Trade Agreements: The Forms and What Each Commits To

A trade agreement is a commitment between countries about how each will treat the other's goods, services or investment. The forms differ in depth: how much comes down, how much is bound against future rises, and how much home rule-making is set aside. The most valuable thing most agreements do is not lower a barrier but make raising one hard.

Two ideas already in place do the load bearing, and neither gets rebuilt. One is what a barrier is and who ends up bearing the cost of a barrier, settled earlier and borrowed intact. The other is that the direction of trade follows from what each side gives up rather than from who is better at making things. A third object sits on top of both, and it is the new one: a country can write down, in advance, what it will and will not do to a barrier, and hand that writing to somebody else. A written commitment of that kind is strange. The whole value of the commitment comes from taking something away from the country that signed it. The forms differ in how much they take away, and in why a country would want anything taken away at all.

What is a country actually agreeing to when it signs a trade agreement?

The everyday version carries almost the whole idea. Start there. A shopkeeper wants to open a small sweet shop in a rented room, and will spend Rs 2,00,000/- on a counter, a display case and a chimney, none of which comes back out of the wall once the room is given up. The landlord says he has no plans to raise the rent. He is telling the truth. He still may raise it next month, and he knows it, and so does the shopkeeper. Now he signs a paper fixing the rent for three years. Nothing about the room changed. Nothing about the rent today changed. The set of things the landlord may do to the shopkeeper changed, and that change is the only reason the chimney gets built.

A trade agreement is a constraint a country accepts on its own future behaviour, and every form of agreement is a different amount of that same one thing. Read the sentence slowly. The definition inverts the way most people picture the object. An agreement is usually imagined as a thing a country gets. An agreement is at least as much a thing a country gives, and what it gives is discretion: the freedom to change its mind later. Sankhya without any arrangement may set the tariffA tax charged on a good when it enters a country, set either as a percentage of the good's value or as a fixed amount for each unit that crosses. on an imported machine at nil next year, at 20 per cent, or at any level its own process will pass. Sankhya inside an arrangement that binds the machine tariff at 10 per cent may not go above 10 per cent, whatever it later wishes.

Two things sit outside the definition. The barrier itself need not be low. An arrangement that binds a tariff at 25 per cent is a perfectly real agreement, and in one important respect it does more work than an arrangement that drops the tariff to nil and says nothing about tomorrow. An agreement need not cover much, either. A single line about a single good is an agreement. The forms in the next section are all built out of two dials, how wide the coverage runs and how tightly the future is tied, and every named form is a setting of those two dials.

What are the main forms, and what does each one set aside?

Line the forms up by one measure only: how much of its own decision-making a country hands over. The shallowest form lowers a barrier on a named list of goods and leaves everything else exactly where it was. Sankhya might agree that a particular grade of machine from Marut enters at 10 per cent instead of 20 per cent, and Sankhya still decides, entirely on its own, what happens to every other good and to every other country. A commitment of that shape is a preference on some goods, and a preference on some goods is the smallest real agreement there is.

Step along the axis and the coverage widens. A free trade area removes barriers on most goods moving between the parties, so the list stops being a list and becomes a default with exceptions. Sankhya still sets its own barrier facing every country outside the arrangement. Keeping that barrier keeps a whole instrument in Sankhya's hands, and the instrument matters more than it sounds. A customs union takes that instrument away: the parties adopt one shared barrier facing the outside world, and from that point Sankhya cannot set its own tariff on a machine from anywhere. A common market goes further again and lets factors of productionThe things used to make anything at all: labour, land, capital and the organising that combines them. Distinct from the goods and services they end up producing. move as well as goods, so workers and money cross as freely as machines do. Deepest of all, the parties coordinate domestic rules, aligning standards and licensing so that a good approved in one place is accepted in the other without a second inspection.

Each step along that axis buys certainty by handing over discretion, so the axis is a trade and not a ladder a country is supposed to climb. Most treatments get this part wrong by accident, simply through the order they present things in. Listed shallow to deep, the forms look like grades, and a reader concludes that deeper is more advanced and therefore better. Nothing in the mechanism says so. A country that keeps its own external tariff has kept a tool it may need; a country that has pooled it has bought a certainty its traders can plan on. Which of those is worth more depends entirely on circumstances no general account can see.

Set out as rows, the same axis reads more precisely. Each form has two descriptions: what it removes, and what it leaves the country holding.

What is the difference between lowering a barrier and binding it?

Here is the distinction everything turns on, and it is worth slowing down for. The way agreements are usually reported makes the distinction invisible. Sankhya can cut the machine tariff from 20 per cent to nil tomorrow morning, on its own, with nobody's agreement, through its own process. Sankhya can put it back to 20 per cent the morning after, on its own, through the same process. Lowering a barrier is an act. A lowering happens on a date and can be undone on another date.

Binding is a different kind of thing. Binding is Sankhya undertaking, to Marut, that the machine tariff will not go above a stated level for a stated period. The level that is bound and the level actually charged need not be the same: Sankhya may charge nil while being bound at 10 per cent, and it is then free to move anywhere from nil up to 10 per cent without breaking anything. Sankhya cannot go to 20 per cent. Lowering changes the price a buyer pays this year, and binding changes what a buyer can plan on, and only the second one survives a change of mind.

Why that difference is worth money becomes clear as soon as it is plain who is deciding what. A buyer placing one order this month cares only about the tariff this month, and for that buyer a lowering is worth exactly as much as a binding. A buyer building a plant cares about the tariff for as long as the plant runs. The plant is bought once and paid for out of many years of output, so the number that decides whether it gets built is not the landed costWhat a buyer actually pays to have a good standing in its own warehouse: the seller's price plus freight, insurance, duty and every other charge added on the way in. today but the worst landed cost that can turn up before the payback periodHow long a project takes to return the money that was put into it, counted in years from the first spending. A rough test of how far ahead a decision has to be right. ends. A lowering says nothing at all about that number. A binding names it.

Why would a country accept a constraint on itself?

Put like that it sounds like a bad bargain, and the puzzle is real. Sankhya wants the freedom to raise the machine tariff if its own machine builders are in trouble in year four. Why sign that freedom away? The answer is that the freedom is exactly what the other side is buying. Marut's machine builders will not put money into a Sankhya-facing production line on the strength of a lowering that can be reversed. The machine builders will put money in against a level that holds. The discretion and the certainty are the same object seen from two sides, so Sankhya cannot sell the certainty without giving up the discretion.

A constraint nobody can escape is worth more than an intention everybody shares, and that is the reason a country pays for its own hands to be tied. The household version is the landlord again, and it is worth noticing what he gets out of signing. He does not sign for nothing. He signs because a shop with a chimney and a display case pays rent for three years and a shop with a folding table leaves in four months. His constraint is what buys him the tenant he wants. The same shape appears in the sweet-maker's own dealings: a caterer who will commit a price for the whole wedding season gets the order that a caterer quoting week by week does not.

The same logic has appeared once already, in a different subject and with different machinery. A central bank saying in advance what it intends to do with its policy rate is doing precisely this: giving up the freedom to surprise, in order to make the expectations of everybody else settle down. Forward guidanceA central bank stating in advance what it expects to do with its policy rate, so that people making decisions today can plan around it rather than guess. works for the same reason a bound tariff works, and fails for the same reason too. The value of an undertaking is exactly equal to how hard it is to walk away from. One idea, two subjects, and recognising it in the second place is most of what makes it stick.

Which of two arrangements can a plant actually be built on?

The point only lands when the cheaper option loses. Work it with figures. The Republic of Sankhya has two arrangements on the table with Marut, and both concern the same machine. The machine lands from Marut at Rs 50,000/- before any tariff is added. Sankhya's standing tariff on that machine, the one that applies with no arrangement in place, is 20 per cent, so the machine currently lands at Rs 60,000/-.

Arrangement one drops the machine tariff from 20 per cent to nil, and expressly reserves Sankhya's right to restore the standing 20 per cent at any time. Arrangement two leaves the machine tariff at 10 per cent and binds it there for ten years. So arrangement one gives a landed cost of Rs 50,000/- today and arrangement two gives Rs 55,000/-. Arrangement one is Rs 5,000/- a machine cheaper right now.

Now put a decision in front of them. A Sankhya household business is deciding whether to build a plant that runs on these machines, taking 20 machines a year through a five year build-out. The build-out swallows 100 machines in all. The plant is financed, and the arithmetic behind the loan works only while a machine lands at Rs 57,500/- or less. Over 100 machines that is a machine budget of Rs 57,50,000/-. Above that the plant does not service its borrowing. The ceiling is the test, and the test is not about today. The test is about every one of the five years.

A decision that has to hold for five years is tested against the worst case rather than the current price, so arrangement one is the cheaper arrangement today and the one the plant cannot be built on. Look at what makes the difference. Arrangement one hands the plant a range, from Rs 50,00,000/- if the tariff stays at nil to Rs 60,00,000/- if Sankhya restores the standing 20 per cent, and the top of that range is Rs 2,50,000/- above what the loan can carry. Arrangement two hands the plant a single number, Rs 55,00,000/-, and that number sits Rs 2,50,000/- under the ceiling. A lender cannot lend against a range whose upper end breaks the plan. A lender can lend against a number.

How does an agreement change who a country buys from?

Everything so far has been about how much a country trades. An agreement also changes who it trades with, and that second effect is easy to miss because the first one is what gets counted. Work it on the machine again. Sankhya has two possible sources: Marut, whose machine costs Rs 50,000/- before any tariff, and a supplier in a third country outside the arrangement, whose machine costs Rs 48,000/- before any tariff. Note which of them is the more efficient producer: the one outside, by Rs 2,000/- a machine.

Before any arrangement exists, Sankhya's standing 20 per cent falls on both. Marut's machine lands at Rs 60,000/- and the outside supplier's at Rs 57,600/-, so Sankhya's buyer takes the outside machine. Now Sankhya and Marut sign an arrangement under which Marut's machines enter at nil. Marut's machine lands at Rs 50,000/-. The outside supplier is not a party, so the standing tariff still falls on it and its machine still lands at Rs 57,600/-. The buyer switches to Marut.

The machine now comes from the less efficient producer, and it comes at a lower landed cost, and both of those are consequences of the same tariff wall standing for outsiders and falling for a party. What did and did not happen there is worth stating plainly. Sankhya's buyer is better off, paying Rs 50,000/- where it paid Rs 57,600/-. The machines are cheaper, so more of them cross a border than before. And the machine is being built in the place that uses more to build it. The shift is real and not a rounding effect. The switch is a mechanism, not a verdict: an agreement moves trade between partners as well as expanding it, and how those two movements net out in any actual case depends on figures no lesson can supply.

What does a trade agreement not do?

An agreement lowers one cost of trading and leaves every other condition of trade exactly where it stood, so the list of things it does not do is longer than the list of things it does. An agreement does not make trade happen. If nobody in Marut wants Sankhya's onions at any price Sankhya will accept, an arrangement covering onions produces no onion trade whatsoever, and a signed text with no shipments under it is a perfectly ordinary outcome. An agreement does not create comparative costWhat a country gives up to make one thing rather than another, compared with what a second country gives up for the same choice. Comparative cost is the reason a trade pattern forms at all.. The reason Sankhya buys machines rather than building them is what a machine costs Sankhya in onions not grown, and no text alters that arithmetic by a single quintal.

An agreement does not protect the people inside a country from the shift. When Sankhya's buyers switch from home-built machines to Marut's, Sankhya as a whole holds more machines for less, and the workshops that built machines in Sankhya hold considerably less work. The agreement is about the border, and it is silent on the workshops. And it does not bind anybody who has not signed: the outside supplier in the last section still faces Sankhya's standing 20 per cent, not because it has been punished, but because it was never a party to anything.

An agreement lowers a cost of trading, and every other thing that has to be true for trade to happen still has to be true. That is not a small claim, and it is the one that keeps a reader from over-reading a signed text. Think of it as an unlocked door rather than a delivery. The door being open is necessary for anybody to walk through it, and it does not mean anybody wants to.

What does an analyst actually check in an agreement?

Not whether an agreement exists. Existence is the least informative fact available, already reflected in every price and every headline. An analyst reads for three narrower facts, and all three sit in the text rather than in the coverage of it. First, which barriers are bound as against merely applied at a lower level today. Only a bound barrier can be carried into a projection. Second, for how long each binding runs. A binding that lapses inside the projection period is not a binding for that purpose at all. Third, what the text allows a party to do in defined circumstances. A level that holds except in stated conditions holds only until those conditions arrive.

A bound level with a stated end date is a fact a five year plan can rest on, and an unbound level is an observation about this year that has to be re-made every year. A lender does the same work with a different purpose. Where a borrower's cost base runs on imported inputs, the loan is exposed to whatever can happen to those inputs, and a bound tariff turns part of that exposure into a known number for a known period. The known number changes the covenant that can sensibly be written and the tenor that can sensibly be lent. An investor reading a company that depends on imported inputs is asking the same question one level up: is the input cost in this projection a bound number, or is it a current number wearing a projection's clothes?

The household version is the shop again, and it generalises perfectly. Before spending Rs 2,00,000/- on a chimney, the question is not whether the rent is low. The question is how long the rent is fixed for, and whether the paper says anything about what happens if the landlord's circumstances change. Everything an analyst does with a trade agreement is that question in a bigger costume.

Where would a reader find what a country has actually committed to?

Everything above is a shape, and a shape is not a text. A particular country's commitment on a particular good is a document with a date on it, and that document can be amended, extended or replaced without anybody's lesson noticing. So the only honest answer to the question is a place to go rather than a fact to carry, and the places exist and can be named.