Collection:
Geoeconomics reading list

61 papers · 3 sessions

Written by Fable 5

Table of Contents

Start here
Block 1 — Trade, production, technology
Block 2 — Interdependence, chokepoints, coercion
Block 3 — Money, finance and macro
The threads, pulled together

Sixty-one papers for a reading group on geoeconomics, in the order the group reads them, each with its own digest one click away. The list has three blocks and the blocks have a logic, which the group’s own spreadsheet states in one line: networks propagation (production and technology) → leverage (interdependence) → macro-financial transmission (money and macro). First the machinery that says what a network does to a shock and what a trade war actually costs, because that is machinery a trade economist already owns. Then the new theory, in which the same networks are read as instruments of state power and the same elasticities become the price of coercion. Then money, where the dollar turns out to be the largest chokepoint of all and fragmentation turns out to be a demand shock in disguise. Each block leans on the one before it, which is the argument for reading in order and the reason the third block, the one closest to the group’s own work, comes last.

Two definitions of the field are in play and the list uses both. Mohr and Trebesch’s is broad — the study of the links between geopolitics and economics, where geopolitics means rivalry for power — and it decides what is in scope, which is why a block on the dollar and on fragmentation as a macro shock can sit next to the coercion theory at all. Clayton, Maggiori and Schreger’s is narrow — a government using the dependence of foreigners on it to extract what it could not otherwise get, with power defined as the gap between the target’s value inside the system and its value outside — and it decides what is central, which is why Block 2 is built around their three papers. Between the two runs a third concept, the IMF’s geoeconomic fragmentation, a policy-driven reversal of integration, which is about consequences rather than power and is the vocabulary Block 1’s empirics are written in. A surprising amount of the field’s apparent disagreement is these three definitions passing one another in a corridor, and the list keeps them labelled.

Every row carries a role. Theory, structural and empirical papers are presented in the group, and the label says what to present: a mechanism, a model taken to data, or a measurement and its identification. Context and data rows are the grey ones on the spreadsheet — read for vocabulary, tools or history, not presented, and, in the group’s own note to itself, đọc cho biết: read so that you know. They are called context for a reason. The survey, the definitions, the network theorem, the one live experiment on elasticities, the sanctions stylized facts and the three datasets are what every presented paper assumes on its first page, and this essay leans on them to frame each section before walking the presented papers in list order. The four books on the list (Hirschman, Kindleberger, Blackwill–Harris, Mulder) are not rows on the spreadsheet, which is papers-only, and appear here in prose.

One thing the list is built to correct. Two kinds of paper sit under the one word and do not meet: papers with a strategic agent choosing threats and dependence, and papers that treat geopolitics as a shock process and estimate its transmission. The survey draws no line between them, and the second kind has none to draw — Ambrosino, Chan and Tenreyro’s small open economy contains no coercer at all. What the halves share is one ancestor: Martin, Mayer and Thoenig is cited by eleven of the presented papers, more than anything else in the pool. Read in this order, the two branches are read together, which is the point.

Start here

A field is a definition plus a reading list, and a survey of a three-year-old field cannot summarise results because there are not many yet. What Mohr and Trebesch’s Geoeconomics does instead is draw the border and furnish the rooms: the broad definition above, five subfields — policy tools such as sanctions; the geopolitics of trade; the geopolitics of finance; geopolitical risk and its spillovers; the economics of war — and roughly 230 references sorted into them, each with a sentence on what is known and one on what is not. The five rooms map almost one-to-one onto the three blocks below: tools and trade are Blocks 1 and 2, finance and risk are Block 3, and war is left out except where it touches trade. Sanctions are the best-known tool and, the survey says, the best-understood: by and large they achieve their stated economic objective, with disagreement only on magnitude. The Kiel draft called the same evidence “overall mixed results”; the published version says sanctions “by and large” hit their economic targets. The citation list under the sentence is identical in both. The section that says “underexplored” is finance, and the closing list of gaps asks for a conceptualisation of economic coercion and a quantification of its rents: the narrow definition’s central object has not been defined, let alone measured. The bibliography has a Kiel accent — one of the two conferences it names is Trebesch’s own institute’s, as are eight of the references — but it no longer says how it was assembled. The working paper stated the rule: highly cited papers, and papers presented at the authors’ own two conferences. The published version kept the conferences and cut the rule. Read all of it; it is short; come back to it when deciding what the list leaves out.

Block 1 — Trade, production, technology

The block asks a question the group already has the tools for: is trade reallocating along geopolitical lines, by how much, through which margins, and at what cost. It opens with the theory of what a network does to a shock and the one experiment in which a network was actually shocked, then forces the reader to say what evidence of fragmentation would look like before reading the papers that claim to find it, then adds a geopolitical term to the tariff theory the group knows, then follows the goods as they reroute, and ends with the export-control instrument. Fourteen presented papers, more than either other block, because this is where presentations go fastest.

1.1 Production networks

Start with a theorem that appears to make the whole subject irrelevant. Hulten’s says that, to first order, the aggregate effect of a shock to one producer is its sales share of GDP — its Domar weight — and nothing else: not its position in the network, not how substitutable its output is. Walmart and electricity generation each sell about four percent of American GDP, so a large negative shock to either is equally bad, and nobody believes that. Baqaee and Farhi’s Beyond Hulten’s Theorem is the reason nobody should: Hulten is a first derivative, and the second derivative is how much the shocked sector’s share moves when it is shocked. With complementarities the share rises after a bad shock, the next percent hurts more than the last, and symmetric shocks produce negatively skewed output out of Gaussian material. The oil exercise needs no model at all: crude oil’s expenditure share went from 1.8 percent of world GDP in 1972 to 7.6 in 1979, and weighting the 1970s shock by the average share rather than the starting one nearly triples its cost. This is the definition of a strategic sector the rest of the list uses without restating it: not a large one, but one with low elasticities at the nodes that would have to substitute, unevenly exposed downstream, hit hard enough that the second-order term is no longer small. The elasticity that sets the cost of the embargo here is the elasticity that sets the coercer’s price in Block 2.

The two theory papers ask what happens when the network is not given but chosen, and they disagree about which way firms err. Acemoglu and Tahbaz-Salehi add customized relationships, bargaining over the surplus and costly link formation, and find that the equilibrium chain is a strict sub-network of the efficient one and that aggregate output is discontinuous in shocks where the planner’s allocation is continuous, because firms sever links at their own indifference point rather than society’s. A finer division of labour raises fragility only where the specialist has bargaining weight enough to reach its cliff early. Grossman, Helpman and Lhuillier then take the word “resilience,” which an executive order defines as supply chains that are “diverse and secure” and operationalizes as rebuilt domestic capacity, and show these are two margins and not one: a reshoring subsidy aims at location and reaches diversification only as a side effect. Under CES the planner never tilts location at all; outside it the sign turns on the demand elasticity rather than on geography, and at the lower of the two elasticities simulated the market over-diversifies and the second best is a tax. The 2021 working paper said subsidize diversification and ranked that instrument first in a welfare panel. The published version re-simulated on a new baseline, dropped the panel, and taxes it. Two externality arguments about one margin, pointing opposite ways.

Then the experiment. Twelve days into the war, nine economists put the cost to Germany of a full stop of Russian energy at half a percent to three percent of output, against sponsored studies of six to twelve and a chancellor warning on television about “the irresponsible use of mathematical models.” (One to three is the retrospective’s restatement of it a year later.) Bachmann and co-authors’ What if? is Beyond Hulten applied to one input, one country and seven or eight months: the loss is the expenditure share times the shock plus half the change in the share times the shock, so every belief about substitutability is a belief about how far a 1.2-percent gas share jumps, and Leontief is the belief that it jumps to a hundred. Russia then cut the gas, and Moll, Schularick and Zachmann’s The Power of Substitution is the autopsy. A 41-point hole in German gas supply was closed by 33 or 34 points of third-country imports and a 20-percent fall in demand, 26 in industry — most of it by pipe, Norway supplying 16 points against 13 for all LNG, of which other countries’ terminals landed about 10. The terminals Germany never built were the tail, not the bulk. The energy-intensive branches fell close to 20 percent and everyone downstream of them did not, the opposite of a cascade; that the gas-intensive intermediate was imported instead is the authors’ hypothesis, on evidence they call illustrative. The number to carry is that a 20-percent cut to a one-percent input costs 20 percent of output at an elasticity of zero and 2.7 percent at 0.05. The rhetoric of the critical input assumes that factor of ten away, and nobody, in the end, substituted away from the belief that the people most exposed to a shock are the best judges of its size.

1.2 Fragmentation

Before you can find fragmentation you need the word and a null. The IMF staff note supplies the word — a “policy-driven reversal of integration, often guided by strategic considerations,” with “often” doing quiet work so that an ordinary protectionist tariff qualifies — and the channel list (trade, capital, migration, payments, public goods) that the empirical papers work through one at a time. It also supplies the number everyone quotes, 0.2 to 7 percent of world GDP, which is narrower than the quoting suggests: it is one study’s range, limited restrictions at the floor and no inter-bloc trade at the ceiling, and its two ends are comparable with each other. What is not comparable is the clause after it, where another paper’s country-level 8-to-12-percent figure is set beside a world-level trade-only range. The health warning sits in a box ten pages on. The note reports “few clear signs of fragmentation in the trade data yet” as of January 2023; what it can count is restrictions, mentions of national security and mentions of reshoring. Goldberg and Reed then supply the null. World imports hit a new high of roughly $27 trillion in 2021; the trade-to-GDP ratio has been flat since 2008 partly because China and India started selling to their own consumers — the paper’s reading, contested in its own published discussion, where Irwin says that fall may be Xi and Modi turning inward by policy, which would carry the IMF’s price rather than being free arithmetic; face-mask imports rose to nearly seven times their 2018 level in 2020, which is a chain working rather than breaking. “Resilience,” they show, cannot be operationalized without saying which shock you want to survive, and no one has said. Their verdict — not in the flows, yes in the policy, flows follow policy with a lag — is the standard the rest of the sub-block must meet.

Gopinath, Gourinchas, Presbitero and Topalova meet it with a gravity regression a graduate student can replicate in an afternoon, which is why it is the first thing a new member should reproduce. Define blocs from UN voting, put pair, source-time and destination-time fixed effects on bilateral flows, and between-bloc trade sits about 11 percent below same-bloc trade since 2022, announced FDI projects about 12 — level gaps against pairs inside a bloc, not growth rates, and the FDI one significant only at ten percent. The IMF working paper put the between-bloc FDI shortfall near 20 percent. The published version extends the window, drops thin pairs and halves it to 12.3, now driven mostly by Chinese flows. Run the same regression on 1920–90 and the Cold War shortfall is 67 percent, which gives the present a ruler. Trade with the nonaligned fell 37 percent in the Cold War and is flat now, and the nonaligned gain American import share in proportion to the Chinese exports and investment they receive — the aggregate Goldberg and Reed could not see, visible only as a relative gap. The sub-block ends with the paper that turns the security rationale against itself. Mayer, Méjean and Thoenig put a bargaining game inside a quantitative trade model — Compte and Jehiel’s protocol, the kind you can walk out of — in which each leader announces a cost of war and shades only its private half, trade exposure being public. De-risking then lowers what you lose in a war, which lowers the cost of war, which raises its probability. Geopolitics argues for less protection, not more. You wanted to stop being a hostage; the hostage was the thing keeping the peace.

1.3 Tariffs and trade wars

The group knows the Mill–Bickerdike tariff; this sub-block adds one term to it. In Becko, Grossman and Helpman the large country values allies and can offer them a preferential deal, so the most-favoured-nation tariff becomes the penalty for declining the carrot and the optimal one exceeds the terms-of-trade level. The proof to remember is Martin’s two-sentence discussion: from the terms-of-trade optimum, raising the tariff costs a second-order loss and buys a first-order alignment gain. Without geopolitics one flat tariff; with it, zero for joiners and a high rate for refusers, and small countries sort by how much they mind aligning. Reading it onto the 2025 regime takes an inversion — the objection in the room was that allies now pay to stay rather than being paid to join, security the cudgel and the tariff the thing extracted — and it survives only because quasi-linearity leaves it indifferent between buying friendship and selling protection. The allies, one suspects, are not. Grossman, Helpman and Redding then show that an input tariff is not a tariff on a good. Supply chains are costly-search matches under renegotiable contracts, so a small tariff worsens the imposing country’s terms of trade by weakening every buyer’s outside option in every existing bargain, and a large one triggers costly re-search in a third country; calibrated to the 2018–19 tariffs, the bill is 0.12 percent of GDP, input use doing most of the damage and search costs the rest. The 2020 draft carried an inelastic-demand branch in which a tariff near 37 percent raised welfare by more than three percent of spending. The published version drops the branch and calibrates every case to a loss. (Lashkaripour and Simonovska’s Optimal Tariffs under Geopolitical Risk, which puts import dependence inside the same optimal-tariff formula, was digested from its NBER talk and left off the list for want of a public paper; it belongs beside these two.)

The third row is the incidence baseline, and the group reversed itself to include it. Fajgelbaum, Goldberg, Kennedy and Khandelwal’s The Return to Protectionism had been left off as known; it came back because the geopolitical tariff papers need a real tariff war to be read against, and this is the one with the identification. A tariff opens a wedge between what the importer pays and what the exporter receives, so the same instrument traces the demand curve from one side and the foreign supply curve from the other; the answer is complete pass-through, foreign prices cut nothing, American buyers lost $51 billion, the Treasury and protected producers recovered most of it, and the net was $7.2 billion, a transfer with a small deadweight tail — plus a retaliation, aimed by foreigners, that landed hardest in the safest Republican counties. Keep the two apart: the $51 billion is precise, the $7.2 billion is not, its interval running through zero, and without the retaliation the tariffs come out a small gain. The Fajgelbaum–Khandelwal survey of the whole incidence literature is digested alongside it and closes on the field’s own admission: unusually confident about the pass-through coefficient, unable to say much about the thing the tariffs were for. That gap is where Blocks 2 and 3 live.

1.4 GVC reallocation, reshoring & friendshoring

Follow the goods. The textbook says a tariff war between two countries diverts trade: bystanders sell more to the taxing country and less elsewhere. Fajgelbaum and co-authors’ sequel finds bystanders sold more to the United States in the products it taxed (elasticity 0.31) and more to everyone else (0.20), and more to the rest of the world in the products China taxed (0.29), so the diversion prior is wrong on average. The paper declines to run that average through its own taxonomy, the average country being revealed neither substitute nor complement for China; applied country by country, the taxonomy makes who gained a property of the country, not of product mix. Alfaro and Chor name the phenomenon and send the bill. China’s share of American goods imports fell from a 21.6-percent peak in 2017 to 16.5 in 2022, and the slogan about who took it does not survive the ranking: Vietnam gained about two points, Taiwan one, India 0.6, Canada and Korea 0.55, Mexico half a point. High-wage Asia took more of the reallocation than Mexico did. The substitutes cost about ten percent more from Vietnam and three from Mexico, though only the Mexican figure clears five percent and the authors cannot separate cost from quality; reshoring is a “perhaps”; and China is now the supplier’s supplier, which is the caution every later paper answers. Baqaee and Farhi’s Networks, Barriers, and Trade is the workhorse behind every decoupling counterfactual: hat-algebra generalized to many countries, nested-CES networks and arbitrary wedges, with the quantitative section showing that on a 60-percent universal trade-cost shock a model without linkages understates the loss by half, and one without nominal rigidities by half again. Domínguez-Iino, Elliott and Hsiao supply the resources case with a mechanism: producers of the same mineral are substitutes and behave like any cartel, but producers of different ones are joined by a chemistry and are complements, so when Indonesia restricts nickel the prices of lithium and cobalt fall. Market power among complements is not OPEC, though the top three producers hold 70 to 85 percent of each.

The hinge to Block 2 is Kleinman, Liu and Redding. Bilateral trade is the wrong measure of dependence twice over; compute instead the general-equilibrium elasticity of one country’s real income to another’s productivity from the whole trade matrix, call positive friend and negative enemy, and use two instruments — China’s emergence fed through the 1980 trade matrix, and air-travel costs interacting with sea-versus-air distances — to show that economic friendship causes political realignment on UN votes, rivalries and alliances. The Block 1 gravity papers regress trade on alignment; this paper reverses the arrow with a model-based regressor, and the next block asks what alignment is worth to whoever buys it.

1.5 Technology, export controls

The most literal chokepoint on the list is a garden hose on the seabed. Porcellacchia, Trebesch and Wache’s Digital Chokepoints counts about 500 submarine cables carrying nearly all intercontinental data, notes that one modern cable exceeds the combined capacity of all seven thousand SpaceX satellites in orbit, and that roughly seventy repair ships exist and none will enter a war zone. It is not background colour but a model with counterfactuals, and the asymmetry is the finding: robust to any single cut, fragile only to coordinated ones, with 44 cables and a quarter of global traffic through the English Channel. Flynn, Levy, Moscona and Wo find the private-sector channel of fragmentation: when a foreign input becomes politically risky — Elon Musk’s 2014 Senate testimony on the Russian engine in the Atlas V is the set piece — domestic firms innovate away from it before any policy acts, and the supplier loses the market whether or not the feared disruption ever occurs. The response is aimed and one-way: essentially all of it is risk in non-allies, and a later fall in risk undoes neither the patents nor the lost exports. Alekseev and Lin then add a defense department as a second final buyer in a network trade model and derive the export-control instrument: the optimal export tax on a good trades its network-adjusted sales to the foreign military against its roundabout sales back to domestic buyers, scaled by the import-demand elasticity. (The thirty-cents-to-the-Chinese-military, twenty-cents-back chip figure is hypothetical, not measured; what is measured are the centralities, which rank aluminium powders, warships and tanks at the top.) The implied military-use measure predicts the EU dual-use list with an R² of 0.85 against 0.35 for raw military sales share; depleted stockpiles double the effect of controls, rerouting through third countries cuts it by more than half, and only a coalition makes American policy outweigh China’s. The sub-block is titled for that paper.

Block 2 — Interdependence, chokepoints, coercion

This is the block the group does not already know. It runs from Hirschman’s 1945 observation that trade is a hold over the partner who needs it more, through Farrell and Newman’s restatement for networks, to Clayton–Maggiori–Schreger’s formalization and the 2025–26 papers that extend it to two hegemons, networks, alignment, finance and firms. Sanctions are its empirical arm; the measurement sub-block is its data; trade-and-conflict is the bargaining logic underneath. Seventeen presented papers, three of them by CMS, which is a deliberate concentration: the CMS machinery — a participation constraint, an inside and an outside option, power as the gap — is the field’s lingua franca and the group needs to own it.

2.1 Lineage and entry points

The lineage is short and mostly not by economists. Hirschman wrote in 1945 about Nazi Germany buying political compliance from its neighbours with trade, and proposed a sum-of-squared-partner-shares index of exposure that escaped into industrial organisation as the HHI, leaving the power question behind. Maggiori’s gloss is that Hirschman had the right instinct with the wrong exponent: concentration matters raised to an elasticity. Economics then let the question lapse for seventy years while political science kept it. Blackwill and Harris revived the word in 2016 with an inventory of instruments, on a definition the survey opening this list reads as a restatement of Baldwin’s economic statecraft. Neither has a model, and the two rows in this sub-block are the bridge from that prose to one.

Farrell and Newman’s Weaponized Interdependence is political science’s statement of the mechanism, and the economics that follows reads best as an answer to it. Networks have hubs because hubs are efficient, and “focal points of cooperation have become sites of control”: a state with jurisdiction over a hub — not merely a large market — can see what flows through it (the panopticon effect) or deny access to it (the chokepoint effect), and the denial devastates the target precisely because the network removed its alternatives. Two conditions must hold at once, and the second is where the variation lives: jurisdiction, and institutions able to use it. SWIFT went from 22 countries and 3,000 messages a year in 1977 to 6.5 billion in 2016, having located itself in Brussels to escape New York and London and having been captured through a backup data centre it prudently built in Virginia; the internet, the largest hub of all, is the case in which the United States ran the panopticon at scale and never used the chokepoint, having no way to order anyone to cut a country off. Jurisdiction without institutions is not power. The paper’s one sentence about states staying in a network “only up to that point where the costs of remaining in them are lower than the benefits” is a participation constraint in prose. Clayton, Maggiori and Schreger’s Putting Economics Back into Geoeconomics is the readable on-ramp to writing that sentence down. The complaint is definitional — “chokepoint” and “strategic sector” have run for three years undefined, which suits everyone lobbying for the label — and the remedy is a truth condition: the loss is the cost of autarky pointed at a subset of goods, so a chokepoint means the hegemon’s share of the target’s spending near one and the elasticity of substitution near one, and power is the gap between inside and outside option. Run over input-output tables it says what no widget lobby wants computed: most sectors generate no dependency at all. The tape measure then fails where Farrell and Newman’s second case lives, services being the least measured input in it. Assign it before the Econometrica paper.

2.2 Coercion, hegemonic models

Khrushchev said you can make anything strategic, even a button, because a soldier without buttons has to hold his trousers up and then what does he do with his rifle. Maggiori opens his talks with that line because it is the problem: “strategic sector,” “chokepoint” and “dependency” are used constantly and defined by nobody, since defining them is in the interest of no one who wants a subsidy. A Framework for Geoeconomics defines them out of three standard pieces of theory — limited enforceability with trigger punishments, input-output amplification, and externalities — and the definition is that power is the slack in a participation constraint. A hegemon that can threaten several relationships jointly collapses several incentive constraints into one, which loosens them all and creates slack it can charge for; a generalized Leontief inverse carries the demanded change into sectors it cannot contract with, which makes the wedge a macro object; and enforcement is positive-sum, so the hegemon spends the surplus it created moving the world inside the frontier it made. The group presents the small model first: The Targets of Geoeconomic Coercion asks when to threaten ASML and when to threaten The Hague, and answers that targeting firms rides a scale externality they do not internalize while targeting the government buys slack-sharing across sectors — twenty-eight pages that teach the machinery in an hour. A Theory of Economic Coercion and Fragmentation is the defence: the economies of scale that make integration valuable are what make the hegemon’s inputs unsubstitutable, so gains from trade and coercion exposure are one quantity seen from two sides; anticipating coercion, each country insulates, shrinking scale for everyone — the doom loop — and the outcome can be worse than no hegemon at all. Its power statistic is a closed form in shares and elasticities, and nonlinear: ninety-five percent control of an input is far more power than eighty-five, and alternatives like SPFS and CIPS are inefficient by construction, which is the hegemon’s power measured in another unit. It also contains the exit: a hegemon that commits to taking only part of the inside option does better than one taking all of it, because unconstrained coercion leaves nobody on its system. The liberal order as hegemonic statecraft, not its absence.

Then the extensions, in list order. Meyer and Wesseler put two hegemons in a procurement problem for a continuum of countries’ alignment, with carrots priced in cash and sticks as the loss from being cut off from that hegemon’s trade, not from autarky. The sticks are cutoff rules never swung in equilibrium, the optimal carrot is hump-shaped in the rival’s — you match until it gets too expensive and throw in the towel — and the estimation runs on digitized Cold War aid. (Levchenko, Pandalai-Nayar and Young’s Geopolitical Payoffs, which asks whether aid pays for itself in exports, lost its slot to this paper and is digested beside it.) Liu, Redding, Smith-Worthington and Yang model foreign policy as a network beauty contest, each country’s action a weighted average of its own interest and its neighbours’, with a prior stage in which countries pay to shift each other’s incentives; the digest follows the July 2026 draft. Broner, Martin, Meyer and Trebesch give Kindleberger a mechanism that needs almost nothing: countries differ in preferred policies, trade rises with alignment, and a large enough hegemon makes aligning worth everyone’s while and everyone else aligns with the alignees — a globalization equilibrium without coercion, tested on some 71,000 bilateral and 6,000 multilateral treaties since 1800, where the distinctive prediction is that two countries aligned with the same hegemon trade about five percent more with each other. A rising rival unravels it only from the opposite end of the circle — one next door merely adds its size — and unravels it into everybody at their own policy, not a rival bloc. Pflueger and Yared close the loop toward Block 3: cheap borrowing funds the military and the military makes the borrowing cheap, a feedback that amplifies small advantages and can flip, as Britain’s real borrowing cost against the Netherlands went from a 1.42-percent disadvantage in the eighteenth century to a 1.98-percent advantage in the nineteenth.

Three papers then answer the CMS question differently, and only one of them offers a rival statistic. Liu and Yang’s International Power is Hirschman with a number attached: ex-ante trade plans, ex-post hold-up in a dispute, and a regression-ready measure — import shares weighted by inverse trade elasticities, directional by construction — computed for every pair-year since 2001, correlating 0.96 with its general-equilibrium version. It disagrees with CMS about who is powerful, the two measuring different objects; power provokes engagement, and partners whose alignment sours build power by cutting imports: the import share moves and the export share does not, so power is built defensively. Kooi’s Power and Resilience offers a rival source for the externality rather than a rival number: bargaining in the shadow of conflict, not contract enforceability. Resilience improves your bargain because it makes you credibly willing to walk, and nobody pays private agents for the bargaining power their locations confer. The optimal policy subsidizes capital in proportion to its appreciation in conflict; trade taxes aimed at your own resilience are exactly zero, though trade policy toward an adversary is not — below the terms-of-trade tariff in peace, above it in conflict. The war game’s answer depends on which war: Taiwan alone puts semiconductors first of 241 sectors at a seven percent subsidy bound, China and Taiwan together sixteenth. Becko and O’Connor’s Strategic (Dis)Integration is the offensive counterpart and offers no statistic at all, only the hegemon’s policy before it threatens: with credible threats the optimal industrial policy is zero, the on-path trade subsidy already pricing the externality, which the authors say calls the geopolitical rationale for CHIPS into question; industrial policy returns only when threats are not credible, to buy credibility. Their theorem pins the target of peacetime trade policy, not its sign; the calibration, not the theorem, promotes trade with China. Antràs and Padró i Miquel end the sub-block with influence running through domestic politics: with mobile capital and immobile labour, foreign tax policy is an ideologically sorted externality, so Chávez backs Morales without believing anything and incumbents buy foreign elections as a costly lottery. (Kempf, Luo and Tsoutsoura’s CEO-ideology paper, in which partisan chief executives run partisan supply chains, is the firm-level pattern this supplies a rationale for; it is digested but off the list.)

2.3 Sanctions

The sanctions literature is an empirical arm attached to a selection problem, and the survey that opens the sub-block says so. Felbermayr, Morgan, Syropoulos and Yotov’s stylized facts come from the Global Sanctions Data Base: programmes in force rose from about 200 to about 600 between 2013 and 2023, some 12 percent of country pairs and 27 percent of world trade are under some sanction, and about four-fifths of regimes never touch trade directly — the treatment is mostly financial and travel measures. The fact to keep is that every sanction in the database is one whose threat already failed: senders must sometimes follow through to keep their threats credible, so the recorded cases are the recalcitrant targets plus the called bluffs, and success rates computed from them are conditional on that. Mulder’s The Economic Weapon is where that selection problem reaches the block as history.

The four presented papers each take one of the survey’s open questions. Egorov, Korovkin, Makarin and Nigmatulina take the distinction between coercion and warfare: export controls are warfare, the right place to measure them is the target’s output rather than its trade flows, and with Russia’s own customs and firm data — sanctioned varieties covering 36 percent of prewar imports — they show directly sanctioned flows fell nearly 60 percent, imports rerouted through friendly countries, and production did not recover. The friendly-country surge, read off the customs form, is almost entirely rerouting. Fernández-Villaverde, Li, Xu and Zanetti measure evasion from the one thing a dark tanker cannot hide, the gap in its AIS signal: a dark fleet averaging 555 tankers, a quarter of the crude tankers they can classify — some 700 a year broadcast nothing and cannot be sorted. Becko takes the sender’s side and proves that sanctions are optimal tariffs with the weight on the target flipped — the trade tax on good $k$ is $(1-\lambda_F)$ times the classic optimal tariff, so sanctions hit the same goods as terms-of-trade manipulation only harder, small sanctions pay for themselves, and the sanctioner’s own elasticities are irrelevant; it is the proof that 1.3 and 2.3 are one machinery. Hausmann, Schetter and Yıldırım take design: the EU supplies about 40 percent of Russia’s imports while Russia takes under two percent of EU exports, a Baqaee–Farhi second-order criterion says the cost to Russia is convex in the coalition’s share of a product, so targeting and coordination beat breadth — and the actual 2022 lists were chosen by import size, not leverage. Re-targeting and coordinating raise Russia’s welfare loss from 0.6 to 1.1 percent, about eighty percent more, at little to no extra cost to sanctioners, who lose about a hundredth as much. Cite the published figure. Several headline numbers moved from the 2022 working paper, and the version of record adds a conclusion conceding Russian GDP fell only 2.1 percent in 2022 — GDP being the wrong scoreboard.

A fifth presented paper, the dataset, and the policy survey. The GSDB is the treatment variable in nearly every gravity paper since, coded from the sender’s own documents by type, objective and success, with threats not recorded at all. Its gravity demonstration is a lesson in separating types: pooled into one “any sanction” dummy the effect is a statistical nothing, while a complete bilateral embargo removes about 77 percent of trade and complete import sanctions about 52. The working paper’s pooled dummy was positive and significant — sanctions apparently raising trade — and no import sanction was distinguishable from zero. Both results are gone in the published version. Lumping types together is what breaks the estimates. Crozet and Hinz, promoted from context to a presentation, are the 2014 baseline: of $96 billion lost over two years Russia bore $53 billion and the 37 sanctioners $42, so “friendly fire” names the composition of the West’s share, not its size. Eighty-seven percent of that share was in goods nobody embargoed, and French customs data trace it to trade finance in the months after the financial sanctions arrived — though the authors will not affirm that the 87 percent is the sanctions alone. Against total exports Russia lost 7.4 percent and the West 0.3; only against predicted bilateral trade does the West lose more. Itskhoki and Ribakova are the policy survey around 2022, a theorist and a practitioner agreeing that the coalition held a strong hand and played its cards in the wrong order, kept the strongest in reserve, and enforced loosely enough that Russia rerouted through China, Turkey and the former Soviet republics within a year; “no pain, no gain” is their summary of what sanctions cost the sender.

2.4 Measuring pressure and alignment

Two datasets, one for each side of the threat. Clayton, Coppola, Maggiori and Schreger’s Geoeconomic Pressure runs open-weight language models over some 785,000 documents — earnings calls, Chinese investor meetings, analyst reports — and classifies who pressures whom with which instrument and, the part the theory needs, whether the pressure was implemented or only threatened. Firms hit by tariffs respond through prices and firms hit by export controls through R&D. And it sees the off-path threat, which is what Block 2 is about: after each Trump election, four-fifths of affected firms are discussing tariffs that do not exist yet. Bailey, Strezhnev and Voeten’s UN ideal points are the alignment variable behind most of Block 1, and in Kleinman, Liu and Redding the outcome rather than the control: a dynamic ordinal model of General Assembly votes in which identically worded resolutions, repeated within five years, anchor the scale across time, so that a position can be told apart from the agenda. The raw agreement scores everyone used before confound a change of opinion with a change in what the Assembly happened to schedule, and the worked example is unkind: two countries agree on nine votes of ten, nobody’s preferences move, the Assembly schedules ten more on the divisive one, and their similarity falls to 50 percent. The caveat that travels with the variable is not the one usually attached. Kuziemko and Werker’s 59-percent rise in American aid when a country rotates onto the Security Council is about the other chamber, and these authors argue vote-buying is rarer in the nonbinding Assembly, checking that against a series built only on the votes Washington lobbied for, which tracks the full one at .92. The problem that remains is that the hegemon who might buy alignment is also the one offering market access, so a coefficient on ideal-point distance is a reduced form for a bargain rather than an elasticity of trade with respect to taste. The remedy is to put the ideal point on the left-hand side, as the carrots-and-sticks papers and Kleinman, Liu and Redding do. Both are tools-session material, not presentations.

2.5 Trade and conflict

The sub-block exists because the presented papers keep citing a bargaining logic that was nowhere on the list; Martin, Mayer and Thoenig alone is cited by eleven of them. Fearon’s Rationalist Explanations for War is the premise: war is costly, so under three mild assumptions — a true probability of winning, no taste for risk, a divisible issue — a bargaining range both sides prefer to fighting exists, and the question is not why states fight but why they fail to settle. The first friction is private information with an incentive to misrepresent — stated as Fearon states it, not as the literature repeats it: a costly signal may raise the risk of war, the informative signals being exactly the ones that court it. July 1914 is the case his own conclusion supports; the Russo-Japanese one turns on updating about capabilities, of which he ends the paper saying he knows no clear instance. Commitment is the second friction, indivisibility mostly demoted. Every trade-and-conflict model picks one friction and attaches trade as the cost of war. Martin, Mayer and Thoenig pick the first and get the result the friendshoring papers do not state: bilateral trade raises the opportunity cost of a war with that partner, multilateral openness makes any single partner replaceable, and for neighbours the second effect wins, thirty years of globalization raising their conflict probability from 3.67 to 4.46 percent — about a fifth, the effect of erasing twenty years of peace — which is the root of the Fragmentation Paradox in 1.2. Two qualifications travel with it. The result is conditional on the kind of war: globalization raises the probability of every bilateral conflict and lowers that of one between large coalitions — more local wars, fewer world wars. And the halves are not equally solid, the multilateral coefficient holding at one percent throughout and the bilateral one only at ten, vanishing under pair fixed effects. The half everyone believed is the fragile half. Thoenig’s handbook chapter is the toolkit: the same result across the whole class of structural gravity models, with named “geoeconomic factors” such as the opportunity cost of war computable for real country pairs. Glick and Taylor anchor the cost, and Thoenig calibrates from them: in gravity since 1870 adversary trade falls 85 percent on the outbreak of war and takes a decade to recover, neutrals losing proportionally more than belligerents, the discounted loss running to about twice the human cost of the First World War and about equal to it in the Second. Rohner, Thoenig and Zilibotti make it dynamic: war signals a low propensity to cooperate, destroys the trust trade needs and so lowers the cost of the next war, so a country can fall into a conflict trap through a run of accidental wars, and one that escapes escapes for good. It is a theory of civil conflict, which is why it stays context, and 68 percent of civil-war outbreaks were in countries with more than one conflict on record.

Block 3 — Money, finance and macro

Read last because it leans on both earlier blocks: currency dominance is the financial form of the chokepoint, and fragmentation macro is what the reallocation of Block 1 looks like from a central bank. Eight presented papers, the smallest block, because two of them are close to the group’s own research and will take more than a session each.

3.1 Currency dominance and sanctions-finance

Gopinath and Stein collapse five stylized facts into one: a claim is safe only if it buys a known basket, so “which deposits are safe” and “which currency are goods priced in” are the same question. Dollar-invoiced imports create demand for dollar deposits, banks over-collateralize to supply them only because dollar funding is cheap, the cheap funding is lent to firms with no dollar revenue (mismatch as an equilibrium), and third-country exporters invoice in dollars because they can borrow in them — sixty percent of Turkey’s imports are dollar-invoiced against six percent bought from the United States. Dominance is a threshold, not a share. The published model differs from the working paper: invoicing is a zero-to-one step rather than a smooth rule, one dominant currency is the unique stable outcome in a middle band, and raising euro safe-asset supply alone now hinders internationalization rather than helping it. Mukhin makes the same complementarity a coordination game through input-output links and competitors’ prices, and gets 65 percent of world invoicing in dollars against a 23 percent trade share for the dollar bloc — a model output, not a measurement, since no invoicing data exist — with the multiplicity that implies for a challenger. Farhi and Maggiori supply the other side: a monopolist issuer of safe debt whose promise not to devalue is only a reputation, three zones of issuance (safe, unstable, collapse), and a Triffin dilemma that is the unconstrained optimum sitting outside the credible zone, so the hegemon either retreats or issues into instability. A multipolar system helps with many issuers and can hurt with two. McLeay and Tenreyro are where the group’s own work starts: the dominant-currency fact does not license the dominant-currency inference. Ask who invoices in dollars and it is homogeneous-goods exporters whose dollar prices are flexible, so low pass-through is an equilibrium outcome, not evidence of rigidity. Expenditure switching survives, and flexible exchange rates still stabilize.

Then the sanctions. Bianchi and Sosa-Padilla write Yellen’s worry down in eleven pages: anticipated confiscation of a foreigner’s dollar assets is a wedge on the convenience yield, and with a convenience yield and a costly supply of safe assets — only then — it raises the dollar return, shrinks holdings, weakens the dollar and lowers welfare for sanctioner and sanctioned alike. Itskhoki and Mukhin are the mirror from the target’s side: the ruble went 75 to 120 to 55 and there was no puzzle, because export and reserve sanctions depreciate while import sanctions appreciate, so the exchange rate is not a sufficient statistic for whether sanctions are working, and in 2022 it pointed the wrong way. Clayton, Dos Santos, Maggiori and Schreger give the challenger’s problem: China opened its bond market to central banks first and mutual funds last, and the mechanism runs opposite to the intuitive one — flighty investors are the creditors least worth expropriating, so admitting them costs most the government that means to renege. You liberalize because you are trusted. The 2022 talk’s story — China opened because it ran out of stable investors — is gone from the published paper, where the two investor types differ only in required pledgeability. The competition-among-reserve-providers question was dropped. Bahaj and Reis are its empirical companion: firms choose the currency of their working-capital credit and their invoicing together, so there is a threshold, and a swap line pushes marginal firms over it by cutting the right tail of borrowing costs, working ex ante even when little used, though “never drawn” overstates it. Signing a People’s Bank line raises the probability of any renminbi use by 11 to 14 points and cuts renminbi funding costs by 115 basis points. The 2020 talk identified this off Xi state visits; the published version drops the instrument entirely and the authors describe the 11.8-point estimate as an association between the policy and renminbi use, not a causal effect. Eichengreen, Mehl and Chiţu close with the prior behind all of it: in the one window where reserve composition is knowable, 1890–1913, a military alliance is worth about thirty points of an ally’s currency share, which is why de-dollarization debates are alliance debates. The working paper’s “thirty points” was one instrumental-variable coefficient; in print that estimate falls to 8 with a standard error of 6. The headline survives only as an average across four methods that disagree fourfold. That thirty is an average across methods disagreeing fourfold, on a small pre-1914 panel. A prior, not a presentation.

3.3 Geopolitics as a macro shock

Nobody in this sub-block threatens anybody, and that is the point of it. Caldara and Iacoviello’s index is the share of articles that talk about war the way people talk about war when they are worried about one — ten newspapers monthly and daily since 1985, three of them back to 1900 — an attention series audited against 44,000 front pages and a human reading of more than 7,000 articles, split into a threats component and an acts component. A two-standard-deviation shock lowers investment about 1.5 percent and hours 0.6 within a year, and the split delivers what Block 2 would want, both ways: a threat shock with acts held fixed is contractionary by nearly as much as the unconstrained one, an act shock with threats held fixed by less. Unexecuted threats are costly; whether they work is not something an index that cannot say who threatened whom can tell you, and the applied-finance literature that regresses everything on it was left off the list as a strand. The index stays because the block needs a shock series. Ambrosino, Chan and Tenreyro are here for the other reason, that the group’s own work sits here, and they treat fragmentation as exactly such a process, in a two-sector small open economy with sticky non-tradables and hand-to-mouth households, finding the central-bank chorus to be partial equilibrium: dearer imports raise marginal cost and make the country poorer, a poorer country spends less, and gradual fragmentation is a demand shock in supply-shock clothing — consumption, inflation and the natural rate all fall. Front-loaded fragmentation is stagflation instead, and the trade-off is specific: the natural rate does not move, so the bank opens a real-rate gap against an unchanged neutral rate. Neither paper is geoeconomics in the narrow sense, and the shock in the model is trade loss where Block 1 says what arrives in a connector economy is trade rerouting, which is the gap a contribution would fill.

3.4 Sanctions macro and open-economy DSGE

The list ends on the model one would extend. Ghironi, Kim and Ozhan put a commodity sector and three switchable sanctions — financial exclusion of some foreign households, trade bans binding on the most productive firms, commodity restrictions — into Ghironi–Melitz. The trade ban is the innovation: where a tariff truncates the exporter distribution from below, this clamps a cutoff on top of it and puts the best one percent of firms out of the export market. Inside one general equilibrium the answers stop being monotone. Sanctioning the target’s comparative-advantage sector is paid for by the sanctioner, 1.7 percent of consumption against the target’s 0.63, because it must move labour into a commodity sector it is bad at. Sanctioning the disadvantage sector depends on direction: the import ban is cheap for the sanctioner, the export ban is not. Partial financial sanctions leak because unsanctioned households intermediate for sanctioned ones, gaining 5.35 percent of consumption for their trouble; and steady-state comparisons get the split wrong, so welfare has to be computed along the transition. And only the export ban depreciates the sanctioner, the other three appreciate it, so the exchange rate is no measure of whether a sanction works. It is the open-economy DSGE the block closes on, and the frontier the published version names is not the CMS threat but optimal sanctions, fiscal effects, and central banks under nominal rigidity.

The threads, pulled together

Four threads run the length of the list. The first is one elasticity, two prices. Beyond Hulten says the network matters only where elasticities are far from one at unevenly exposed nodes and the shock is large; the German gas pair puts a number on it, a 20-percent cut to a one-percent input costing 20 percent of output at an elasticity of zero and 2.7 at 0.05. The same parameter reappears as the coercer’s leverage in the CMS statistic, as the inverse-elasticity weight in Liu and Yang, and as the convexity in Hausmann, Schetter and Yıldırım’s export-ban criterion. Every chokepoint claim in Block 2 is a second-order claim from Block 1, and every “cost of fragmentation” in Block 1 is the coercer’s price from Block 2, which is why the branches should cite each other and mostly do not.

The second is the off-path threat as the object of study. The GSDB records imposed sanctions and no threats at all, which is to say the threats that failed; Mulder’s interwar history is where the same problem reaches the block; Meyer and Wesseler’s sticks are cutoff rules never swung; Caldara and Iacoviello find that a threat never carried out shows up in investment nearly as much as one that is; and Geoeconomic Pressure is the dataset built to see the threat nobody imposed. A field whose central mechanism lives off the equilibrium path is a field whose data are, by construction, about the cases where the mechanism broke.

The third is trade as the cost of war, and de-risking as its subsidy. Fearon’s bargaining range, Martin–Mayer–Thoenig’s opportunity cost, Glick and Taylor’s decade of lost trade, the Fragmentation Paradox’s finding that lowering what you lose in a war makes war likelier, and Kooi’s resilience that is valuable because it improves the bargain — read together they say that the security case for friendshoring is a case for making conflict cheaper, and that the hostage was doing work nobody had put on the ledger. Note which half of Martin, Mayer and Thoenig it rests on: not the fragile finding that bilateral dependence deters, but the robust one that openness makes any single partner replaceable.

The fourth is dominance as a chokepoint, and what coercion costs the coercer. Gopinath–Stein and Mukhin explain why there is one dominant currency; Farrell and Newman explain what jurisdiction over its plumbing is worth; Bianchi–Sosa-Padilla and Crozet–Hinz price what using it costs the user, through the convenience yield and through trade finance; Pflueger and Yared show the privilege and the military funding each other in a loop that has flipped twice, Amsterdam to London to New York; and Bahaj–Reis and Internationalizing Like China describe a challenger crossing thresholds one firm and one central bank at a time. The dollar is the largest hub on Farrell and Newman’s map, and the papers that name it as a weapon are also the ones that put a depreciation schedule on it.

Which leaves the definitional joke the list opened with. The broad definition lets in everything from submarine cables to a small open economy with hand-to-mouth households; the narrow one insists on a coercer and a participation constraint; and the survey that supplies the broad definition closes its list of gaps by asking for a conceptualisation of economic coercion and a quantification of its rents: the narrow definition’s central object is not merely unmeasured but undefined. Sixty-one papers, three definitions, and the number the field is named for is still the one nobody has.