The Poverty of Methodological Nationalism

Who exchanges?

China exports electric vehicles to Europe. Germany runs a surplus. The United States loses jobs to Mexico. Three sentences of a kind one reads every day, each with a grammatical subject, and none of those subjects has ever bought or sold anything. Buying and selling are done by persons, and by firms belonging to persons. A country can levy taxes, wage war, close its borders and sign treaties, all of which is real and consequential. Exchange it cannot. An exchange is a transaction between at least two wills, both of which must consent, and a country has no will. It has a procedure in which the wills of some persons prevail over those of others.

This is not pedantry, because the choice of accounting unit decides which questions can be asked at all. Whoever installs the nation as the acting subject has claimed, silently, that it makes sense to add the advantages and disadvantages of every inhabitant of a territory into a single number and to treat that number as a measure of success. The claim may be right. It is a claim and not an observation, and in the trade literature it is almost never marked as one.

Consider the baker. Each of us runs a permanent and growing deficit with his baker: we buy bread there for decades and sell nothing in return. Nobody demands corrective measures, and nobody asks how the bakery sector is draining household purchasing power. The very same number becomes a political object of the first rank as soon as the accounting boundary is drawn differently and all the inhabitants of a state territory are pooled. The transaction has not changed. The boundary has changed.

The working rule I use is methodological individualism, a term from Schumpeter since freighted with meanings I do not intend. Explanations begin with persons and their reasons; collective magnitudes such as the trade balance or the savings rate are results rather than starting points; and a statement about an aggregate counts as understood only once one can say who did what, and why. None of this denies that groups act. Duties are levied, borders policed, capital movements restricted, and the second half of this essay is about little else. What is denied is only that the group may be installed as the beginning of an explanation. It is the thing to be explained.

How far it reaches

The opposing position has a name, though not one from economics. Wimmer and Glick Schiller call methodological nationalism the assumption that the nation-state is the natural social and political form of the modern world, together with the practice of treating it as the self-evident unit of analysis. Of the three variants they distinguish, the one that matters here is naturalisation: state-generated categories adopted as given reality.

Note where the term was developed — in sociology and migration studies, and not in the discipline constituted by the boundary in question. The name says as much. “International economics” presupposes an inside and an outside before its first sentence is written, and there is no field called interpersonal economics, because there nobody suspects a special case.

The habit can be shown in the best literature, and that is the point. Irwin’s Free Trade under Fire has been for five editions the authoritative defence of open exchange, and its architecture follows the structure of its opponents: the United States in the world economy, trade and jobs, developing countries and open markets. The tactical reason is good, since Irwin means to refute the arguments as they are actually made. The price is high. Whoever adopts the national accounting unit has made the decisive concession before the first argument falls, and that this happens to the best book of its kind is an argument against the architecture rather than against the author. The pattern is sharper in Baldwin, whose analysis of the second unbundling states expressly that offshoring concerns individuals rather than firms or groups, and that comparative advantage has been denationalised: it is not Japan and Germany that compete but value chains, each combining the know-how of one place with the wages of another. In the policy chapters of the same book, industrial policy, national production sites and the broken social contract between national technology and national labour all return. The break runs through the middle of a single work. The purest form of the habit sits in international macroeconomics, where the current-account balance is derived from the optimising calculus of a representative household, one per country: the model does not dissolve the nation into persons but makes the nation into a person, and looks for all the world like microfoundation.

Nor is the framing merely academic. In March 2026 the United States Trade Representative opened investigations into structural excess capacity in sixteen economies, and among the evidence recited against several of them is that they run large or persistent goods trade surpluses. An accounting identity has become a legal standard: that the residents of a territory save more than they invest is treated as evidence of an actionable practice.

None of this is an accusation. It is not that trade economists are nationalist in sentiment, which would be false and would only provoke resistance. It is that a discipline adopts the categories in which its object is administered, and that this happens even to those who argue substantively for the opposite.

Why people exchange, and why no nation appears in the answer

Human beings do not differ from other species by any biological endowment acquired in the last ten thousand years. The genome has barely changed; prosperity has changed by orders of magnitude. What changed is not the equipment of the individual head but the way heads combine. Ridley’s formula captures it: ideas must prove themselves, and they carry further when they join with other ideas, because only recombination allows advantages that arose in different places and at different times to accumulate. Historically that recombination happened through the exchange of objects, since whoever acquires an object acquires with it the ideas embodied in it. Exchange moves not merely goods but knowledge. And exchange in turn arises from its advantages, chief among them the specialisation it permits. Read’s essay on the pencil showed where this leads: no single person on earth knows how to make one, the knowledge exists only distributed, and it comes together only through exchange.

One property of this story matters here, and it is easy to miss because nothing marks it. No nation occurs in it. Neither the emergence of specialisation, nor the recombination of ideas, nor the self-reinforcing circle between them requires a category above the person. One could insert such a category, but the explanation does not thereby improve. It only becomes coarser.

A qualification is owed, and it protects the argument rather than weakening it. It is not the case that every reason for exchange sits with the person and every reason against it with the group. Individuals decline advantageous-looking offers for good reasons of their own: because dependence on a sole supplier creates a quasi-rent the counterparty can appropriate once one’s specific investment is sunk, or because one does not hand a neighbour with whom one is in dispute a tool he can use against one. Several of the group-level grounds below are these same reasons scaled up. An essay that placed every reason for exchange with the person and every reason against it with the collective would have settled its outcome by its architecture.

What actually has to be justified

Groups are nonetheless unimpressed by the individual gains from exchange. They tax exchange between members and non-members, sometimes forbid it, and have done so for as long as there have been groups. It would be complacent to assume that so persistent a practice rests on nothing at all.

Two preliminaries fix the object. The first is an asymmetry. A person may at any time decline to exchange, and that requires no justification, no reason and no theory. What requires justification is exclusively the coercive restriction a collective imposes on its own members. This identifies the relevant group without presupposing it: it is the group that can compel. Whoever can evade the restriction without disadvantage is not subject to it; whoever cannot is a member in the sense meant here.

The second is five properties in place of a size. It would be natural to fix the size of the group — family, commune, province, country — but size gives the wrong information. What an argument needs is one or more of five properties: the capacity to compel, which binds members and raises the cost of exit; the capacity to compensate, that is, to redistribute internally and actually indemnify losers; market power, a share of the relevant market large enough that the group’s joint conduct moves the prices at which it deals with outsiders; the capacity to bind itself externally, so as to contract, to ally, and to answer for its members’ conduct; and a decision procedure, settling who decides, by what rule, and whether the decision can be revised. The properties are gradual rather than binary, and they can be held by different entities.

One further distinction decides the fairness of the enquiry. It is tempting to understand by the restricting group always a protection group, a community that keeps imports out. The same logic permits a compensation group, which bears the costs of adjustment instead of impeding imports; it needs the second property rather than the first, and it leaves exchange untouched. Whoever admits only the first form has handed the group its worst instrument and withheld its best. Both are carried below, and the payoff is that the targeting principle — treat a distortion at its source — arrives not as an imported maxim but as a plain comparison: which of the two achieves the stated goal at lower cost to its own members?

One tool is used implicitly throughout. Take an argument held compelling at the national level and lay it along a nested series — person, family, village, town, region, nation, continent, world — and ask on which steps it holds. Whoever thinks protection against foreign competition sensible and protection of one city against another absurd owes an account of which of the five properties changes between the steps. The test yields no verdict; it assigns a burden of proof. And it bites only if the series is carried past the nation and one then asks why one should stop precisely there. Within the series that question has no answer.

Eight grounds, examined

Each ground below goes through the same four steps. State it precisely. Ask whether it can really be a problem from the group’s point of view. Ask which instrument solves that problem most effectively. Only then judge whether a tariff or a prohibition is ever the right tool.

(a) Protecting a subgroup within

The commonest argument in practice: an industry, a region or an occupation is to be preserved from displacement.

It presupposes the capacity to compel. The goal is a pure internal redistribution from buyers to producers, and no external relation occurs in it at all, although the instrument used is an external one. If the group also possesses the capacity to compensate, the instrument becomes superfluous, because a transfer accomplishes the same thing without destroying the gains from exchange. The tariff raises prices for everyone in order to favour a few; the transfer also burdens, but it leaves prices alone.

The ground fails. It also owes an explanation that is rarely offered: why does this subgroup deserve protection rather than every other group affected by a change? Whoever protects against import competition but not against technical progress, shifting taste or domestic competition has not a principle but a selection.

(b) Security of supply

A serious argument, and it survives the examination — in transformed shape.

Remarkably little is presupposed. Stockholding, supplier diversification, offtake agreements and reserve capacity are a club good that can be organised voluntarily, needing neither coercion nor market power. The functionally right group is moreover often transnational: a secure supply of medicines as a purely national undertaking makes little sense, while as a European one, or as an association along the supply chain, it makes a great deal.

With that, the ground migrates out of trade policy altogether. It justifies an insurance club, not a tariff. This is a stronger result than a refutation, because the purpose remains legitimate and only the instrument changes.

One concession. Where capacity can be built only with long lead times and cannot be stored — production plant, trained personnel — supply cannot simply be bought when it is needed, and preserving existing capacity is genuinely an option. Even there the tariff is not the targeted means. A capacity premium costs less and is visible, whereas the tariff hides its costs in prices and preserves the capacity only as a side effect.

(c) Defence capability

The classical case, and the hardest concession. There are capabilities one cannot procure from a potential adversary, and goods whose absence in an emergency cannot be corrected by prices. This is the scaled-up version of not arming one’s counterparty, and it is as plausible for a group as for a person.

What it presupposes above all is the capacity to bind externally. It is a public good with strong free-riding, and its functionally correct group is almost never the nation but the alliance; the recurring burden-sharing quarrels within alliances are the free-rider problem in visible form. That alliances function although exit stands open shows at the same time that no capacity to compel is required here.

The real difficulty lies not in the warrant but in the extent. The ground is indefinitely elastic, because almost any good can be built into a chain of defence, and the security exception is among the most abused provisions in trade law. The usable test is therefore not whether a good is security-relevant, since nearly everything is, but whether in an emergency there would be time to procure it. Everything that can be stored, or built within months, falls back to (b).

(d) Bargaining leverage and reciprocity

Maintain one’s own restrictions in order to trade them against those of others. The architecture of the GATT and the WTO rests on this.

It presupposes the capacity to bind externally together with the capacity to compel, and it is the only ground on this list that genuinely requires the features of a state: without the ability to bind oneself credibly and to enforce the binding on one’s own members, there is nothing to negotiate.

The counter-position is strong. Unilateral liberalisation does one good regardless of what others do, and the notion that removing a restriction is a concession for which one must be indemnified is a mercantilist inheritance. In the accounting unit used here the point becomes sharper still. Concession, giving ground and bad deal are expressions that presuppose a national actor; drop the premise and the whole vocabulary of negotiation becomes unintelligible. Lowering one’s own duty on fish permits one’s own residents to buy fish more cheaply, so the operation chiefly benefits those who are said to be paying its price.

What remains is worth keeping. Reciprocity has worked historically, and the individualist explanation of why is instructive: it is not an argument in trade theory but a technique of collective decision-making, since negotiations mobilise exporters as a domestic interest group and set them against import-competing ones, and without that counterweight the concentrated interests prevail. Reciprocity justifies no restriction. It explains how restrictions can be dismantled when the decision procedure works against dismantling them.

(e) Infant industry

This ground cannot be waved away, because the account of exchange given above produces it. If one becomes better at a thing by doing it, then the order of specialisation is not a matter of indifference, and an early start can secure a lead a late starter never makes up.

The examination is nonetheless clear, in four steps. First, if the future gains exceed the initial losses, this is an investment project like any other and ought to be financeable; the argument therefore presupposes not a trade problem but a capital-market failure, as Baldwin put it in 1969, and the relocation matters because it shifts the jurisdiction. Second, even where the capital-market failure is real, the tariff is not the targeted means, since it burdens buyers in order to favour a producer, whereas a production subsidy or a loan achieves the same at lower cost to the group’s own members. Third, the classical test of Mill and Bastable demands two things: the protected activity must one day be able to stand without protection, and the discounted future gains must exceed the accumulated losses of the protection phase. Both conditions are formulable, both are testable, and in practice they are never tested. Fourth, the reach of the learning effect settles the jurisdiction. If it stays inside the firm there is no argument at all, since the firm bears the cost and reaps the return. If it spills over to other firms there is a genuine externality — but its reach is wherever the employees and the suppliers are, an agglomeration or an industry, and not the customs frontier.

An honest balance. The historical record is not unambiguous, and both readings are defended for Japan and Korea. The ground has not been refuted in theory; it demands a test almost never carried out; and it is regularly claimed for activities long since grown up.

(f) The speed and the cost of adjustment

The empirically weightiest ground of the present, and the one on which the defence of exchange most often fails.

Its starting point lies inside the account of exchange itself. Where open exchange shifts market shares towards more productive suppliers, the less productive drop out; selection has losers by construction. Autor, Dorn and Hanson showed that adjustment to a rapid import surge is regionally concentrated, slow, and for those affected lastingly costly. That is not an argument against exchange. It is an argument about speed, and about the distribution of burdens.

What it presupposes is the capacity to compensate rather than the capacity to compel, and with that the case is decided: the responsible body is the compensation group, not the protection group. Temporary protection can buy time, but it destroys precisely those gains out of which the indemnity would have to be financed, whereas compensation leaves the gains intact and redistributes them.

Here the strongest open objection must be admitted, because at this point the theoretical answer is better than the reality. Compensation programmes exist and work badly; American Trade Adjustment Assistance is the most-studied example. Whoever points to compensation points to an instrument superior in principle and weak in practice. The explanation lies again in the decision procedure. Compensation is visible and must be quantified, voted and defended, whereas protection is a concealed transfer that generates no budget line. The better option is the politically dearer one, and that asymmetry is itself a finding rather than an incidental.

(g) The optimal tariff and collective market power

Theoretically the most robust argument for a tariff there is. A sufficiently large group can, by levying a charge on its imports, depress the price at which outsiders sell to it, and pocket the difference.

It presupposes market power together with the capacity to compel, and it is the only ground that genuinely needs the group as a group. Translated into individual terms it is a buyers’ cartel — and that translation disposes of it. The gain arises not from anything created but from a transfer out of outsiders who took no part in the decision procedure. And it does not generalise: if everyone applies it, everyone loses. Johnson worked out the retaliation dynamics, and the outcome is an equilibrium of high tariffs in which, as a rule, even the party that began is worse off.

The case is instructive for a further reason. It is the purest example of the accounting unit determining the result: only if one adds up the advantages of the group’s members and leaves out the disadvantages of outsiders does the optimal tariff appear as an improvement at all.

(h) Fiscal grounds

Historically the principal reason for tariffs, and for some poor countries significant still: revenue can be raised at the border where there is no internal tax administration. Strictly this is not a ground for restriction at all but a choice of tax. The restriction is the unwanted side effect and not the purpose, and the test question is correspondingly different. Is the tariff a good tax? The answer turns on the collection costs of the alternatives, and it is clearly negative for a country with a functioning value-added tax, possibly positive for one without a revenue service.

The case is useful because it shows how strongly enforcement technology shapes policy — and because the same collection that makes the tariff possible produces the statistic in which the tariff later appears as trade policy.

Interim balance

Of eight grounds, one fails without remainder (a). Two survive in transformed shape and migrate out of trade policy altogether (b and c). Two are not, on inspection, grounds for restriction at all and belong in other chapters (d and h). One is theoretically sound but not generalisable (g). Two remain seriously open: the infant-industry question in its capital-market form (e), and above all the question of adjustment (f).

The pattern is more remarkable than the tally. In not one case is the verdict that the goal is unwarranted. It is always that the instrument is wrongly chosen, and always for the same reason. The group that can compel is not the group that is affected.

What follows

The examination does not vindicate the national accounting unit. Only one of the eight grounds requires the group as a group, and that one turns out to be a cartel. The rest need properties held variously by clubs, alliances, supply-chain consortia, agglomerations and revenue authorities. What the nation-state contributes is that it happens to hold all five at once — a coincidence that says nothing about language, descent or cultural belonging, and one that is contingent. Before the nineteenth century customs areas, currency areas and legal orders frequently fell apart, and they fall apart again today. The European Union holds market power and the capacity to bind externally at Union level while the capacity to compensate lies almost entirely with the member states; it can therefore threaten, negotiate and levy duties, but it cannot indemnify the losers of its own trade policy. What is usually described as political failure is a structural deficit.

The persistence of the nation-state in trade policy therefore has an explanation other than the one usually supposed. It withstands criticism not because it is the functionally right level but because it is the only level at which the instrument bites. That is a statement about the tool and not about the matter.

Which leaves the question of why, if the grounds do not carry the framing, the framing survives. There is an old joke about a man searching at night under a street lamp for a key he lost in the park; asked why he is looking here, he says: because this is where the light is. Kaplan brought the figure into the philosophy of science as the principle of the lamp, and in the usual telling the lamp stands there by accident and the searcher is merely lazy.

In this case it does not stand there by accident. Trade data arise at the border because duty is levied at the border. The statistic is a by-product of tax collection, and the measuring boundary is therefore not a boundary of knowledge but a boundary of coercion — the word statistics does not come from state by chance. The lamp is operated by precisely the party whose warrant is under examination, and a feedback loop follows: the coercion boundary generates data, the data generate categories, the categories generate questions, and the questions confirm the boundary. The tariff finances its own justification by producing the numbers in which it appears to make sense.

The mechanism predicts something, and the prediction has been borne out: wherever finer data appeared, the unit of analysis moved after them. Firm-level customs records produced the literature on firm heterogeneity within a few years and revealed that a tiny fraction of firms handles almost all trade. Measuring in value added rather than gross flows made the American deficit with China considerably smaller. And McCallum found a startlingly strong border effect between Canadian provinces and American states, visible only because Canada, exceptionally, records trade between its provinces. The corresponding blind spot follows at once: nobody knows anything about the exchange between Lyon and Marseille, because nobody measures it, because nothing is levied there. The nested test set out above is therefore, at present, a logical instrument and not an empirical one. It fails on data that are missing only because the practice whose warrant it examines is missing there.

An essay of this kind should say what it is asking for. The book implied by the foregoing has not been written. Its first part would set out why persons specialise and exchange, entirely at the level of the person. Its second would introduce the group without fixing its size and would test every ground for restricting exchange against the properties that ground requires, carrying the protection group and the compensation group side by side. Its third would treat the nation-state last and as a derived case. Mises, George and Bastiat refute, but they do not catalogue; Irwin catalogues, but nationally. The shelf where such a book would stand is empty.

The poverty in my title, finally, is not the charge that methodological nationalism is false. The nation-state is entirely real. It compels, it binds, it redistributes, and a currency area imposes on everyone inside it the same interest rate whether or not anybody measures the fact. The charge is that as a starting point for explanation it is barren. It manufactures questions that cannot be answered — whether a country is competitive, whether a bilateral balance is fair — and it hides the ones that can: who saves here, and did they choose to; who bears this cost, and did they vote for it; who is the creditor and who the debtor, and was it in either case a decision. A framework is poor when it makes the answerable questions invisible. That is the case against this one, and settling it will take the book.

Discussed and written with Claude.