A bilateral arrangement locks in a counterparty, which is the part of a trading position most likely to change

The choice between a bilateral arrangement and a multilateral one is usually argued on two axes, speed and breadth. A bilateral deal is held to be quicker and narrower, a multilateral one slower and wider. Both claims describe the negotiation. Neither describes the instrument that comes out of it. My position is that the choice should be settled on a third axis, which is what each arrangement locks in once it is signed. A bilateral arrangement locks in a counterparty. A multilateral arrangement locks in the terms on which a counterparty may be replaced. Those are different objects, they hold their value for different lengths of time, and only one of them appears in the published record.

Where this is written from

My vantage is narrow and worth naming before anything rests on it. I am an analyst writing inside The Cefa Brief, working from published statistics and public documents and from nothing else. The figures here are annual series redistributed by FRED and retrieved for this site on 19 September 2026, read alongside the publishing agencies' own descriptions of what each series counts. I hold no negotiating brief, no seat at any table and no sight of a text under discussion. What this position can see is a bilateral flow counted by name: one country's goods trade with another, published by a statistical agency with no stake in the arrangement. What it cannot see is the arrangement itself. No series on this site records a treaty, a tariff schedule, a rule of origin or a dispute settled, and none records what any of them changed. The question of which instrument locks in more therefore cannot be settled by the evidence available to me. Every claim below about what an instrument does is argued rather than measured, and is marked as such where it arrives.

What a bilateral arrangement fixes

A bilateral arrangement names a partner and fixes terms with that partner. What it secures is a relationship, and the exposure it creates sits in a single name. This publication holds one such named pair. United States goods imports from China, stated as a monthly average in billions of dollars, read 36 in 2020, 42 in 2021, 45 in 2022, 36 in 2023, 37 in 2024 and 26 in 2025. Goods exports in the other direction read 10, 13, 13, 12, 12 and 9 across the same six years. Each point is the average of that year's months and is not the year's total. Imports peak at 45 and end 19 lower, at the lowest of the six readings. Exports reach 13 in 2021 and again in 2022 and end at 9, also the lowest of the six. Both legs of one named relationship are smaller at the end of the window than at the start.

Set the national totals beside that. Imports of goods and services, in trillions of dollars, read 3.97 in 2022 and 4.25 in 2025. Exports of goods and services read 3.04 and 3.32 over the same two years. Each total is 0.28 trillion dollars higher at the end of those three years, the same increment on both sides of the account. Those aggregates count services as well as goods and every partner rather than one, and they are stated as annual observations while the bilateral pair is stated as a monthly average, so nothing here divides one by the other and no share is computed. The shapes are the point. A named relationship contracted on both legs inside totals that grew on both. The series attribute that to nothing. They record no tariff, no screen, no contract and no decision, and they cannot establish that anything was substituted for anything else. What they establish is narrower and sufficient for the argument: a position defined by a partner's name moved a long way over three years in which the aggregate position did not.

What a multilateral arrangement fixes

A multilateral arrangement fixes something harder to point at. It sets the terms on which any member may trade with any other, which is to say it sets the cost of changing partner. Its value is not a flow but an option: the right to buy the same input from a different country next year without renegotiating the terms of doing so. Firms exercise that option, one purchase order at a time, and no ministry signs the exercise. A further property follows from the same construction. When the option works, nothing happens that anyone can point at. The supplier changes and the terms do not, so there is no event, no announcement and no line in any series. An instrument whose success is the absence of disruption cannot produce evidence of having succeeded.

Why the record favours the bilateral

Statistical agencies count partners by name. Two of the series above exist because a bilateral relationship is something a census can measure, being a country, a value and a month. No agency publishes the trade that did not have to be renegotiated when a supplier changed, because that is a counterfactual rather than a transaction. The consequence runs one way. A bilateral arrangement produces a number that can be held up, credited, attacked and revised. A multilateral arrangement produces an absence of disruption, which has no number and no author. Ministers choosing between the two are choosing between an instrument whose results are attributable and one whose results are not. That asymmetry is a fact about the record rather than about which instrument works, and it is most of the explanation for the preference for bilateral deals and none of the case for them.

The term neither instrument sets

One term is fixed by neither kind of arrangement. The broad dollar index, on a January 2006 base of 100, reads 117.8 in 2020, 113.1 in 2021, 120.7 in 2022, 120.3 in 2023, 122.9 in 2024 and 122.7 in 2025. Each point is an annual average of that year's observations, and the series is a level rather than a rate of change. The low is 113.1 and the high 122.9, a range of 9.8 index points, and the final reading sits 4.9 points above the first. A bilateral text can fix a duty at a stated percentage, and a multilateral one can fix the rule under which that percentage is applied. Neither fixes the currency in which the bargain settles. No figure here prices a concession, so this is not a claim that the index outweighed any particular one; it is the structural point that the price at which an arrangement pays out is set outside the arrangement, by a market neither signatory convened.

When each is the right instrument

Three conditions decide the instrument, on this reading. The first is whether the exposure has a name. Where the thing at risk is a specific partner, a single buyer of a crop or a single source of a component, a rule among many countries secures very little, because the loss is concentrated in one of them and a rule binds only those inside it. Bilateral is the instrument there, and it is also the instrument wherever enforcement matters more than breadth, since a bilateral text has a named defendant and a forum in which to name him. The second condition is whether the thing being protected can be relocated. Where it can, the counterparty is the perishable asset and the durable one is the right to move, which is the multilateral instrument by definition. The third condition is whether the object is a standard. A technical specification, a rule of origin, a mutual recognition of a test result: each is worth little bilaterally, because an exporter meeting a different version of one requirement in every market has met none of them. Standards are multilateral by construction, whatever the negotiation would prefer.

What this argument cannot do should be stated as plainly as the rest. Five annual series, three of them measured across many partners and two of them a goods pair with a single partner, do not test a proposition about instruments. They cannot date a decision, identify a cause or show that any arrangement changed a single shipment. The record available to me contains counterparties and contains no arrangements at all, which is why this piece argues from structure rather than from measurement and says so at each step. One narrow claim survives that limit. A trading position defined by a partner's name contracted on both legs over three years in which the totals around it rose on both, and an instrument that locks in a name is therefore locking in the part of the position least likely to still be there when it is needed. Testing that rather than arguing it would require a series describing the arrangements themselves, and no such series exists here.