Strategic (Dis)Integration

John Sturm Becko (Princeton) and Daniel G. O’Connor (MIT). Working paper, first version March 2024; this digest uses the January 2025 draft (39 pages of main text, appendices to p. 115). The acknowledgements list NBER SI International Economics and Geopolitics among the venues where it was presented, but no recording could be found, so this is a PDF-only digest. Figure 6 is cropped from the PDF.

Preparing to use trade as a weapon

Every paper in block 2 has a hegemon making a threat. This one asks what the hegemon should do before it threatens — how to build domestic industry and foreign trading relationships in peacetime so that a future threat has more bite — and the answer comes out at an angle to the policies being pursued. A country that can credibly threaten trade taxes in a conflict should have no geopolitical industrial policy at all. Its peacetime trade policy should aim not at how much the rival gains from trade but at how manipulable the rival’s prices will be once the conflict starts, and in the paper’s calibration of the United States against China that prescription comes out as promoting trade with China on both the import and the export margin. The authors say Theorem 1 calls into question the geopolitical rationale for the CHIPS Act, and read the case where the hegemon cannot credibly threaten as consistent with the semiconductor and energy-independence subsidies. Reading de-risking the same way is my extension, not theirs — and it fits less well, since de-risking cuts trade and the model’s trade policy does no such thing.

The model

Two periods, peacetime and conflict, and two countries, Home and Foreign, each with a household, a goods producer and a capital producer exchanging many goods and many domestic capital varieties; capital is built and rented at home and does not cross borders. Capital is slow to adjust — the conflict-period stock is what was built in peace, which fits factories and stockpiles — so peacetime decisions determine the conflict-period economy. In the second period a geopolitical game runs first: Home announces trade taxes as a function of Foreign’s geopolitical action (respect Hong Kong’s autonomy, stop arming North Korea), possibly subject to a credibility constraint that Home will not impose taxes that cost itself more than some bound; Foreign picks its action weighing exogenous geopolitical benefits against the endogenous economic cost of the threatened taxes; then the economy trades under the realized taxes. Markets are efficient, and for the trade results and the quantification the planner weights Home and Foreign economic welfare equally (the weight is free in the model; the first theorem holds for any value of it), so any policy that emerges is geopolitical and not a disguised terms-of-trade or market-failure motive. In the deterministic model the on-path instrument is a carrot: Home never actually sanctions in equilibrium, and when the threat alone is not enough it subsidizes trade conditional on good behaviour, with the stick living off the equilibrium path. Home’s optimal threat is closely related to Becko’s (2024) optimal sanction; the paper’s subject is the peacetime policies that precede it.

Theorem 1: no industrial policy if your threats are credible

Private investment has a geopolitical externality — a stockpile at home changes what Foreign loses from a trade cut-off and so changes Foreign’s incentives — and the natural instinct is to subsidize or tax it. Theorem 1 says the optimal capital subsidy is zero. The argument is a revealed-preference trick. Home’s planner already chooses an on-path trade subsidy, and the first-order condition for that subsidy trades the geopolitical benefit of moving Foreign’s surplus against the deadweight loss of distorting trade; that condition reveals the shadow value Home places on a marginal improvement in Foreign’s incentives. A small change in Home’s stockpile has costs and benefits that are exactly proportional to those of a small change in the trade subsidy, so at the optimum its net value is zero. That is the targeting principle: Home’s only channel to Foreign’s action is trade, actual or threatened, and the instrument that acts on trade directly is already set correctly. A footnote adds that the capital subsidy also carries a second-order domestic misallocation cost, so it is not merely redundant but strictly worse.

Theorem 2: when threats are not credible, subsidize what pays in your worst case

If Home cannot credibly threaten taxes that would hurt itself beyond some limit — the EU’s decision not to sanction Russian gas is the example, citing Moll et al. — then industrial policy returns as a second-best instrument, because capital subsidies can now move trade in ways the constrained threat cannot. Theorem 2 gives its form: subsidies proportional to each capital variety’s rental rate in the state where Foreign takes the action that leaves Home economically worse off, $r_{H1v}\,s_{1v} \propto \tilde r_{H2v}(a_L)$. Two things follow. The subsidies are positive, since more capital raises Home’s welfare under any given trade and so loosens the credibility constraint. And in the typical case where Home is worse off punishing than rewarding, they fall on the capital that produces substitutes for Foreign’s exports — the goods that get scarce under sanctions. The authors find this consistent with the strategic petroleum reserve and the semiconductor plant subsidies. The subsidies buy credibility, not leverage.

Theorem 3: peacetime trade policy targets manipulability, not welfare

Trade taxes in peacetime move Foreign’s capital stock, and the question is which movements Home wants. Not, it turns out, the ones that raise the gap in Foreign’s welfare across its actions — that gap is better targeted by the second-period threat. Theorem 3 (eq. 13 in the paper) says the peacetime tax on good $g$ satisfies, up to Lerner symmetry,

where $\tau^{ToT}_{1g}$ is the conventional terms-of-trade tax (zero when Home is indifferent to redistribution in both periods), $\kappa$ is the multiplier on Foreign’s incentive constraint, $\tilde p_{F2}$ is Foreign’s inverse net-export supply curve, the integral runs from Foreign’s conflict-period trade under the bad action to its trade under the good one, and the last factor is how peacetime trade in $g$ moves Foreign’s capital. The object in the middle is a second derivative of Foreign’s prices — not how elastic Foreign’s demand is, which is what Dixit’s or Becko’s optimal tariffs care about, but how much the capital induced by peacetime trade changes that elasticity along the path between the off-path and on-path conflict allocations. In words: Home uses peacetime trade to make Foreign’s terms of trade more manipulable in conflict. The two-good picture behind this decomposes any shift in Foreign’s export supply curve into a uniform shift plus a rotation around the final price-quantity point. The uniform shift raises Foreign’s gains from trade but offers exactly the cost-benefit trade-off the second-period threat already exploits, so it is worth nothing at the margin; the rotation costs nothing and raises the gap in Foreign’s surplus between aggression and restraint, so it is pure gain. Hence the striking case where Home does nothing: if Foreign is a gas exporter whose pipeline uniformly lowers its export costs, subsidizing that trade raises Foreign’s stake in the relationship, yet Home declines, because pipelines do not make Foreign’s export supply less elastic. Note that the theorem pins the target of peacetime trade policy, not the sign; whether Home ends up subsidizing or taxing trade with the rival depends on which way induced investment bends Foreign’s demand.

The contrast the paper draws is clean. Industrial policy, where it is used, reduces Home’s dependence on Foreign to make Home’s threats credible; trade policy increases Foreign’s dependence on Home to make Home’s threats effective. And Proposition 1 says the trade formula survives constraints on the threat, provided the set of trade quantities Home can credibly threaten depends only on Home capital (which covers the credibility constraint of Theorem 2): such constraints affect peacetime trade taxes only indirectly, through which off-path quantities Home can threaten, and not through the formula itself. The reason is again targeting — Home capital affects credibility directly, while Foreign capital affects it only through second-period trade, which the threat targets on its own.

The United States and China, calibrated

The last part specializes the many-country extension to a multi-sector model with intermediates and two kinds of capital — sector-specific production capital that lowers variable cost, and sector-by-origin relationship capital that lowers the cost of sourcing from a particular origin — calibrated to 2017 trade, tariffs, value added and IO tables for 39 countries and 26 sectors, with Boehm et al.’s short-run and long-run trade elasticities (1.25 and 2) and Ding’s investment composition. US threats are assumed fully credible, so there is no industrial policy and the exercise is peacetime trade policy only. Two caveats shape how to read it. The geopolitical block is calibrated only through the limit in which the incentive multiplier goes to zero, so every on-path tax also goes to zero and the exercise pins the relative sizes of taxes across goods and partners, not their levels — Figure 6 normalizes the largest tax to one. And the off-path punishment is extreme: because the US can subsidize trade with bystanders during conflict, its threat can in principle put China in autarky, and the authors flag the case of partly credible threats as future work.

 and export subsidies (right) by partner country and sector, normalized so the largest tax equals one; the exercise pins relative sizes only. Blue means trade promotion in both panels, but the colourbars run opposite ways: a blue import cell is a negative tariff (−1 at the bottom of the left bar) and a blue export cell is a positive subsidy (+1 at the top of the right bar). The China row is predominantly blue on both margins, with a few orange investment-sector cells on the export side. Gray: origin-sectors with no exports to the US.’)

The US’s optimal policy promotes bilateral trade with China on both margins. Import subsidies are close to uniform, from 40 to 58 percent of the largest tax; export subsidies range from a 52 percent tax to the full 100 percent subsidy, and the variation is almost entirely the share of Chinese expenditure that goes, directly or through intermediates, to investment goods. Consumption sectors like food and services are subsidized; investment sectors like construction and minerals are positively taxed, not merely subsidized less, because Chinese purchases from the US there crowd out Chinese imports from third countries in sectors where China’s elasticities are more manipulable. The mechanism is relationship capital: subsidized trade shifts China’s sourcing investments toward foreign partners and away from domestic relationships, which makes its conflict-period import demand less elastic. The paper’s Shapley decomposition has three components — manipulability through China’s relationship capital, manipulability through its production capital, and the many-country adjustment — and the relationship channel alone explains 111 percent of the bilateral taxes, with the other two of little quantitative importance. (There is no terms-of-trade component to compare against; the equal-weights assumption switches it off.) Taxes on third countries average about 4.3 percent of the China taxes in magnitude and follow a pattern: the US makes the goods of countries that run bilateral deficits with China, Mexico above all, scarce — import subsidy plus export tax — and the goods of surplus countries like Taiwan abundant, because scarcity appreciates a country’s goods and pushes it to import more, which for a deficit country means more trade with China. The same logic makes textiles and electrical equipment, where China’s share of world output is large, scarce. Everything serves the one end of Chinese relationship capital pointed outward.

Where it sits

Read it right after Coercion and Fragmentation and Kooi. CMS’s defence paper has each target insulating itself against a hegemon’s threat and finds over-fragmentation; Kooi has a country buying resilience because bargaining rewards it, and the authors note in a footnote that Kooi’s capital subsidies look like their own limited-credibility case; Becko–O’Connor have the hegemon deciding how integrated to be. Their general answer is that trade policy should target the manipulability of the rival’s prices and that industrial policy is unnecessary as long as the threats are credible; when they are not, industrial policy comes back, in a form the authors find consistent with the policies we observe, while the trade prescription is unchanged. In the US–China calibration the trade prescription is deeper integration with the rival, at least in relative terms. It is also the paper that makes the block’s recurring variable, dependence, precise: what the coercer wants is not that the target gains a lot from trade, but that the target’s demand is inelastic when the trade is cut. Liu–Yang’s elasticity-weighted measure is the empirical shadow of that statement.