The short-run impact of US tariffs on an interconnected world

JORGE ALONSO-ORTIZ JOSÉ-MARÍA DA-ROCHA

Documento de Trabajo 2025/09

Agosto de 2025

fedea

Las opiniones recogidas en este documento son las de sus autores y no coinciden necesariamente con las de Fedea.

Jorge Alonso-Ortiz

José-María Da-Rocha

ITAM-CIE^†

Universidade de Vigo^‡

July 30, 2025

Abstract

We quantify the short-run impact of U.S. tarifs with a 77-country, 11-sector input–output model calibrated to OECD’s ICIOs. All intermediate inputs are locked in fixed Leontief proportions, so relative prices—not quantities—absorb the shock. A uniform 10% duty trims world GDP by 0.73% and U.S. GDP by 0.83%; ratcheting the schedule to 25% on NAFTA, to 15% on EU partners, and to 145% on China deepens those losses to 3.38% and 3.78%, respectively. Global trade flows in value, measured as exports plus imports over GDP, drop by 4.60% on average, but the U.S. is the biggest loser as its trade flows drop by -2.40% with a uniform duty, but drops can be as large as -11.60%.

^∗We thank A. de la Fuente for helpful comments.

^†Centro de Investigación Económica. Av. Camino Santa Teresa, 930 C.P., CDMx. 10700, Mexico. E-mail: jorge.alonso@itam.mx.

^‡Facultad de Comercio. Rúa do Conde de Torrecedeira, 105, 36208 Vigo, Spain. E-mail: jmrocha@uvigo.es.

1 Introduction

The World is experiencing a new wave of protectionism, initiated by the United States in 2018 during President Trump’s first term, following disputes over China’s alleged trade dumping and currency manipulation. This protectionist shift has intensified and broadened since the beginning of his non-consecutive second term, reaching levels not seen since the Smoot-Hawley Tarif Act of 1930. This paper aims to quantify the short-run efects of tarifs on real GDP in international dollars¹, and global trade flows in a world that is more interconnected than ever before, where retooling production lines is costly and time-consuming.

To this end, we develop a static computable general equilibrium model (CGE) of the global economy, comprising 76 countries and 11 sectors. The model is calibrated using Inter-Country Input-Output tables curated by the Organization for Economic Co-operation and Development (OECD). In our calibration, we abstract from diferences in factor intensities across sectors and countries, as well as from variation in capital-to-labor ratios. We assume perfect complementarity between intermediate goods. These simplifications allow us to isolate the short-run, first-order efects of tarifs on real quantities—specifically, GDP, and trade flow intensity.

Given that the United States accounted for 25% of the World’s GDP in 2019, its tarif policies have significant global efects that are worth examining in isolation to understand their magnitude before any retaliation, which yields a sort of upper-bound to the efect of tarifs on GDP and trade openness. Our assumption of a world with one bully also delivers a stark message to the American public. Even with no retaliation, the US economy sufers considerably more than most countries, including China, Canada, and Mexico.

¹We use the methodology of Geary (1958) and Khamis (1972) to compute purchasing power parities, and a brief description is included in Appendix 6.

We model five such scenarios. In all the scenarios, we are assuming that there is no retaliation. In the first one, the US imposes a 10% tarif on all goods from all countries. Then, it raises tarifs to 25%, targeting NAFTA partners, and China, and imposes a 15% tarif on the European Union that will be held fixed across the next scenarios. In the remaining three, the U.S. increases tarifs on Chinese goods to 54%, 125%, and 145%, respectively. These rates reflect tarifs that have either been agreed, enacted, or publicly proposed by the U.S. administration, and provide ready-to-use estimates of the short-run impacts of tarifs on GDP and global trade flows over GDP in a tightly interconnected world.

A uniform 10% tarif reduces the World’s GDP in international dollars by 0.73%. The losses are broadly distributed around the World, but the U.S. sufers a bigger GDP loss, measured in international dollars, of 0.83%, whereas Canada sufers losses of 0.67%, China 0.64%, Mexico 0.45%, and the average EU country sufers losses of 0.72%, with little variance. France loses 0.70%, Germany and Portugal 0.72%, and Spain loses 0.74% of its GDP, measured in international dollars. Everybody shares similar losses because fixed input-output linkages propagate price distortions well beyond the initially targeted bilateral trade flows.

Escalation to the extreme 145% tarif schedule significantly deepens the contraction: GDP falls by 3.38%, with the U.S. losing 3.78%, Canada 3.14%, China 2.93%, Mexico 2.40%, and the European Union 3.38%. The geographical dispersion of trade costs in terms of international GDP increases across their members, which highlights how rigid global supply chains amplify tarifs into first-order real shocks.

Global trade flows drop by 0.33% in the baseline 10% tarif only. Even in the worst case scenario, the trade openness index for the World, including the US, drops by 1.46%. These small, short-run drops in trade openness across scenarios hide a considerable redistribution across countries, where the U.S. turns out to be the biggest loser. In the baseline scenario, it loses 2.40% of trade flows, but ratcheting the schedule to 25% on NAFTA, to 15% on EU partners, and to 145% on China, increases the loss to 11.60%.

The emptiness left by the U.S. as a trade actor is partially filled by other countries that reroute trade value through the network. In particular, Mexico increases its trade flows by 2.84% in the baseline scenario, up to 7.80% if the U.S. imposed a 145% tarif on China. Therefore, Mexico absorbs two-thirds of the trade volume that the U.S. relinquishes, constituting one of the biggest shots to the foot of a country in recent economic history. The EU index barely changes, showing the resilience of a sub-network that is a trade and monetary union.

These results position our paper at the intersection of several strands of the trade literature. Early work in CGE models already showed that even a narrow tarif wedge reverberates through the whole economy: Johansen (1960) calibrated a multi-sector model for Norway, Harberger (1962) traced incidence across factors, and the applied CGE wave that followed—exemplified by Shoven and Whalley (1984) and the GTAP synthesis of Hertel (1997)—quantified the Kennedy- and Tokyo-round reforms. Historical episodes were dissected in the same spirit: Magee and Young (1987) revisited Smoot–Hawley in an Armington CGE, while Eichengreen and Irwin (2010) mapped the 1930’s beggar-thy-neighbor spiral.

Mitchener et al. (2022) constructs a monthly bilateral trade panel to study how retaliatory behavior impacted U.S. imports during the Smoot-Hawley war. Another strand revisits tarif incidence using scanner-level evidence, which shows near-unit pass-through into landed costs as in Amiti et al. (2019) or Cavallo et al. (2021), that use pricing information at the border, or (Bown and Kolb, 2023) that explore how firms use third-country detours (Bown and Kolb, 2023) to reduce tax incidence.

A new generation of Armington and Ricardian CGEs with flexible sourcing finds far smaller welfare losses once suppliers can be re-optimized—Fajgelbaum et al. (2020) and Caliendo and Parro (2015, 2021) cut U.S. losses to well under one-half of one percent—yet their equilibrium is explicitly long-run. We find larger upper bounds when we assume that countries and industries, are densely interconnected, which has been explored by yet another strand of the trade literature.

This network-based strand restores short-run rigidity: non-parametric formulas in Baqaee and Farhi (2020), or Baqaee and Farhi (2024), predict higher-order spillovers that plant-level data confirm as in (Atalay, 2017; Carvalho et al., 2021). Pandemic and prospective trade-war simulations by Bonadio et al. (2021) and Almazán-Gómez et al. (2025) reveal losses of an order of magnitude once input–output links are fixed.

Evidence on the network origins of business cycle asymmetries (Miranda-Pinto et al., 2023) and game-theoretic work on optimal retaliation (Fajgelbaum and Khandelwal, 2021) further motivate our dense bilateral IO framework with zero substitution, which provides an operational upper bound for the short-run cost of tarifs in terms of GDP, and complements the lower-bound estimates from elastic-sourcing models.

Taken together, our paper incorporates recent findings from trade literature and points to three insights: tarifs of even 10% have first-order real efects when contracts are rigid. Escalation multiplies GDP losses, although marginal damage tapers once critical links snap. Trade balances redistribute immediately, not without a net loss, but with a redistribution of the value of trade to countries that the US had in the spotlight. The remainder of the paper details the data, the model, the calibration, and the policy implications of these findings.

2 Data

Measuring the short-run efect of tarifs requires detailed information on transactions of goods and services across sectors and countries, including destinations. Inter-country input-output databases (ICIOs) serve that purpose. Still, only the OECD’s ICIO contains information for the most significant number of countries in 2019, right before the pandemic, but capturing the beginning of the trade dispute between China and the United States. We use the 2023 edition, a release that records bilateral flows of intermediate and final goods for 45 industries across 76 sovereign economies, plus a residual “Rest of the World” aggregate.

In its raw form, the matrix contains more than 150k non-zero cells. The sheer granularity would sink the numerical simulations of our model through near collinearity. Thus, we aggregate the data into 11 sectors that are listed in Appendix A so that it preserves the cross-country-sector fabric, while respecting the pragmatic limits of a static computable general equilibrium model.

After aggregation, each national block of the matrix spans 11 rows of inputs and 11 columns of outputs, while the global Leontief inverse—the object we ultimately feed to the model—bloats to 836 × 836 once all bilateral permutations are enumerated, which makes it still tractable. Every entry in ICIO is denominated in billions of current US dollars, which will allow us to compute magnitudes in international US dollars within the model.

A 2 × 2 example. To illustrate how we deal with 77 countries and 11 sectors, we shrink the economy into the U.S. and the rest of the World, and we collapse sectors into manufacturing and services. The US represents 25% of the worlds GDP in nominal $ US, and it is the instigator of this unfolding trade war, so it makes sense to track these two blocks to explain the structure of ICIO and national accounts, and the statistical imputations that need to done to make both congruent.

Table 1: A illustration.

Note: We aggregate the economy into a two country: the US and The World, and two sectors: manufacturing, into which we drop agriculture, and services. This way we can easily explain the imputations that need to be done to homogenize input-output information, from that of national accounts. Nevertheless, it is only a useful illustration of our more complicated model.

Table 1 contains information on intermediate use across countries and industries, final destination of goods, taxes, capital income, and labor income. The first upper left quadrant is the cross-intermediate use matrix, which will turn into the Leontief matrix in the calibration. Letting denote the monetary flow from sector i in country to sector j in country the ICIO provides

\[y _ {c j} = \sum_ {c ^ {\prime}, i} x _ {i j} ^ {c c ^ {\prime}} + T _ {c j} + V A B _ {c j},\]

so that dividing each intermediate entry by yields the Leontief coeficient . Once intermediate consumption is subtracted from output, we have GDP , which has to be equal to taxes on products net of subsidies T , capital income K net of depreciation, and labor income L.

The upper right quadrant of Table 1 captures the expenditure side of value added, produces the coeficients and that will allow us to characterize parameters in a household’s utility problem. Conceptually, we interpret these ratios as short-run contracts: they are locked in until physical retooling or renegotiation takes place. A version of this table for Canada, China, the European Union, Mexico, the U.S., and the Rest of the World is spliced in tables 9 and 10 in Appendix A.

Table 2: Imports and Exports % National Accounts

Reconciling national accounts. While the ICIO embeds capital formation and changes in inventories in its final-demand block, those series occasionally diverge from the System of National Accounts when statistical agencies revise investment upstream. We smooth the mismatch by proportionally scaling each country’s investment column so that C + I + G + XN matches GDP, where XN is the net-export vector derived from the bilateral matrix. In the United States, the adjustment is less than 0.3 percent of baseline GDP, whereas in Singapore—an archetypal entrepot economy—it reaches 1.7 percent. These adjustments are also needed so we can take the data to the model that we present in the

next section.

3 Model economy

There are C countries in the world. Each country c has J sectors that produce local varieties of goods that are traded globally. These goods are bought by all industries across all countries that are integrated into the global value chain, and by importers that combine local and foreign varieties into investment and final consumption goods. This section describes the basic features of the model and its equilibrium and relegates algebraic details to Appendix A.

Sectoral technology. We assume that there is a price-taking representative firm in each country that produces using a combination of varieties from diferent sectors and countries, capital , and labor with the technologies:

\[y _ {c j} = \min \left\{\left(\frac {x _ {i j} ^ {c c ^ {\prime}}}{a _ {i j} ^ {c c ^ {\prime}}}\right) _ {(c ^ {\prime}, i) \in C \times J}, \beta_ {c j} k _ {c j} ^ {\alpha_ {c j}} l _ {c j} ^ {1 - \alpha_ {c j}} \right\}\]

We assume that sector good producers use intermediate goods in fixed combinations, given by the inverse coeficients of the Leontief matrix . The combination of capital and labor generates part of the value added through a Cobb-Douglas technology with a country-specific sectoral productivity

These firms minimize the costs of renting capital and labor from the representative household and purchasing intermediate goods from diferent sectors around the world, at international prices for . Intermediate goods from abroad must pay a product and country-specific tarif . These firms pay a production tax net of subsidies, and operate under free entry. We demonstrate that this zero-profit condition enables us to establish an equilibrium relationship between local prices and international prices

\[(1 + t _ {c j}) p _ {c j} = \frac {1}{\beta_ {c j}} \left[ \left(\frac {r _ {c}}{\alpha_ {c j}}\right) ^ {\alpha_ {j}} \left(\frac {w _ {c}}{1 - \alpha_ {c j}}\right) ^ {1 - \alpha_ {c j}} \right] + \sum_ {i = 1} ^ {J} p _ {i} \sum_ {c ^ {\prime} = 1} ^ {C} (1 + \tau_ {c c ^ {\prime}} ^ {i}) a _ {i j} ^ {c c ^ {\prime}}\]

which is a matrix equation with an analytic solution as a function of country-specific unit production costs.

Final demand assemblers. We assume that two firms operate a technology that combines the same variety from diferent countries to produce a final consumption good

\[y _ {u j} ^ {c} = \min \left\{\left(\frac {x _ {u j} ^ {c c ^ {\prime}}}{a _ {u j} ^ {c c ^ {\prime}}}\right) _ {(c ^ {\prime}, j) \in C \times J} \right\}\]

or an investment good , also in fixed combination with intensity given by the inverse coeficients of the input-output matrix for final demand for . These firms are also price takers, operate under free entry, and must pay tarifs for imported goods. They sell a final variety of goods to the representative household at price , and an investment composite at price . We show that their zero-profit conditions allow us to write prices for the final consumption of good as:

\[p _ {f j} ^ {c} = p _ {j} \sum_ {c ^ {\prime} = 1} ^ {C} (1 + \tau_ {c c ^ {\prime}} ^ {j}) a _ {f j} ^ {c c ^ {\prime}}\]

and the final price of the investment good as

\[p _ {c I} = \sum_ {j = 1} ^ {J} p _ {j} \sum_ {c ^ {\prime} = 1} ^ {C} (1 + \tau_ {c c ^ {\prime}} ^ {j}) a _ {I j} ^ {c c ^ {\prime}}\]

Representative households. These final consumption goods for and investment goods are valued by consumers around the world with country-specific utility:

\[U _ {c} = \sum_ {j = 1} ^ {J} \theta_ {c j} \log (d _ {f j} ^ {c}) + \theta_ {c I} \log (d _ {c I})\]

These consumers maximize utility subject to a budget constraint.

\[\sum_ {j = 1} ^ {J} p _ {f j} ^ {c} (1 + \tau_ {f j} ^ {j}) d _ {c j} + p _ {c I} d _ {c I} = w _ {c} \bar {l} _ {c} + r _ {c} \bar {k} _ {c} + T _ {c}\]

where are value-added taxes. These consumers receive the proceeds from the collection of government taxes

Government. The final important element in our model is a country-specific government that collects value-added taxes , production taxes , and tarifs such that budget balances:

\[T _ {c} = \tau_ {c f} p _ {c f} d _ {c f} + \sum_ {j = 1} ^ {J} \left[ p _ {c j} t _ {c j} y _ {c j} + \sum_ {c ^ {\prime} = 1} ^ {C} \sum_ {i = 1} ^ {J} p _ {i} \tau_ {c c ^ {\prime}} ^ {i} x _ {i j} ^ {c c ^ {\prime}} \right]\]

These proceeds are rebated in lump sum to the representative consumer in each country. Therefore, three forces shape the equilibrium in our model: technological/contractual fixed trade relationships, tastes over consumption and investment goods, and trade policy that interact in a competitive general equilibrium, as we allow all markets to clear.

The world product markets are clear for all

\[\sum_ {c = 1} ^ {C} y _ {c j} = \sum_ {c = 1} ^ {C} \sum_ {c ^ {\prime} = 1} ^ {C} \left(x _ {f j} ^ {c c ^ {\prime}} + x _ {I j} ^ {c c ^ {\prime}} + \sum_ {i = 1} ^ {J} x _ {i j} ^ {c c ^ {\prime}}\right)\]

as well as local capital and labor markets . The final product markets are also cleared, as the local supply of the final products and

This is a large nonlinear problem of equations and unknowns; for 77 countries and 11 sectors, it turns into 1859 equations and unknowns. We describe the algorithmic solution in the Appendix A.

4 Calibration

Following Kehoe and Kehoe (1995), we calibrate the model so that, in equilibrium, the agents’ transactions in the model reproduce the baseline inter-country Input-Output tables (ICIO) for 11 industries and 76 countries. Our calibration pins down every parameter so that the model reproduces the 2019 benchmark economy exactly. Here we present a example that we were using for illustrative purposes.

Table 3: Calibrated coeficients

Table 3 presents a reduced version of the calibration inputs that we need for the 77 countries and 11 sectors in our model: the input-output coeficients, including those for final destination that are informative of utility weights, technological parameters, and taxes.

We begin with the technical coeficients of the Leontief matrix . Each dollar flow in the upper-left block of Table 1 is divided by the gross output of its destination sector.

For instance

\[a _ {\mathrm{SER} \leftarrow \mathrm{MAN}} ^ {\mathrm{US} \leftarrow \mathrm{RoW}} = \frac {x _ {\mathrm{SER} \leftarrow \mathrm{MAN}} ^ {\mathrm{US} \leftarrow \mathrm{RoW}}}{y _ {\mathrm{US,SER}}} = . 0 1 1 4\]

Similarly, we need to select the final demand shares of columns headed C, I, and normalized by their respective totals to give the consumption and investment coeficients that enter the household utility function θ. With prices normalized to one, utility weights follow directly from observed expenditures. Such as the appetite in the U.S. for their own services.

\[\theta_ {U S, S E R} ^ {U S} = 1 2 3. 3 / 1 8 9. 1 = 0. 6 7\]

We assume that capital shares are constant across sectors and countries and set to a third, which is a common value in the macro literature. We also assume that the endowment of capital relative to labor is the same across countries. This way we focus on the real efect of the price transmission of tarifs, instead of efects that arise in the presence of diferential factor intensities and endowments, contained in the classical and seminal work of Heckscher (1919), Ohlin (1933), Stolper and Samuelson (1941) or Rybczynski (1955).

Sectoral productivity across countries is obtained using standard growth accounting techniques, given , so that:

\[\beta_ {c j} = \frac {y _ {c j}}{k _ {c j} ^ {\alpha_ {c j}} l _ {c j} ^ {1 - \alpha_ {c j}}}\]

Finally, we use row of Table 9 over output to compute tax rates on products . Endowed with these parameters we can simulate our model to study several U.S. trade policies that we undertake in the next section.

5 Results

We simulate the efect of five diferent unilateral trade policies on GDP in international dollars and trade flows, measured by the ratio of nominal imports plus exports to nominal GDP. In the baseline scenario (I), the U.S. sets a 10% tarif on every country and good. Next (II), the U.S. diferentially raises tarifs to 25% on Canada, China, and Mexico, and imposes a 15% tarif to the European Union, that is held constant for the rest of the cases. In scenario (III) the U.S. imposes a 54% tarif to China, followed by an increase to (IV) 125% and (V) 145%. There is no retaliation as we want an upper bound of the cost of U.S. tarifs to the rest of the World, and to itself.

The efect of U.S. tarifs on GDP growth. Most papers that quantify U.S. tarif shocks work with real GDP in national prices or chain-weighted constant U.S. dollars, not international dollars. We can use the model to transform production in nominal US dollars into endogenously defined international prices, using the methodology described in (Geary, 1958) and (Khamis, 1972) to study how U.S. tarifs are transmitted through international prices to the rest of the world.

This twist seems to matter for policy decisions. Estimates from Yale’s Budget Lab set $100 billion as the yearly costs of Trump tarifs in 2024, or 0.3% of their GDP. Our simulations in the mildest scenario set the costs of tarifs to 0.83%, almost three times the level found in the Yale study, and in large CGE and Armington models such as Fajgelbaum et al. (2020), or the GTAP updates.

Table 4 illustrates that the U.S. 0.83% drop in the mildest scenario is larger than the World’s average loss, and certainly bigger than its targets in this trade war: NAFTA and EU partners, and China. They sufer smaller declines on the order of 0.45 to 0.67%. These fairly even losses reflect the fixed Leontief production structure, without scope for input substitution, a tarif-induced price wedge in any bilateral trade flow propagates

downstream and across borders.

Table 4: GDP growth (%)

Note: Each entry reports the percentage change in GDP in international dollars relative to a no-tarif baseline in 2019. All scenarios are unilateral U.S. import duties with no foreign retaliation. All tarifs are homogeneous across the board. (I) applies a uniform 10% tarif to all countries. (II) from (I) raises tarifs to 25% on Canada, Mexico, and 15% to the EU. (III)–(V) retain the structure of (II) while escalating the rate on Chinese goods to 54%, 125%, and 145%, respectively. EU countries are listed in Table 8 in the Appendix. Groups of countries are weighted by their relative size in nominal GDP in $US.

GDP losses in international dollars are fairly similar across European countries. France loses 0.70%, Germany (0.72%), Italy (0.75%), Portugal (0.72%), and Spain (0.74%), below the self inflicted harm that the U.S. imposes upon itself, but above the costs experienced by its NAFTA partners or China.

Table 5: Change in openness

Note: Each entry reports the percentage change in a trade openness index, defined as exports plus imports over GDP, all in nominal terms, relative to a no-tarif baseline in 2019. All scenarios are unilateral U.S. import duties with no foreign retaliation. All tarifs are homogeneous across the board. (I) applies a uniform 10% tarif to all countries. (II) from (I) raises tarifs to 25% on Canada, Mexico, and 15% to the EU. (III)–(V) retain the structure of (II) while escalating the rate on Chinese goods to 54%, 125%, and 145%, respectively. EU countries are listed in Table 8 in the Appendix.

Escalation to scenario V paints a much starker picture. Here, the U.S. maintains its 10% general duty, but imposes a 25% tarif on Canada and Mexico, 15% on European countries, and imposes a tarif of 145% on China. In such a case, the World’s GDP falls by 3.38%, more than four times the baseline loss. The U.S. now forfeits 3.78% of GDP, a considerably bigger fall than China 2.83%, Mexico 2.40%, or Canada 3.14%. This geographical fanning out of GDP losses in international dollars is displayed in Figure 1. As tarif raises for Canada, the European Union, Mexico, and particularly for China, color heterogeneity increases.

Fajgelbaum et al. (2020) and Caliendo and Parro (2015) find that the first Trump-era tarifs shaved at most 0.4% of U.S. real income, barely half the 0.82% percent contraction we measure under a uniform 15% percent tarif. The gap widens dramatically under escalation: our extreme scenario of a 3.78% U.S. GDP loss is almost four times larger than the 1 percent upper bound those elastic-sourcing models typically report, although previous works don’t compute GDP in international dollars, which lay closer to a measure of welfare change.

Non-linear amplification is equally striking. Our four- to five-fold jump in global GDP losses between the baseline and the 145% China tarif mirrors the convex responses predicted by the network formulae of Baqaee and Farhi (2020, 2024). Their analytical wedges imply that once a critical edge in the production graph is severed, the marginal damage of additional wedges decays—a pattern that Table 4 confirms and that Figure 1 visualizes for its spatial dispersion.

Micro-data studies of tarif incidence ofer a complementary lens. Near-unit pass-through at the border in Amiti et al. (2019) or Cavallo et al. (2021), and firm-level detours via third countries (Bown and Kolb, 2023) suggest that price wedges materialize quickly, yet firms start rerouting within months. Our short-run framework captures the immediate price shock but locks in intermediate inputs, thereby fixing an upper bound on the firstyear welfare hit. In this sense, the 3.38% global GDP decline under the harshest schedule should be interpreted as a worst-case snapshot before strategic sourcing or capacity investment can unfold.

The efect of U.S. tarifs on trade openness. Table 5 shows the efect of the diferent scenarios of U.S. trade policy on the ratio of exports plus imports over GDP, which is a textbook measure of trade openness. It has the advantage that it is a very intuitive measure of trade disengagement, which has been overlooked in previous studies.

Bleaney and Tian (2023) point that trade openness is sensitive to geographical factors, unrelated to trade policies, and exchange rate swings. Whereas we can not change geography, its natural limitation is contained in the fixed IO coeficients in the short run. Given that we use international dollars in our model, our measure of trade openness is thus neutral to exchange rate swings. With that in mind, we can ask how far the U.S. tarif wall pushes economies toward autarky, or makes them trade more.

Table 5 shows how the U.S. is shooting itself in the foot. Even a mild homogeneous tarif increase to 10%, as in scenario (I), reduces the volume of trade relative to GDP by 2.40%. However, the World’s trade openness only falls by 0.33%, and if we exclude the U.S., it increases. Therefore, other countries fill the emptiness left by the U.S. retreat from world trade through tarifs. Mexico increases its trade openness by 2.84%. China, despite being the primary foe in the U.S. policy narrative, barely sufers any change in trade openness, with a loss of only 0.08%. Trade openness in the EU barely changes as well, but there are winners and losers within the union.

As the U.S. escalates its tarif policy, it becomes more autarkic and poorer in the short run, and exacerbates the changes in the trade openness index. In the worst-case scenario, where the U.S. charges a 145% tarif on China, it loses 11.60% of it, and Mexico plays once more a pivotal role in the short-run rerouting of trade value, increasing its trade openness index by 7.80%.

The maps in Figure 2 reveal the short-run reordering of trade through imports and export flows for intermediate goods and final demand goods in the first scenario. The tarifs barely change the U.S. imports of intermediates, but the import of final goods drops substantially. Conversely the U.S. exports less intermediates and more final goods to the rest of the world. Spare parts are used by the rest of the world that buys more final-use products from the U.S..

Figure 3 shows the world redistribution of trade flows, as measured by the trade openness index, in the first scenario. The trade openness index falls by 2.40% in the U.S.. However, the rest of the world does not de-globalize; other countries fill the space left by the U.S. retreat. Instead, Mexico gains 2.84% in its trade openness index, and the rest of the world absorbs the remainder, so there is barely a change in the trade openness index of a world without the U.S. trade partner.

This pattern complements (Bleaney and Tian, 2023), warning that the headline openness ratio masks where transactions land. Openness changes little because fixed Leontief link ages keep total intermediate demand almost constant; what shifts is who clears the orders. The migration of flows to already-dense hubs echoes (Rodrik, 2000) insight that unilateral protection often enlarges third-country trade and aligns with the shock-propagation evidence of (Amiti et al., 2020).

Our simulations are also consistent with plant-level evidence from the T¯ohoku earthquake used in (Carvalho et al., 2021) and pandemic disruptions studied by (Bonadio et al., 2021), both of which document that network centrality—not headline openness—predicts the depth of local output shortfalls. Therefore, the evidence shows that when a central buyer withdraws, peripheral suppliers pivot to the next-best central node instead of shrinking outright.

In sum, the baseline and extreme scenarios underscore three lessons. First of all, tarifs have first-order real efects when sourcing is rigid: a uniform 10% duty sufices to shave almost three-quarters of a per cent of world income. Second, escalation multiplies those losses non-linearly—output drops quadruple and trade–balance swings widen—even though the last increments of tarif rates add diminishing marginal harm once key links have already snapped. Third, tarifs act like a redistributive valve—they divert containers to the nearest viable hub instead of choking them of altogether. It seems that the rest of the world can sustain similar trade flows relative to its size, without the U.S. as a trading partner in the short run.

6 Conclusions

We build and solve a short-run, multi-country input–output model in which every sector sources intermediate goods at fixed Leontief proportions, and quantities are expressed in international dollars. The framework nests 77 economies and 11 industries, capturing not only bilateral trade in final products but also the cross-border supply links that amplify relative-price shocks—an amplification channel that recent empirical work has highlighted as central to understanding the efect of tarifs.

Tarifs behave in today’s production network much less like a barrier, and much more like a valve that diverts value chains. In our baseline simulation, the United States erects a uniform 10% wall. Because all 77 economies are locked into fixed Leontief proportions, the price shock propagates instantly and trims world GDP by 0.73%; the U.S. itself loses 0.83%, while Mexico, Canada, and China record contractions between 0.45% and 0.67%. Escalating the schedule to 25% on NAFTA partners, 15% to EU countries, and 145% on China magnifies those numbers almost fourfold—global output falls by 3.38% and the U.S. output by 3.78%. The aggregate volume of trade falls by 1.46% across scenarios, yet the aggregate volume of trade does not decrease if we exclude the U.S. from the sample.

Instead, the geography of commerce is upended: Mexico’s openness index jumps by 7.80%, while the U.S. ratio plunges by 11.60%. The pattern echoes (Johansen, 1960) early multi-sector warning that even narrow wedges reverberate across an economy, and it corroborates more recent network formulations by (Baqaee and Farhi, 2019) that predict convex amplification once a critical edge is severed. Network centrality, rather than headline openness, dictates exposure.

These findings complicate the benign view that long-run Armington or Ricardian models ofer, because they make clear that the first year after a tarif hike is dominated by rigidity, not substitution. They also cut against the political argument that tarifs repatriate pro duction: the American economy emerges less open, poorer in real terms, and surrounded by neighbors that trade more, not less, despite showing a trade surplus.

Only once contracts expire, capital is reallocated, and firms experiment with new suppliers will the losses move toward the smaller numbers reported by elastic-sourcing CGEs. Until that adjustment unfolds, protectionism remains a costly gamble whose immediate dividend is a reshufling of container routes rather than a return of manufacturing jobs.

Future extensions that let coeficients drift, model retaliatory games, or incorporate intangible flows could trace the transition path from this upper-bound snapshot to longer-run equilibrium. For now, the evidence is unambiguous: in a densely wired world economy, even a seemingly moderate tarif shock leaves global trade volumes largely intact. Still, it hollows out the income of the country that fires the first shot. The lesson that emerges from our numbers, and from the historical record they echo, is that trade wars are not won; they are paid for.

References

Almazán-Gómez, Miguel Angel, Fatima El Khatabi, Carlos Llano, and Julián ^´ Pérez, “Modelling Regional Exposure to New Trade Wars,” Journal of Policy Modeling, 2025. forthcoming.

Amiti, Mary, Stephen J. Redding, and David E. Weinstein, “The Impact of the 2018 Trade War on U.S. Prices and Welfare,” Journal of Economic Perspectives, 2019, 33 (4), 187—-210. Earlier version: NBER Working Paper 25672 (2019).

, , and , “Who’s Paying for the US Tarifs? A Longer-Term Perspective,” AEA Papers and Proceedings, 2020, 110, 541–546.

Atalay, Enghin, “How Important Are Sectoral Shocks?,” American Economic Journal: Macroeconomics, 2017, 9 (4), 254–280.

Baqaee, David Rezza and Emmanuel Farhi, “The Macroeconomic Impact of Microeconomic Shocks: Beyond Hulten’s Theorem,” Econometrica, 2019, 87 (4), 1155–1203.

and , “Nonlinear Production Networks with Quasi-Linear Utilities with an Application to the COVID-19,” Working Paper 27281, National Bureau of Economic Research 2020.

and , “Networks, Barriers and Trade,” Working Paper, UCLA 2024.

Bleaney, Michael and Mo Tian, “The Trade–GDP Ratio as a Measure of Openness,” The World Economy, 2023, 46 (5), 1319–1332.

Bonadio, Barthelemy, Zhen Huo, Andrei A. Levchenko, and Nitya Pandalai-Nayar, “Global Supply Chains in the Pandemic,” Journal of International Economics, 2021, 133, 103534.

1. Bown, Chad P. and Yeling Kolb, “Supply-Chain Detours: How Global Firms Respond to Tarifs,” PIIE Working Paper 23-6, Peterson Institute for International Economics 2023.
2. Caliendo, Lorenzo and Fernando Parro, “Estimates of the Trade and Welfare Efects of NAFTA,” Review of Economic Studies, 2015, 82 (1), 1–44.
3. and , “Trade Policy, Structural Change and Welfare,” mimeo, Yale University 2021.
4. Carvalho, Vasco M., Makoto Nirei, Yukiko Saito, and Alireza Tahbaz-Salehi, “Supply Chain Disruptions: Evidence from the Great East Japan Earthquake,” Quarterly Journal of Economics, 2021, 136 (2), 1255–1321.
5. Cavallo, Alberto, Gita Gopinath, Brent Neiman, and Jenny Tang, “Tarif Pass-Through at the Border and at the Store,” American Economic Review: Insights, 2021, 3 (1), 19–34.
6. Eichengreen, Barry and Douglas A. Irwin, “The Slide to Protectionism in the Great Depression: Who Succumbed and Why?,” Journal of Economic History, 2010, 70 (4), 871–897.
7. Fajgelbaum, Pablo D. and Amit K. Khandelwal, “The Economic Impacts of the US-China Trade War,” Technical Report 29315, National Bureau of Economic Research 2021.
8. , Pinelopi K. Goldberg, Patrick J. Kennedy, and Amit K. Khandelwal, “Return to Protectionism,” Quarterly Journal of Economics, 2020, 135 (1), 1–55.
9. Geary, Roy C., “A Note on the Comparison of Exchange Rates and Purchasing Power between Countries,” Journal of the Royal Statistical Society. Series A (General), 1958, 121 (1), 97–99.

Harberger, Arnold C., “The Incidence of the Corporation Income Tax,” Journal of Political Economy, 1962, 70 (3), 215–240.

1. Heckscher, Oli F., “The Efect of Foreign Trade on the Distribution of Income,” Ekonomisk Tidskrift, 1919, 21 (4), 497–512. Reprinted as chapter 13 in AEA (1949) and in Flam & Flanders (1991).

Hertel, Thomas W., Global Trade Analysis: Modeling and Applications, Cambridge: Cambridge University Press, 1997.

1. Johansen, Leif, A Multi-Sectoral Study of Economic Growth, Amsterdam: North-Holland, 1960.

Kehoe, Patrick J. and Timothy J. Kehoe, “A Primer on Static Applied General Equilibrium Models,” in Patrick J. Kehoe and Timothy J. Kehoe, eds., Modeling North American Economic Integration, Dordrecht: Springer Netherlands, 1995, pp. 1–31.

1. Khamis, Salem Hanna, “A New System of Index Numbers for National and International Purposes,” Journal of the Royal Statistical Society. Series A (General), 1972, 135, 96–121.

Magee, Stephen P. and Leslie Young, “Endogenous Protection, Tarif Reform, and the U.S. Smoot-Hawley Tarif,” Journal of International Economics, 1987, 23 (1–2), 295–315.

1. Miranda-Pinto, Jorge, Alvaro Silva, and Eric R. Young, “Business cycle asymmetry and input-output structure: The role of firm-to-firm networks,” Journal of Monetary Economics, 2023, 137 (C), 1–20.

Mitchener, Kris James, Kevin H. O’Rourke, and Kirsten Wandschneider, “The Smoot–Hawley Trade War,” The Economic Journal, 2022, 132 (647), 2500–2533.

1. Ohlin, Bertil G., Interregional and International Trade Harvard Economic Studies, no. 40, Cambridge, MA: Harvard University Press, 1933. Ohlin credited Heckscher as co-developer of the model.
2. Rodrik, Dani, “How Far Will International Economic Integration Go?,” Journal of Economic Perspectives, 2000, 14 (1), 177–186.

Rybczynski, T. M., “Factor Endowment and Relative Commodity Prices,” Economica, 1955, 22 (88), 336–341.

1. Shoven, John B. and John Whalley, “Applied General-Equilibrium Models of Taxation and International Trade: An Introduction and Survey,” Journal of Economic Literature, 1984, 22 (3), 1007–1051.

Stolper, Wolfgang F. and Paul A. Samuelson, “Protection and Real Wages,” The Review of Economic Studies, November 1941, 9 (1), 58–73.

A Appendix

A.1. Data construction and code mappings

A.1.1 Transformation pipeline

The raw 2023 edition of the OECD ICIO provides ∼150 000 non–zero cells: 45 ISIC Rev.4 industries (ROW). To keep the numerical Leontief inversion tractable we compress those data into the 77 × 11 structure used in the model. Table 6 summarizes the workflow; the *.xlsx scripts mentioned are supplied with the replication material.

Table 6: Automated pipeline that converts raw ICIO blocks into the model matrices

A.1.2 Industry aggregation: the 11-sector key

Table 7 lists the correspondence between the 2-digit ISIC codes and the composite sectors (AGRI . . . OTHS) employed in the simulations. The same mapping drives all three Excel helpers cited above.

A.1.3 Regional aggregation: the EU-27 block

Following common practice in multi-regional IO work, all members of the European Union are collapsed into a single “EUR” economy. Table 8 lists the ISO-3 codes that are summed before

Table 7: Mapping from ISIC Rev.4 industries to the 11 composite sectors

the Leontief inversion; bilateral flows between those members disappear from the final dataset, but intra-EU trade is preserved in the EUR row/column.

Table 8: Countries aggregated into the EUR region

Any country not enumerated above—or absent from OECD ICIO—is absorbed into the residual ROW entity to keep the global transaction matrix square.

_erm^ediate ^transactions ^by ^origin ^and ^destinat

_onsumption _an^d ^investment ^by ^origin ^and ^destina

A.2. Model and solution algorithm

In this section, we provide details on the decision problems faced by participants in the global economy and how we aggregate them into the world’s competitive equilibrium.

Local producers problem. There are J local producers in each country that demand capital , labor and intermediate products from diferent sectors and countries, taking local cost of capital and labor and the international price of intermediates as given:

\[\min _ {\left\{x _ {i j} ^ {c c ^ {\prime}} \right\} _ {(c ^ {\prime}, i) \in S \times J}, k _ {c j}, l _ {c j}} r _ {c} k _ {c j} + w _ {c} l _ {c j} + \sum_ {c ^ {\prime} = 1} ^ {C} \sum_ {i = 1} ^ {J} p _ {i} (1 + \tau_ {c c ^ {\prime}} ^ {i}) x _ {i j} ^ {c c ^ {\prime}}\]

subject to:

\[\min \left\{\left(\frac {x _ {i j} ^ {c c ^ {\prime}}}{a _ {i j} ^ {c c ^ {\prime}}}\right) _ {(c ^ {\prime}, i) \in C \times J}, \beta_ {c j} k _ {c j} ^ {\alpha_ {c j}} l _ {c j} ^ {1 - \alpha_ {c j}} \right\} \geq y _ {c j}\]

The solution to this problem is a set of intermediate input demand functions:

\[x _ {i j} ^ {c c ^ {\prime}} = a _ {i j} ^ {c c ^ {\prime}} y _ {c j}\]

and demands for capital

\[k _ {c j} = \frac {1}{\beta_ {c j}} \left[ \frac {\alpha_ {c j} w _ {c}}{(1 - \alpha_ {c j}) r _ {c}} \right] ^ {1 - \alpha_ {c j}} y _ {c j}\]

and labor

\[l _ {c j} = \frac {1}{\beta_ {c j}} \left[ \frac {(1 - \alpha_ {c j}) r _ {c}}{\alpha_ {c j} w _ {c}} \right] ^ {\alpha_ {j}} y _ {c j}\]

These demands and those of intermediate inputs are homogeneous of degree 1 in industry production, so theoretically firms could be making profits in these sectors. We impose a zero profit condition in each sector, which allows us to define local prices as a function of international prices and the unit cost of production using capital and labor through the profit condition:

\[(1 + t _ {c j}) p _ {c j} = \frac {1}{\beta_ {c j}} \left[ \left(\frac {r _ {c}}{\alpha_ {c j}}\right) ^ {\alpha_ {j}} \left(\frac {w _ {c}}{1 - \alpha_ {c j}}\right) ^ {1 - \alpha_ {c j}} \right] + \sum_ {i = 1} ^ {J} p _ {i} \sum_ {c ^ {\prime} = 1} ^ {C} (1 + \tau_ {c c ^ {\prime}} ^ {i}) a _ {i j} ^ {c c ^ {\prime}}\]

Final consumption producers problem. There is a representative firm that buys products at international prices and supply final products to local consumers at final domestic prices to maximize profits:

\[\max \sum_ {j = 1} ^ {J} \left[ p _ {f j} ^ {c} y _ {f j} ^ {c} - \sum_ {c ^ {\prime} = 1} ^ {C} (1 + \tau_ {c c ^ {\prime}} ^ {j}) p _ {j} x _ {f j} ^ {c c ^ {\prime}} \right]\]

subject to:

\[y _ {f j} ^ {c} \leq \min \left\{\left(\frac {x _ {f j} ^ {c c ^ {\prime}}}{a _ {f j} ^ {c c ^ {\prime}}}\right) _ {c ^ {\prime} \in C} \right\}\]

Following the same zero profit principle, we can write the prices that the final consumer faces in each country, for each product, as:

\[p _ {f j} ^ {c} = p _ {j} \sum_ {c ^ {\prime} = 1} ^ {C} (1 + \tau_ {c c ^ {\prime}} ^ {j}) a _ {f j} ^ {c c ^ {\prime}}\]

Investment producers problem. There is a representative firm that buys products at international prices and transforms them into sectoral investment goods that have value to local consumers and buys them at final domestic prices to maximize profits:

\[\max \sum_ {j = 1} ^ {J} \left[ p _ {I j} ^ {c} I _ {c j} - \sum_ {c ^ {\prime} = 1} ^ {C} (1 + \tau_ {c c ^ {\prime}} ^ {j}) p _ {j} x _ {I j} ^ {c c ^ {\prime}} \right]\]

subject to:

\[I _ {c j} \leq \min \left\{\left(\frac {x _ {I j} ^ {c c ^ {\prime}}}{a _ {I j} ^ {c c ^ {\prime}}}\right) _ {c ^ {\prime} \in C} \right\}\]

Relying, once again, on a zero profit condition, we can write the price of investment in each sector as:

\[p _ {I j} ^ {c} = p _ {j} \sum_ {c ^ {\prime} = 1} ^ {C} (1 + \tau_ {c c ^ {\prime}} ^ {j}) a _ {I j} ^ {c c ^ {\prime}}\]

Representative household problem. There is a representative household in each country that values the demand of goods for final consumption and for investment

\[\max _ {d _ {f j} ^ {c}, d _ {I j} ^ {c}} \sum_ {j \in J} \left[ \theta_ {f j} \log (d _ {f j} ^ {c}) + \theta_ {I j} \log (d _ {I j} ^ {c}) \right]\]

subject to:

\[\sum_ {j \in J} \left[ p _ {f j} (1 + \tau_ {c f} ^ {j}) d _ {f j} ^ {c} + p _ {I j} d _ {I j} ^ {c} \right] = w _ {c} l _ {c} + r _ {c} k _ {c} + T _ {c}\]

The solution of this problem results in a set of demands for final consumption:

\[d _ {f j} ^ {c} = \frac {\theta_ {f j} (w _ {c} l _ {c} + r _ {c} k _ {c} + T _ {c})}{p _ {f j} ^ {c} (1 + \tau_ {c f} ^ {j})}\]

and investment:

\[d _ {I i} ^ {c} = \frac {\theta_ {I j} (w _ {c} l _ {c} + r _ {c} k _ {c} + T _ {c})}{p _ {I j}}\]

Government. There is a government that collect taxes on products , consumption , and tarifs , so that tax proceeds are rebated in full to the representative household in each country:

\[T _ {c} = \sum_ {j = 1} ^ {J} \left[ p _ {c j} t _ {c j} y _ {c j} + \tau_ {c f} ^ {j} p _ {f j} ^ {c} d _ {f j} ^ {c} + \sum_ {c ^ {\prime} = 1} ^ {C} \sum_ {i = 1} ^ {J} p _ {i} \tau_ {c c ^ {\prime}} ^ {i} x _ {i j} ^ {c c ^ {\prime}} \right]\]

Numerical solution

We solve the equilibrium by following the following steps:

(i) Make a guess of international prices , productions , trade balances and tax proceeds across countries

(ii) We compute using international price and production guesses in (i) and free entry conditions of local good producers

(iii) We compute final prices of final consumption, investment and foreign goods using the free entry conditions of the final goods assemblers

(iv) We obtain final demands by solving the household’s problems based on the guesses in (i)

(v) We compute tax proceeds across countries

(vi) We compute demand of intermediate goods derived from local producer’s minimization problem

(vii) We compute exports and imports using final demands from (iv) and intermediate demands from (vi)

(viii) This recursion can be summarized by a newton algorithm applied to the following set of equations

\[\sum_ {c = 1} ^ {C} y _ {c j} ^ {(0)} = \sum_ {c = 1} ^ {C} y _ {c j} \sum_ {c ^ {\prime} = 1} ^ {C} \sum_ {i = 1} ^ {J} a _ {i j} ^ {c c ^ {\prime}} + \sum_ {c = 1} ^ {C} y _ {c j} ^ {f} + \sum_ {c = 1} ^ {C} y _ {c j} ^ {I}\tag{1}\]

\[p _ {j} = p _ {j} ^ {(0)} \forall j \in J\tag{2}\]

\[\sum_ {j = 1} ^ {J} l _ {c j} (w _ {c}, r _ {c}) = \bar {l} _ {c} \text {and} \sum_ {j = 1} ^ {J} k _ {c j} (w _ {c}, r _ {c}) = \bar {k} _ {c} \forall c \in C\tag{3}\]

\[X _ {c} = X N _ {c} ^ {(0)} \forall c \in C\tag{4}\]

\[T _ {c} = T _ {c} ^ {(0)} \forall c \in C\tag{5}\]

A.3. Impact by region, income level and country

In the scenarios (I)-(V), international prices

\[p _ {j} = \sum_ {c = 1} ^ {C} (p _ {c j} / \mathrm{PPP} _ {c}) \frac {q _ {i j}}{\sum_ {j = 1} ^ {J} q _ {c j}}\]

are computed as the weighted average of international prices using given by

\[\mathrm{PPP} _ {c} = \frac {\sum_ {j = 1} ^ {J} p _ {c j} q _ {c j}}{\sum_ {j = 1} ^ {J} p _ {j} q _ {c j}}.\]

We use these international prices to compute GDP in Geary-Khamis dollars. Note that our calibration implies that international prices and are also equal to one in the baseline World.

Table 11: GDP growth by income group and region (%)

Table 12: Trade openness change by income group and region (%)

Table 13: Change in GDP in Geary-Khamis dollars (%)

ME. NA. Af & Pa = Middle East, North Africa, Afghanistan & Pakistan

Table 14: Change in GDP in Geary-Khamis dollars (%) (cont)

ME. NA. Af & Pa = Middle East, North Africa, Afghanistan & Pakistan

Table 15: Trade openness change (%)

ME. NA. Af & Pa = Middle East, North Africa, Afghanistan & Pakistan

Table 16: Trade openness change (%) (cont)

ME. NA. Af & Pa = Middle East, North Africa, Afghanistan & Pakistan