of a tool is part of knowing the tool.

What does this argument leave out?

Set the boundary out with the same care as the argument. Take the argument for more than it is and the error shows up in public. Start with the shape of it. Two goods. Two countries. One period. Labour as the only input. No transport cost, no tariffA tax charged on a good as it crosses a border. What a tariff does, who collects it and who ends up paying it are taken up separately., no shipping delay, no exchange rateThe price of one currency in terms of another. An exchange rate turns a price quoted in one place into a price payable in another, and exchange rates are covered separately. and no money at all. Every one of those absences was a simplification chosen to make the cost gap visible, and every one of them is a real feature of a real trade.

Then the three that matter most. First, the argument counts a country as a single body with a single interest, and no country is that. When Sankhya moves its workers out of machine shops and into onion fields, the country as a whole holds more of both goods and the people who ran the machine shops hold considerably less. The argument shows a bigger total. The argument shows nothing at all about the distribution of that total, and the loudest politics of trade lives in exactly that gap.

Second, moving from one pattern to another is not free and is not instant. A worker-year is not a coin to be slid from one pile to another. Retraining takes years, workshops sit empty, and a district built around machine building does not become a farming district because a ratio said it should. The argument compares two end states and prices the journey between them at nil.

Third, a pattern can lock in. Specialising in the thing a country currently gives up least for says nothing about whether that thing will still be worth making in twenty years, or whether the skills it builds lead anywhere. A country that gives up least to grow onions today may be a country that never learns to build anything. The argument as stated has one period in it and no opinion on the matter.

None of these is a refutation of the argument, and every one of them is a limit on what it can be asked to settle. The cost gap is real, the gain is real, and the arithmetic does not stop being right because it is narrow. The argument answers one question, and several other questions are in the room.

What does an analyst actually do with a cost gap?

None of this stays theoretical for long, and the practical use of it is narrower and more useful than most readers expect. An analyst covering a company does not compute opportunity costs for countries. They sort a company's inputs and outputs into two buckets, the ones traded across borders and the ones that are not. The sort changes how every other number behaves.

A cost base that never crosses a border is governed by domestic conditions. Local wages, local rent, local power tariffs, the domestic policy rate feeding into the cost of borrowing. A domestic cost base is not comfortable, but it is legible, and it moves with conditions that can already be read. A traded cost base is governed by something else entirely. A traded cost base moves when a currency moves, when a barrier is imposed somewhere, when a shipping route reroutes, and when a supplier's own government changes a rule. Two companies with identical margins today behave completely differently the moment either of those worlds shifts, and the only way to know which is which is to have done the sort.

The productivityHow much output a given quantity of input produces, usually stated per worker or per hour. Productivity measures the level of output, and levels do not decide a trade pattern. comparison is the second use, and it is used as a question rather than an answer. If a company's competitor abroad gives up far less to make the same thing, that gap does not disappear because the company works harder. The gap is a structural feature of where each of them sits, and it tends to show up eventually in prices, in margins, or in which of them is still making that product in five years. An analyst treats a cost gap as a thing to be explained rather than a thing to be judged. A gap with a reason behind it behaves very differently from one that arrived by accident.

A lender does the same sort for a different purpose. When a borrower's revenue is earned abroad and its costs are incurred at home, or the other way round, the loan carries a risk that has nothing to do with how well the business is run. A household reading its own budget is doing a smaller version of the same thing. The part of the monthly spend that comes from an imported good behaves differently from the part that comes from the vegetable market down the road, and knowing which part is which is the whole skill.